WO2004086972A9 - Methods for the compensation of imaging technique in the processing of radiographic images - Google Patents

Methods for the compensation of imaging technique in the processing of radiographic images

Info

Publication number
WO2004086972A9
WO2004086972A9 PCT/US2004/009165 US2004009165W WO2004086972A9 WO 2004086972 A9 WO2004086972 A9 WO 2004086972A9 US 2004009165 W US2004009165 W US 2004009165W WO 2004086972 A9 WO2004086972 A9 WO 2004086972A9
Authority
WO
WIPO (PCT)
Prior art keywords
bone
image
ray
fracture
parameters
Prior art date
Application number
PCT/US2004/009165
Other languages
French (fr)
Other versions
WO2004086972A2 (en
WO2004086972A3 (en
Inventor
Philipp Lang
Daniel Steines
Siau-Way Liew
Rene Vargas-Voracek
Original Assignee
Imaging Therapeutics Inc
Philipp Lang
Daniel Steines
Siau-Way Liew
Rene Vargas-Voracek
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Imaging Therapeutics Inc, Philipp Lang, Daniel Steines, Siau-Way Liew, Rene Vargas-Voracek filed Critical Imaging Therapeutics Inc
Priority to CA002519187A priority Critical patent/CA2519187A1/en
Priority to EP04758337A priority patent/EP1605824A2/en
Priority to JP2006509289A priority patent/JP2007524438A/en
Publication of WO2004086972A2 publication Critical patent/WO2004086972A2/en
Publication of WO2004086972A9 publication Critical patent/WO2004086972A9/en
Publication of WO2004086972A3 publication Critical patent/WO2004086972A3/en

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/0002Inspection of images, e.g. flaw detection
    • G06T7/0012Biomedical image inspection
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/45For evaluating or diagnosing the musculoskeletal system or teeth
    • A61B5/4504Bones
    • A61B5/4509Bone density determination
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00Measuring for diagnostic purposes; Identification of persons
    • A61B5/74Details of notification to user or communication with user or patient ; user input means
    • A61B5/7475User input or interface means, e.g. keyboard, pointing device, joystick
    • A61B5/748Selection of a region of interest, e.g. using a graphics tablet
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00Apparatus for radiation diagnosis, e.g. combined with radiation therapy equipment
    • A61B6/44Constructional features of apparatus for radiation diagnosis
    • A61B6/4423Constructional features of apparatus for radiation diagnosis related to hygiene or sterilisation
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00Apparatus for radiation diagnosis, e.g. combined with radiation therapy equipment
    • A61B6/46Apparatus for radiation diagnosis, e.g. combined with radiation therapy equipment with special arrangements for interfacing with the operator or the patient
    • A61B6/467Apparatus for radiation diagnosis, e.g. combined with radiation therapy equipment with special arrangements for interfacing with the operator or the patient characterised by special input means
    • A61B6/469Apparatus for radiation diagnosis, e.g. combined with radiation therapy equipment with special arrangements for interfacing with the operator or the patient characterised by special input means for selecting a region of interest [ROI]
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00Apparatus for radiation diagnosis, e.g. combined with radiation therapy equipment
    • A61B6/50Clinical applications
    • A61B6/505Clinical applications involving diagnosis of bone
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00Apparatus for radiation diagnosis, e.g. combined with radiation therapy equipment
    • A61B6/58Testing, adjusting or calibrating apparatus or devices for radiation diagnosis
    • A61B6/582Calibration
    • A61B6/583Calibration using calibration phantoms
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/10Segmentation; Edge detection
    • G06T7/11Region-based segmentation
    • GPHYSICS
    • G09EDUCATION; CRYPTOGRAPHY; DISPLAY; ADVERTISING; SEALS
    • G09BEDUCATIONAL OR DEMONSTRATION APPLIANCES; APPLIANCES FOR TEACHING, OR COMMUNICATING WITH, THE BLIND, DEAF OR MUTE; MODELS; PLANETARIA; GLOBES; MAPS; DIAGRAMS
    • G09B23/00Models for scientific, medical, or mathematical purposes, e.g. full-sized devices for demonstration purposes
    • G09B23/28Models for scientific, medical, or mathematical purposes, e.g. full-sized devices for demonstration purposes for medicine
    • G09B23/30Anatomical models
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B2576/00Medical imaging apparatus involving image processing or analysis
    • A61B2576/02Medical imaging apparatus involving image processing or analysis specially adapted for a particular organ or body part
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B6/00Apparatus for radiation diagnosis, e.g. combined with radiation therapy equipment
    • A61B6/50Clinical applications
    • A61B6/508Clinical applications for non-human patients
    • AHUMAN NECESSITIES
    • A61MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61BDIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B8/00Diagnosis using ultrasonic, sonic or infrasonic waves
    • A61B8/08Detecting organic movements or changes, e.g. tumours, cysts, swellings
    • A61B8/0875Detecting organic movements or changes, e.g. tumours, cysts, swellings for diagnosis of bone
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/30Subject of image; Context of image processing
    • G06T2207/30004Biomedical image processing
    • G06T2207/30008Bone

Definitions

  • the present invention is in the field of imaging and analysis thereof.
  • methods and compositions for accurately analyzing images to determine bone mineral density and/or bone structure are described.
  • Osteoporosis is a condition that affects millions of Americans.
  • Osteoporosis refers to a condition characterized by low bone mass and microarchitectural deterioration of bone tissue, with a consequent increase of bone fragility and susceptibility to fracture. Osteoporosis presents commonly with vertebral fractures or hip fractures due to the decrease in bone mineral density and deterioration of structural properties and microarchitecture of bone.
  • Imaging techniques are important diagnostic tools, particularly for bone related conditions.
  • DXA dual x-ray absorptiometry
  • QCT quantitative computed tomography
  • pDXA peripheral DXA
  • pQCT peripheral QCT
  • DXA of the spine and hip has established itself as the most widely used method of measuring B D. Tothill, P. and D.W. Pye, (1992) BrJ Radiol 65:807-813.
  • the fundamental principle behind DXA is the measurement of the transmission through the body of x-rays of 2 different photon energy levels. Because of the dependence of the attenuation coefficient on the atomic number and photon energy, measurement of the transmission factors at 2 energy levels enables the area densities (i.e., the mass per unit projected area) of 2 different types of tissue to be inferred. In DXA scans, these are taken to be bone mineral (hydroxyapatite) and soft tissue, respectively.
  • Quantitative computed tomography is usually applied to measure the trabecular bone in the vertebral bodies.
  • QCT studies are generally performed using a single kV setting (single-energy QCT), when the principal source of error is the variable composition of the bone marrow.
  • a dual-kV scan dual-energy QCT is also possible. This reduces the accuracy errors but at the price of poorer precision and higher radiation dose.
  • QCT are very expensive and the use of such equipment is currently limited to few research centers.
  • Quantitative ultrasound is a technique for measuring the peripheral skeleton. Njeh et al. (1997) Osteoporosis Int 7:7-22; Njeh et al. Quantitative Ultrasound: Assessment of Osteoporosis and Bone Status. 1999, London, England: Martin Dunitz. There is a wide variety of equipment available, with most devices using the heel as the measurement site. A sonographic pulse passing through bone is strongly attenuated as the signal is scattered and absorbed by trabeculae. Attenuation increases linearly with frequency, and the slope of the relationship is referred to as broadband ultrasonic attenuation (BUA; units: dB/MHz).
  • BOA broadband ultrasonic attenuation
  • BUA is reduced in patients with osteoporosis because there are fewer trabeculae in the calcaneus to attenuate the signal.
  • most QUS systems also measure the speed of sound (SOS) in the heel by dividing the distance between the sonographic transducers by the propagation time (units: m/s). SOS values are reduced in patients with osteoporosis because with the loss of mineralized bone, the elastic modulus of the bone is decreased.
  • SOS speed of sound
  • Radiographic absorptiometry is a technique that was developed many years ago for assessing bone density in the hand, but the technique has recently attracted renewed interest. Gluer et al. (1997) Semin Nucl Med 27:229-247. With this technique, BMD is measured in the phalanges.
  • the principal disadvantage of RA of the hand is the relative lack of high turnover trabecular bone. For this reason, RA of the hand has limited sensitivity in detecting osteoporosis and is not very useful for monitoring therapy-induced changes.
  • Peripheral x-ray absorptiometry methods such as those described above are substantially cheaper than DXA and QCT with system prices ranging between $15,000 and $35,000.
  • epidemiologic studies have shown that the discriminatory ability of peripheral BMD measurements to predict spine and hip fractures is lower than when spine and hip BMD measurements are used. Cummings et al. (1993) Lancet 341:72- 75; Marshall et al. (1996) BrMedJ 312:1254-1259. The main reason for this is the lack of trabecular bone at the measurement sites used with these techniques.
  • the disclosure provides a method to derive information regarding one or more bone parameters from an image, the method comprising the steps of: (a) obtaining an image comprising bone from a subject; (b) defining two or more regions of interest (ROIs) in the image; and (c) analyzing a plurality of positions in the ROIs to determine one or more parameters selected from the group consisting of bone microarchitecture, bone macro-anatomy, biomechanical parameters and combinations thereof of the ROIs.
  • the ROIs are overlapping.
  • the positions analyzed in the ROIs may be at regular intervals relative to one another or, alternatively, may be irregularly spaced relative to each other.
  • the methods involve determining bone micro-architecture, for example by analyzing positions at regular intervals.
  • the methods involve determining bone macro-anatomy, for example by analyzing positions at irregular intervals in the image.
  • the image can be two- dimensional (2D) or three-dimensional (3D).
  • the images may be x-rays, MRI images, CAT scan images, or any other image including bone.
  • the image may be an electronic image.
  • the subject can be, for example, an osteoporosis subject.
  • this disclosure relates to a method of generating a map of one or more bone parameters, the method comprising the steps of (a) obtaining information on bone parameters according to the method of any of methods described herein; and (b) identifying regions of the image that exhibit similar parameter characteristics, thereby creating a parameter map of the image.
  • a method of predicting a fracture path in a subject comprising the steps of: (a) generating multiple parameter maps according to any of the methods of generating parameters maps described herein; (b) generating a composite parameter map from the multiple parameters maps of step (a); and (c) analyzing the composite parameter map to identify possible fracture paths.
  • the invention includes a method of predicting a fracture path in a subject, the method comprising the steps of: (a) analyzing of one or more parameter maps preparing according to any of the methods described herein, wherein the analysis is watershed segmentation analysis or Markov random field analysis; and (c) identifying possible fracture paths based on the analysis of step (a), thereby predicting a fracture path in the subject.
  • the invention includes a method of predicting the risk of fracture in a subject, the method comprising the steps of: (a) generating a finite element model from one or more parameter maps obtained according any of the methods described herein; (b) applying simulated force vectors that would occur during a fracture incident to the finite element model generated in step(s); and (c) determining the minimum forces required for fracture to occur, thereby estimating the risk of fracture.
  • the invention includes a method of determining the risk of fracture in a subject comprising: (a) predicting a fracture path according to any of the methods of predicting fracture path as described herein; (b) evaluating one or more selected bone parameters along the predicted fracture path, thereby estimating the risk of fracture.
  • the invention includes a method of treating a subject with bone disease comprising (a) obtaining an image from a subject; (b) analyzing the image obtained in step (a) using any of the methods described herein; (c) diagnosing a bone disease based on the analysis of step (b); and (d) selecting and administering a suitable treatment to said subject based on said diagnosis.
  • FIG. 1 shows an example of a dental x-ray.
  • a calibration phantom 110 is seen.
  • Regions of interest 120 have been placed for measurement of bone mineral density or structure.
  • FIG. 2 shows another example of a dental x-ray.
  • a calibration phantom A calibration phantom
  • Regions of interest 120 have been placed for measurement of bone mineral density or structure.
  • FIG. 3 shows an example of an analysis report resulting from a measurement of mandibular or maxillary bone mineral density.
  • a subject (X) is more than one standard deviation below the mean of age-matched controls (x-axis age, y- axis arbitrary units BMD).
  • FIG. 4 shows an example of a V-shaped calibration phantom 110 mounted on a tooth 120. Gums are also shown 130.
  • FIG. 5 shows an example of a holder 115 for a calibration phantom 110.
  • the holder 115 is mounted on a tooth 120. Gums are also shown 130.
  • FIG. 6 panels B through E shows gray value profiles along different rows of pixels used for locating dental apices. From top to bottom, the characteristic peaks for the dental roots (shown in dental x-ray panel A) gradually disappear.
  • FIG. 7 shows a Hough transform (panel A) of a test image (panel B). All collinear points from the same line are transformed into sinusoidal curves that intersect in a single point (circles).
  • FIG. 8 shows a Hough transform (panel A) of a skeletonized trabecular bone x-ray image (panel B).
  • panel A a skeletonized trabecular bone x-ray image
  • panel B a skeletonized trabecular bone x-ray image
  • FIG. 9 shows the effect of varying size of structuring element E 2 ; calibration phantom image with lines of varying width (1 , 3, 5, 7, 9, 11 , 13 pix) (top left); skeleton operation performed using E 2 with a diameter of 3 pix (top right), 7 pix (bottom left), and 11 pix (bottom right), respectively.
  • FIG. 10 shows the effect of varying size of structuring element E ⁇ ; gray scale image of trabecular bone (top left, panel A); skeleton operation performed using E 2 with a diameter of 3 pix (top right, panel B); 7 pix (bottom left, panel C) and 11 pix (bottom right, panel D), respectively.
  • FIG. 11 shows gray value surface plot of an anatomical region of interest from a dental x-ray (inset) used for fractal analysis.
  • FIG. 12 shows an example of a hygienic cover holder that includes compartments for a calibration phantom and a fluid-filled bolus back.
  • FIG. 13 shows an example of an anatomical region of interest (black dot), determined relative to the teeth or to the convexity/concavity of the mandible.
  • FIG. 14 shows an example of three anatomical region of interests (black dots), determined relative to the teeth or to the convexity/concavity of the mandible.
  • Fig. 15 is a side view of an exemplary system for minimizing tube angulation as described herein.
  • the system is shown as a dental x-ray system.
  • An extension tubing (200) is attached to a ring-shaped Rinn holder (102).
  • the outer diameter of the extension tubing is slightly smaller than the inner diameter of the tube located in front of the dental x-ray system/dental x-ray tube.
  • the extension tubing can then be inserted into the metal tube thereby reducing tube angulation and resultant errors in bone apparent density and bone structural measurements.
  • FIG. 16 depicts an example of a regular interval sampling field for microarchitecture (+) and a higher density sampling field for macro-anatomical features(*) on a femur radiograph.
  • White rectangles are examples of overlapping window positioning.
  • FIG. 17 depicts watershed segmentation boundaries superimposed on a parameter map.
  • the two white lines are the actual fracture paths resulted from an in- vitro mechanical loading test.
  • FIG. 1 is a flowchart depicting an exemplary process to determine fracture risk using overlapping window processing and fracture paths prediction.
  • FIG. 19 depicts a Markov random field analysis by modeling particular joint feature distributions as they are estimated at each image element or image neighborhood.
  • FIG. 20 depicts an exemplary model definition for trabecular pattern density characterization in a region of interest (ROI) with a noise model P(N) and characteristic structure pattern given a density level P(l
  • ROI region of interest
  • FIG. 21 depicts exemplary Bayes' Rule analysis.
  • FIG. 22 depicts an example of a regular interval sampling field for microarchitecture (+) and a higher density sampling field for macro-anatomical features (*) on a spine radiograph. White rectangles are examples of overlapping window positioning.
  • FIG. 23 depicts an example of a sampling field of varying density for microarchitecture (+, x, diamond) and a regular sampling field for macro-anatomical features( * ) on a knee radiograph. White rectangles are examples of overlapping window positioning.
  • FIG. 24 depicts an example of an application of structure extraction and measurement for therapeutic monitoring using spine x-ray.
  • White outline of extracted structure are show in (a) before treatment, and (b) after treatment.
  • the invention includes methods of obtaining and/or deriving information about bone mineral density and/or bone structure from an image. Additionally, the present invention relates to the provision of accurate calibration phantoms for use in determining bone structure and methods of using these calibration phantoms. In particular, the present invention recognizes for the first time that errors arising from misplacement of interrogation sites in dental or hip x-rays of bone density and/or bone structure can be corrected by positioning the x-ray tube, the detector and/or the calibration reference with respect to an anatomical landmark (or anatomical region of interest).
  • Advantages of the present invention include, but are not limited to, (i) providing accessible and reliable means for analyzing x-rays; (ii) providing non-invasive measurements of bone structure and architecture and macro-anatomy; (iii) providing methods of diagnosing bone conditions (e.g., osteoporosis, fracture risk); (iv) providing methods of treating bone conditions; and (iv) providing these methods in cost-effective manner.
  • An image can be acquired using well-known techniques from any local site.
  • imaging techniques suitable for acquiring images from which data can be obtained include, ultrasound, CAT scan, MRI and the like. See, also, “Primer of Diagnostic Imaging,” 3rd edition, eds. Weissleder et al. (2002), Mosby Press; and International Publication WO 02/22014.
  • 2D planar x-ray imaging techniques are used.
  • 2D planar x-ray imaging is a method that generates an image by transmitting an x-ray beam through a body or structure or material and by measuring the x-ray attenuation on the other side of said body or said structure or said material.
  • 2D planar x-ray imaging is distinguishable from cross-sectional imaging techniques such as computed tomography or magnetic resonance imaging. If the x-ray image was captured using conventional x- ray film, the x-ray can be digitized using any suitable scanning device. Digitized x-ray images can be transmitted over a networked system, e.g. the Internet, into a remote computer or server.
  • x-ray images can also be acquired using digital acquisition techniques, e.g. using photostimulable phosphor detector systems or selenium or silicon detector systems, the x-ray image information is already available in digital format which can be easily transmitted over a network.
  • 3D images are acquired, for example, using 3D imaging techniques and/or by creating 3D images from 2D images.
  • any images can be used including, but not limited to, digital x-rays and conventional x-ray film (which can be digitized using commercially available flatbed scanners).
  • the x-ray is of the hip region, for example performed using standard digital x-ray equipment (Kodak DirectView DR 9000, Kodak, Rochester, NY). Patients are typically positioned on an x-ray table in supine position, parallel to the long axis of the table, with their arms alongside their body.
  • the subject's feet may be placed in neutral position with the toes pointing up or in internal rotation or may be placed in a foot holder such that the foot in a neutral position (0° rotation) or in any desired angle of rotation (e.g., internal or external) relative to neutral (see, also Example 8 below).
  • Foot holders suitable for such purposes may include, for example, a base plate extending from the foot, for example, from the mid to distal thigh to the heel. The base plate preferably sits on the x-ray table. The patients' foot is positioned so that the posterior aspect of the heel is located on top of the base plate.
  • the medial aspect of the foot is placed against a medial guide connected rigidly to the base plate at a 90° angle by any suitable means (e.g., straps, velcro, plastic, tape, etc.).
  • a second, lateral guide attached to the base plate at a 90° angle with a sliding mechanism can then be moved toward the lateral aspect of the foot and be locked in position, for example when it touches the lateral aspect of the foot.
  • the use of a foot holder can help improve the reproducibility of measurements of bone structure parameters or macro-anatomical and/or biomechanical parameters.
  • the patient or subject can be any warm-blooded animal.
  • patients, or subjects are chosen from the class Mammalia.
  • patients, or subjects would include humans and nonhuman primates such as chimpanzees and other apes and monkey species; farm animals such as cattle, sheep, pigs, goats and horses; domestic mammals such as dogs and cats; laboratory animals including rodents such as mice, rats and guinea pigs, and the like.
  • other non-mammals can be subjected to the protocols described herein without departing from the scope of the invention,
  • macro-anatomical parameters generally describe the shape, size or thickness of bone and/or surrounding structure. Oftentimes the typical parameters are, but need not be, greater than 0.5mm in size in at least one dimension.
  • macro-anatomical parameters include thickness of the femoral shaft cortex, thickness of the femoral neck cortex, cortical width, hip axis length, CCD (caput-collum-diaphysis) angle, neck-shaft angle and width of the trochanteric region.
  • macro-anatomical parameters include thickness of the superior and inferior endplate, thickness of the anterior, lateral and posterior vertebral walls, diameter and height of the vertebral body, dimensions of the spinal canal and the posterior elements.
  • the ray is centered onto the hip joint medial and superior to the greater trochanter.
  • a calibration phantom such as an aluminum step wedge may also be included in the images to calibrate gray values before further image analysis.
  • dental x-rays are preferred because of the relative ease and lack of expense in obtaining these images.
  • the mandible and maxilla are primarily composed of trabecular bone. Since the metabolic turnover of trabecular bone is approximately eight times greater than that of cortical bone, areas of predominantly trabecular bone such as the vertebral body are preferred sites for measuring bone mineral density. Lang et al. (1991) Radiol Clin North Am 29:49-76.
  • trabecular bone is clearly visible on the dental x-ray image, thus facilitating quantitative analysis of bone mineral density and structure.
  • Jeffcoat et al. (2000) Periodontol 23:94-102; Southard et al. (2000) J Dent Res 79:964-969.
  • the earliest bone loss in osteoporosis patients occurs in areas of trabecular bone.
  • Multiple dental x-ray images are commonly made in most Americans throughout life. Indeed, there are approximately 750 million U.S. dental visits annually and 150 million of these patients result in more than 1 billion dental x-rays taken each year.
  • the ability to diagnose osteoporosis on dental x-rays would be extremely valuable since it would create the opportunity for low-cost mass screening of the population.
  • x-ray imaging is performed using standard x-ray equipment, for instance standard dental x-ray equipment (e.g. General Electric Medical Systems, Milwaukee, Wl).
  • standard dental x-ray equipment e.g. General Electric Medical Systems, Milwaukee, Wl.
  • X-rays of the incisor region and canine region are acquired using a standard x-ray imaging technique with 80 kVp and automatic exposure using a phototimer or using a manual technique with 10mA tube current.
  • X-ray images are acquired, for example, on Kodak Ultraspeed film (Kodak, Rochester, NY).
  • X-ray images may be digitized using a commercial flatbed scanner with transparency option (Acer ScanPremio ST).
  • other imaging techniques are typically performed using standard equipment, for instance, MRI or CAT equipment.
  • the images include accurate reference markers, for example calibration phantoms for assessing bone mineral density and/or bone structure and/or one or more macro-anatomical and/or biomechanical parameters on any given image.
  • Calibration references also known as calibration phantoms
  • U.S. Patent No. 5,335,260 discloses a calibration phantom representative of human tissue containing variable concentrations of calcium that serves as reference for quantifying calcium, bone mass and bone mineral density in x-ray and CT imaging systems.
  • currently available calibration phantoms are not always accurate.
  • some of the methods and devices described herein are designed to assess not only bone mineral density but also bone structure and, in addition, macro-anatomical and/or biomechanical parameters. By assessing two or more of these parameters, more accurate testing and screening can be provided for conditions such as osteoporosis.
  • the current invention provides for methods and devices that allow accurate quantitative assessment of information contained in an x-ray such as density of an anatomic structure and/or morphology of an anatomic structure.
  • Any suitable calibration phantom can be used, for example, one that comprises aluminum or other radio-opaque materials.
  • U.S. Patent No. 5,335,260 describes other calibration phantoms suitable for use in assessing bone mineral density in images.
  • suitable calibration reference materials can be fluid or fluid-like materials, for example, one or more chambers filled with varying concentrations of calcium chloride or the like.
  • the system used to monitor bone mineral density and/or bone structure and/or one or more macro-anatomical and/or biomechanical parameters in a target organism comprises an image (e.g., a dental or hip radiograph), which provides information on the subject; an assembly including a calibration phantom, which acts as a reference for the data in the image; and at least one data processing system, which evaluates and processes the data from the image and/or from the calibration phantom assembly.
  • an image e.g., a dental or hip radiograph
  • an assembly including a calibration phantom which acts as a reference for the data in the image
  • at least one data processing system which evaluates and processes the data from the image and/or from the calibration phantom assembly.
  • a calibration phantom can contain a single, known density or structure reference. Furthermore, a gradient in density can be achieved by varying the thickness or the geometry of the calibration phantom along the path of the x-ray beam, for example, by using a V-shape of the calibration phantom of varying thickness (Fig. 4).
  • the calibration phantom can also include angles. For example, the calibration phantom can be "T"-shaped or "L"-shaped thereby including one or more 90 degree angles.
  • the calibration phantom can contain several different areas of different radio-opacity.
  • the calibration phantom can have a step-like design, whereby changes in local thickness of the wedge result in differences in radio-opacity.
  • Stepwedges using material of varying thickness are frequently used in radiology for quality control testing of x-ray beam properties.
  • the intensity and spectral content of the x-ray beam in the projection image can be varied.
  • Stepwedges are commonly made of aluminum, copper and other convenient and homogeneous materials of known x-ray attenuation properties.
  • Stepwedge-like phantoms can also contain calcium phosphate powder or calcium phosphate powder in molten paraffin.
  • the calibration reference may be designed such that the change in radio-opacity is from periphery to center (for example in a round, ellipsoid, rectangular, triangular of other shaped structure).
  • the calibration reference can also be constructed as plurality of separate chambers, for example fluid filled chambers, each including a specific concentration of a reference fluid (e.g., calcium chloride).
  • a calibration phantom can also contain metal powder, e.g. aluminum or steel powder, embedded within it (for example, embedded in a plastic).
  • the calibration phantom is specifically designed to serve as a reference for bone structure (e.g., trabecular spacing, thickness and the like).
  • the calibration wedge can contain one or more geometric patterns with known dimensions, e.g. a grid whereby the spacing of a grid, thickness of individual grid elements, etc. are known.
  • This known geometric pattern of radio-opaque elements in the calibration phantom can be used to improve the accuracy of measurements of trabecular bone structure in an x-ray.
  • Such measurements of trabecular bone structure can include, but are not limited to, trabecular spacing, trabecular length and trabecular thickness.
  • Such measurements of trabecular spacing, trabecular length and trabecular thickness can, for example, be performed in a dental or spine or hip x-ray.
  • These calibration phantoms can be made up of a variety of materials include, plastics, metals and combinations thereof.
  • the reference components can be solid, powdered, fluid or combinations thereof.
  • the calibration wedge can also be used to improve measurements of bone structure.
  • the calibration phantom is specifically designed to serve as a reference for macro-anatomical parameters (e.g., in the hip joint, thickness of the femoral shaft cortex, thickness of the femoral neck cortex, cortical width, hip axis length, CCD (caput-collum-diaphysis) angle, neck-shaft angle and width of the trochanteric region; and in the spine, thickness of the superior and inferior endplate, thickness of the anterior, lateral and posterior vertebral walls, diameter and height of the vertebral body, dimensions of the spinal canal and the posterior elements).
  • the calibration wedge can contain one or more geometric patterns with known dimensions, e.g.
  • the calibration phantom can be used to improve the accuracy of measurements of macro-anatomical and/or biomechanical parameters in an x-ray, for example by aiding in the correction of image magnification.
  • Such measurements of macro-anatomical parameters can, for example, be performed in a dental or spine or hip x-ray.
  • These calibration phantoms can be made up of a variety of materials include, plastics, metals and combinations thereof.
  • the reference components can be solid, powdered, fluid or combinations thereof.
  • the calibration wedge can also be used to improve measurements of bone structure.
  • the present invention contemplates analysis of dental x-ray images for information on bone structure, bone mineral density or both structure and density, it will be apparent that calibration phantoms will be selected based on whether structure, density or both are being measured. Thus, one or more calibration phantoms may be present.
  • the at least one marker when present, be positioned at a known density and/or structure in the phantom. Furthermore, it is preferred that at least one geometric shape or pattern is included in the calibration phantom. Any shape can be used including, but not limited to, squares, circles, ovals, rectangles, stars, crescents, multiple-sided objects (e.g., octagons), V- or U-shaped, inverted V- or U-shaped, irregular shapes or the like, so long as their position is known to correlate with a particular density of the calibration phantom. In preferred embodiments, the calibration phantoms described herein are used in 2D planar x-ray imaging.
  • the calibration phantoms can be imaged before or after the x-ray image is taken. Alternatively, the calibration phantom can be imaged at the same time as the x- ray image.
  • the calibration phantom can be physically connected to an x-ray film and/or film holder. Such physical connection can be achieved using any suitable mechanical or other attachment mechanism, including but not limited to adhesive, a chemical bond, use of screws or nails, welding, a VelcroTM strap or VelcroTM material and the like.
  • a calibration phantom can be physically connected to a detector system or a storage plate for digital x-ray imaging using one or more attachment mechanisms (e.g., a mechanical connection device, a VelcroTM strap or other VelcroTM material, a chemical bond, use of screws or nails, welding and an adhesive).
  • attachment mechanisms e.g., a mechanical connection device, a VelcroTM strap or other VelcroTM material, a chemical bond, use of screws or nails, welding and an adhesive.
  • the external standard and the film can be connected with use of a holding device, for example using press fit for both film and external standard.
  • the calibration phantom assembly can be attached to an anatomical structure, for example one or more teeth, mucus membranes, the mandible and/or maxilla.
  • the calibration phantom can be attached (e.g., via adhesive attachment means) to the epithelium or mucous membrane inside overlying the mandible or the maxilla.
  • the calibration phantom can be placed on or adjacent to a tooth, for example, a V- or U-shaped (in the case of the maxilla) or an inverted V- or U-shaped (in the case of the mandible) calibration phantom can be used.
  • the opening of the V or U will be in contact with the free edge of at least one tooth or possibly several teeth (Fig.
  • a calibration phantom is included in the field of view.
  • Any suitable calibration phantom can be used, for example, one that comprises aluminum or other radio-opaque materials.
  • U.S. Patent No. 5,335,260 describes other calibration phantoms suitable for use in assessing bone mineral density in images.
  • Examples of other suitable calibration reference materials can be fluid or fluid-like materials, for example, one or more chambers filled with varying concentrations of calcium chloride or the like.
  • the material of the phantom is stainless steel (e.g., AISI grade 316 comprising carbon (0.08%); manganese (2%); silicon (1%); phosphorus (0.045%); sulphur (0.03%); nickel (10-14%); chromium (16-18%); molybdenum (2-3%); plus iron to make up 100%).
  • AISI grade 316 comprising carbon (0.08%); manganese (2%); silicon (1%); phosphorus (0.045%); sulphur (0.03%); nickel (10-14%); chromium (16-18%); molybdenum (2-3%); plus iron to make up 100%.
  • the relative percentages of the components may be with respect to weight or volume.
  • calibration phantoms suitable for attachment to an anatomical structure can have different shapes depending on the shape of the anatomical structure (e.g., tooth or teeth) on which or adjacent to which it will be placed including, but not limited to, U-shaped, V-shaped, curved, flat or combinations thereof.
  • U-shaped (or inverted U-shaped) calibration phantoms can be positioned on top of molars while V-shaped (or inverted V-shaped) calibration phantoms can be positioned on top of incisors.
  • the calibration phantom can rest on top of the tooth just based on its gravity or it can be attached to the tooth (e.g., using adhesive). In the case of the teeth on the maxilla, the calibration phantom will typically be attached to the tooth, for example with use of an adhesive.
  • Any of these attachments may be permanent or temporary and the calibration phantom can be integral (e.g., built-in) to the film, film holder and/or detector system or can be attached or positioned permanently or temporarily appropriately after the film and/or film holder is produced.
  • the calibration phantom can be designed for single-use (e.g., disposable) or for multiple uses with different x-ray images.
  • the calibration phantom is reusable and, additionally, can be sterilized between uses.
  • Integration of a calibration phantom can be achieved by including a material of known x-ray density between two of the physical layers of the x- ray film. Integration can also be achieved by including a material of known x-ray density within one of the physical layers of the x-ray film.
  • the calibration phantom can be integrated into the film cover.
  • a calibration phantom or an external standard can also be integrated into a detector system or a storage plate for digital x-ray imaging.
  • integration can be achieved by including a material of known x-ray density between two of the physical layers of the detector system or the storage plate. Integration can also be achieved by including a material of know x-ray density within one of the physical layers of the detector system or the storage plate.
  • cross-hairs, lines or other markers may be placed on the apparatus as indicators for positioning of the calibration phantom. These indicators can help to ensure that the calibration phantom is positioned such that it doesn't project on materials that will alter the apparent density in the resulting image.
  • any of the calibration phantom-containing assemblies described herein can be used in methods of analyzing and/or quantifying bone structure and/or one or more macro-anatomical and/or biomechanical parameters (or bone mineral density) in an x-ray image.
  • the methods generally involve simultaneously imaging or scanning the calibration phantom and another material (e.g., bone tissue from a subject) for the purpose of quantifying the density of the imaged material (e.g., bone mass).
  • the calibration phantom, the x-ray tube or dental x-ray film is typically positioned in a manner to ensure inclusion of the calibration phantom and a portion of the mandible and/or maxilla on the dental x-ray image.
  • the calibration phantom, the x-ray tube and the dental x-ray film are positioned so that at least a portion of the section of the mandible or maxilla included on the image will contain predominantly trabecular bone rather than cortical bone.
  • the calibration phantom is preferably imaged or scanned simultaneously with the individual subject, although the invention allows for non-simultaneous scanning of the phantom and the subject.
  • Methods of scanning and imaging structures by x-ray imaging technique are well known.
  • reference calibration samples allow corrections and calibration of the absorption properties of bone.
  • X-ray imaging assemblies that are currently in use do not take into account the position of the calibration phantom in relation to the structures being imaged. Thus, when included in known assemblies, calibration phantom(s) are often positioned such that they project on materials or structures (e.g., bone) that alter apparent density of the calibration phantom in the resulting x-ray image.
  • the methods described herein ensure that the calibration phantom projects free of bone (e.g., teeth, jaw) tissue. This can be accomplished in a variety of ways, for example, positioning the calibration phantom in the x-ray film or in the x-ray film holder such that it will appear between the teeth in the dental x-ray.
  • the calibration phantom materials and methods of the present invention are preferably configured to be small enough and thin enough to be placed inside the mouth, and the method of the present invention can be used to quantify bone mass using standard dental x-ray systems, for example by including temporary or permanent calibration phantoms in dental x-ray film holders. Further, it is highly desirable that the calibration phantom be positioned so that at least a portion doesn't project on structures or materials that will alter the apparent density or structural characteristics of the calibration phantoms.
  • calibration phantom at a defined distance relative to at least one tooth or the mandible or the maxilla whereby a substantial portion of the calibration phantom projects free of said tooth, said mandible or said maxilla on the x-ray image.
  • Any suitable distance can be used, for example between about 1 mm and 5 cm or any value therebetween.
  • a cross-calibration phantom can be used to optimize system performance, e.g. x-ray tube settings or film processor settings, or to improve the comparability of different machines or systems, typically located at different sites.
  • a separate image may be obtained which does not include a patient or a body part.
  • the image includes the primary calibration phantom used in patients, e.g. a step-wedge of known density, and the cross-calibration phantom.
  • the apparent density of the primary calibration phantom is then calibrated against the density of the cross-calibration phantom.
  • the resultant cross-calibration of the primary phantom can help to improve the accuracy of measurements of bone density, bone structure and macro-anatomical and/or biomechanical parameters. It can also help improve the overall reproducibility of the measurements.
  • an x-ray technologist or a dental hygienist will perform a cross-calibration test once a day, typically early in the morning, prior to the first patient scans.
  • the results of the cross-calibration or the entire cross-calibration study can be transmitted via a network to a central computer.
  • the central computer can then perform adjustments designed to maintain a high level of comparability between different systems. 1.2.
  • information inherent in the anatomic structure or the non-living object can be used to estimate the density and/or structure and/or macro-anatomy of selected bone regions of interest within the anatomic structure or the non-living object.
  • the density of muscle, fat, water (e.g., soft tissue), metal (e.g., dental fillings) and air are typically known, the density of air surrounding an anatomic structure or non-living object, the density of subcutaneous fat, and the density of muscle tissue can be used to estimate the density of a selected region of bone, for example within the distal radius.
  • a weighted mean can be determined between one or more of the internal standards (e.g., air, water, metal, and/or fat) and used as internal standards to determine bone density in the same x-ray image.
  • the density of a tooth or a portion of a tooth can be used to estimate the density of a selected region of bone, e.g. an area in the mandible.
  • a holder can be used to position the calibration phantom.
  • the holder can be U-shaped or V-shaped (Fig. 5) for ease in attachment to a tooth.
  • the attachment can be, for example, with an adhesive.
  • the calibration phantom in turn, can be attached to the holder.
  • the calibration phantom can be attached to holders comprising one or more molds of at least one or more teeth.
  • the holder can be used to position both the film and the calibration phantom relative to the osseous structure that will be included in the x-ray image.
  • a holding device that can hold the x-ray film is integrated in the calibration phantom. This holding device can hold the film in place prior to taking the x-ray.
  • the holding device can be spring-loaded or use other means such as mechanical means of holding and stabilizing the x-ray film.
  • the holder may comprise a disposable or sterilizeable hygienic cover. See, e.g., WO 99/08598, the disclosure of which is incorporated by reference herein in its entirety.
  • the holder may comprise multiple components, for example, the calibration phantom and a integrated or insertable bolus back that can serve to enhance the accuracy of the calibration phantom by accounting for the effect of soft tissue that may project with the calibration phantom and/or with the bone.
  • the calibration phantom can be configured so that it stabilizes against the surrounding tissues on its own without the use of an additional holder.
  • the calibration phantom can be protected with a hygienic cover.
  • the holder e.g., hygienic cover
  • the holder may be comprised of a rigid material, a flexible material or combinations thereof.
  • the holder may include one or more pockets/compartments adapted to receive additional components such as the calibration phantom, a bolus back or the like. Additionally, one or more portions of the holder may be radiolucent. 2.0. Analysis md Manipulation ⁇ f Data [0084] The data obtained from images taken as described above is then preferably analyzed and manipulated.
  • the systems and assemblies described herein can also include one or more computational units designed, for example, to analyze bone density or bone structure or macro-anatomical and/or biomechanical data in the image; to identify an anatomical landmark in an anatomical region; to correct for soft tissue measurements; and/or to evaluate bone density and structure and macro- anatomy of the image.
  • the computational unit can include any software, chip or other device used for calculations. Additionally, the computational unit may be designed to control the imaging assembly or detector (as well as other parameters related to the detector(s)). Other applications of the computational unit to the methods and devices described herein will be recognized by those skilled in the art. The computational unit may be used for any other application related to this technology that may be facilitated with use of computer software or hardware.
  • the computational unit can also further comprise a database comprising, for example, reference anatomical maps and the computational unit is further designed to compare the anatomical map with the reference anatomical map.
  • the reference anatomical map may be historic (from the same or another patient, generated as part of an interrogation protocol), or theoretical or any other type of desired reference map.
  • Any image can be analyzed in order to obtain and manipulate data.
  • data points, derived data, and data attributes database may comprise the following: (1) the collection of data points, said data points comprising information obtained from an image, for example, bone mineral density information or information on bone structure (architecture); and (2) the association of those data points with relevant data point attributes.
  • the method may further comprise (3) determining derived data points from one or more direct data points and (4) associating those data points with relevant data point attributes.
  • the method may also comprise (5) collection of data points using a remote computer whereby said remote computer operates in a network environment.
  • the information is obtained from a dental x-ray image.
  • dental x-ray images can be acquired at a local site using known techniques. If the x-ray image was captured using conventional x-ray film, the data points (information) of the x-ray image can be digitized using a scanning device. The digitized x-ray image information can then be transmitted over the network, e.g. the Internet, into a remote computer or server.
  • the x-ray image information is already available in digital format.
  • the image can be transmitted directly over the network, e.g. the Internet.
  • the information can also be compressed and/or encrypted prior to transmission. Transmission can also be by other methods such as fax, mail or the like.
  • the methods of and compositions described herein make use of collections of data sets of measurement values, for example measurements of bone structure and/or bone mineral density from x-ray images. Records may be formulated in spreadsheet-like format, for example including data attributes such as date of x-ray, patient age, sex, weight, current medications, geographic location, etc.
  • the database formulations may further comprise the calculation of derived or calculated data points from one or more acquired data points. A variety of derived data points may be useful in providing information about individuals or groups during subsequent database manipulation, and are therefore typically included during database formulation.
  • Derived data points include, but are not limited to the following: (1) maximum bone mineral density, determined for a selected region of bone or in multiple samples from the same or different subjects; (2) minimum bone mineral density, determined for a selected region of bone or in multiple samples from the same or different subjects; (3) mean bone mineral density, determined for a selected region of bone or in multiple samples from the same or different subjects; (4) the number of measurements that are abnormally high or low, determined by comparing a given measurement data point with a selected value; and the like.
  • Other derived data points include, but are not limited to the following: (1) maximum value of a selected bone structure parameter, determined for a selected region of bone or in multiple samples from the same or different subjects; (2) minimum value of a selected bone structure parameter, determined for a selected region of bone or in multiple samples from the same or different subjects; (3) mean value of a selected bone structure parameter, determined for a selected region of bone or in multiple samples from the same or different subjects; (4) the number of bone structure measurements that are abnormally high or low, determined by comparing a given measurement data point with a selected value; and the like.
  • Other derived data points include, but are not limited to the following: (1) maximum value of a selected macro-anatomical and/or biomechanical parameter, determined for a selected region of bone or in multiple samples from the same or different subjects; (2) minimum value of a selected macro-anatomical and/or biomechanical parameter, determined for a selected region of bone or in multiple samples from the same or different subjects; (3) mean value of a selected macro-anatomical and/or biomechanical parameter, determined for a selected region of bone or in multiple samples from the same or different subjects; (4) the number of macro-anatomical and/or biomechanical measurements that are abnormally high or low, determined by comparing a given measurement data point with a selected value; and the like.
  • the amount of available data and data derived from (or arrived at through analysis of) the original data provide provides an unprecedented amount of information that is very relevant to management of bone related diseases such as osteoporosis. For example, by examining subjects over time, the efficacy of medications can be assessed.
  • Measurements and derived data points are collected and calculated, respectively, and may be associated with one or more data attributes to form a database.
  • the amount of available data and data derived from (or arrived at through analysis of) the original data provide provides an unprecedented amount of information that is very relevant to management of bone related diseases such as osteoporosis. For example, by examining subjects over time, the efficacy of medications can be assessed.
  • Data attributes can be automatically input with the x-ray image and can include, for example, chronological information (e.g., DATE and TIME). Other such attributes may include, but are not limited to, the type of x-ray imager used, scanning information, digitizing information and the like. Alternatively, data attributes can be input by the subject and/or operator, for example subject identifiers, i.e. characteristics associated with a particular subject.
  • identifiers include but are not limited to the following: (1) a subject code (e.g., a numeric or alpha-numeric sequence); (2) demographic information such as race, gender and age; (3) physical characteristics such as weight, height and body mass index (BMI); (4) selected aspects of the subject's medical history (e.g., disease states or conditions, etc.); and (5) disease-associated characteristics such as the type of bone disorder, if any; the type of medication used by the subject.
  • a subject code e.g., a numeric or alpha-numeric sequence
  • demographic information such as race, gender and age
  • physical characteristics such as weight, height and body mass index (BMI)
  • BMI body mass index
  • selected aspects of the subject's medical history e.g., disease states or conditions, etc.
  • disease-associated characteristics such as the type of bone disorder, if any; the type of medication used by the subject.
  • each data point would typically be identified with the particular subject, as well as the demographic, etc. characteristic of that subject.
  • the databases comprise data points, a numeric value which correspond to physical measurement (an "acquired” datum or data point) or to a single numeric result calculated or derived from one or more acquired data points that are obtained using the various methods disclosed herein.
  • the databases can include raw data or can also include additional related information, for example data tags also referred to as "attributes" of a data point.
  • the databases can take a number of different forms or be structured in a variety of ways.
  • the most familiar format is tabular, commonly referred to as a spreadsheet.
  • a variety of spreadsheet programs are currently in existence, and are typically employed in the practice of the present invention, including but not limited to Microsoft Excel spreadsheet software and Corel Quattro spreadsheet software.
  • association of data points with related attributes occurs by entering a data point and attributes related to that data point in a unique row at the time the measurement occurs.
  • Relational databases typically support a set of operations defined by relational algebra. Such databases typically include tables composed of columns and rows for the data included in the database. Each table of the database has a primary key, which can be any column or set of columns, the values for which uniquely identify the rows in a table. The tables in the database can also include a foreign key that is a column or set of columns, the values of which match the primary key values of another table. Typically, relational databases also support a set of operations (e.g., select, join and combine) that form the basis of the relational algebra governing relations within the database.
  • a set of operations e.g., select, join and combine
  • relational databases can be implemented in various ways. For instance, in Sybase® (Sybase Systems, Emeryville, CA) databases, the tables can be physically segregated into different databases. With Oracle® (Oracle Inc., Redwood Shores, CA) databases, in contrast, the various tables are not physically separated, because there is one instance of work space with different ownership specified for different tables. In some configurations, databases are all located in a single database (e.g., a data warehouse) on a single computer. In other instances, various databases are split between different computers.
  • Sybase® Sybase Systems, Emeryville, CA
  • Oracle® Oracle Inc., Redwood Shores, CA
  • databases are all located in a single database (e.g., a data warehouse) on a single computer. In other instances, various databases are split between different computers.
  • data may be aggregated, sorted, selected, sifted, clustered and segregated by means of the attributes associated with the data points.
  • Relationships in various data can be directly queried and/or the data analyzed by statistical methods to evaluate the information obtained from manipulating the database.
  • a distribution curve can be established for a selected data set, and the mean, median and mode calculated therefor. Further, data spread characteristics, e.g. variability, quartiles and standard deviations can be calculated.
  • Non-parametric tests may be used as a means of testing whether variations between empirical data and experimental expectancies are attributable merely to chance or to the variable or variables being examined. These include the Chi Square test, the Chi Square Goodness of Fit, the 2 x 2 Contingency Table, the Sign Test, and the Phi Correlation Coefficient.
  • tools and analyses include, but are not limited to, cluster analysis, factor analysis, decision trees, neural networks, rule induction, data driven modeling, and data visualization. Some of the more complex methods of data mining techniques are used to discover relationships that are more empirical and data-driven, as opposed to theory- driven, relationships.
  • Exemplary data mining software that can be used in analysis and/or generation of the databases of the present invention includes, but is not limited to: Link Analysis (e.g., Associations analysis, Sequential Patterns, Sequential time patterns and Bayes Networks); Classification (e.g., Neural Networks Classification, Bayesian Classification, k-nearest neighbors classification, linear discriminant analysis, Memory based Reasoning, and Classification by Associations); Clustering (e.g., k-Means Clustering, demographic clustering, relational analysis, and Neural Networks Clustering); Statistical methods (e.g., Means, Std dev, Frequencies, Linear Regression, non-linear regression, t-tests, F-test, Chi2 tests, Principal Component Analysis, and Factor Analysis); Prediction (e.g., Neural Networks Prediction Models, Radial Based Functions predictions, Fuzzy logic predictions, Times Series Analysis, and Memory based Reasoning); Operating Systems; and Others (e.g., Parallel Scalability, Simple
  • data e.g., bone structural information or macro- anatomical and/or biomechanical information or bone mineral density information
  • reference databases can be used to aid analysis of any given subject's x-ray image, for example, by comparing the information obtained from the subject to the reference database.
  • the information obtained from the normal control subjects will be averaged or otherwise statistically manipulated to provide a range of "normal" (reference) measurements. Suitable statistical manipulations and/or evaluations will be apparent to those of skill in the art in view of the teachings herein.
  • the comparison of the subject's x-ray information to the reference database can be used to determine if the subject's bone information falls outside the normal range found in the reference database or is statistically significantly different from a normal control. Data comparison and statistical significance can be readily determined by those of skill in the art using for example the z-test or t-test statistics for continuous variables, the chi-square test or Fisher's exact test for categorical data and the rank-sum test or Kruskal-Wallis test for ranked data.
  • the use of reference databases in the analysis of x-ray images facilitates that diagnosis, treatment and monitoring of bone conditions such as osteoporosis.
  • the data is preferably stored and manipulated using one or more computer programs or computer systems. These systems will typically have data storage capability (e.g., disk drives, tape storage, CD-ROMs, etc.). Further, the computer systems may be networked or may be stand-alone systems. If networked, the computer system would be able to transfer data to any device connected to the networked computer system for example a medical doctor or medical care facility using standard e-mail software, a central database using database query and update software (e.g., a data warehouse of data points, derived data, and data attributes obtained from a large number of subjects). Alternatively, a user could access from a doctor's office or medical facility, using any computer system with Internet access, to review historical data that may be useful for determining treatment.
  • database query and update software e.g., a data warehouse of data points, derived data, and data attributes obtained from a large number of subjects.
  • the application includes the executable code required to generate database language statements, for example, SQL statements. Such executables typically include embedded SQL statements.
  • the application further includes a configuration file that contains pointers and addresses to the various software entities that are located on the database server in addition to the different external and internal databases that are accessed in response to a user request.
  • the configuration file also directs requests for database server resources to the appropriate hardware, as may be necessary if the database server is distributed over two or more different computers.
  • each networked computer system includes a World Wide Web browser that provides a user interface to the networked database server.
  • the networked computer system is able to construct search requests for retrieving information from a database via a Web browser.
  • users can typically point and click to user interface elements such as buttons, pull down menus, and other graphical user interface elements to prepare and submit a query that extracts the relevant information from the database.
  • Requests formulated in this manner are subsequently transmitted to the Web application that formats the requests to produce a query that can be used to extract the relevant information from the database.
  • the Web application accesses data from a database by constructing a query in a database language such as Sybase or Oracle SQL which is then transferred to a relational database management system that in turn processes the query to obtain the pertinent information from the database.
  • a database language such as Sybase or Oracle SQL
  • the present invention describes a method of providing data obtained from x-ray images on a network, for example the Internet, and methods of using this connection to provide real-time and delayed data analysis.
  • the central network can also allow access by the physician to a subject's data. Similarly, an alert could be sent to the physician if a subject's readings are out of a predetermined range, etc. The physician can then send advice back to the patient via e-mail or a message on a web page interface. Further, access to the entire database of data from all subjects may be useful for statistical or research purposes. Appropriate network security features (e.g., for data transfer, inquiries, device updates, etc.) are of course employed.
  • a remote computer can be used to analyze the x-ray that has been transmitted over the network automatically. For example, x-ray density information or structural information about an object can be generated in this fashion. X-ray density information can, for example, be bone mineral density. If used in this fashion, the test can be used to diagnose bone-related conditions such as osteoporosis.
  • an interface such as an interface screen that includes a suite of functions is included to enable users to easily access the information they seek from the methods and databases of the invention.
  • Such interfaces usually include a main menu page from which a user can initiate a variety of different types of analyses.
  • the main menu page for the databases generally include buttons for accessing certain types of information, including, but not limited to, project information, inter-project comparisons, times of day, events, dates, times, ranges of values, etc.
  • Computer Program Products A variety of computer program products can be utilized for conducting the various methods and analyses disclosed herein.
  • the computer program products comprise a computer-readable medium and the code necessary to perform the methods set forth supra.
  • the computer-readable medium on which the program instructions are encoded can be any of a variety of known medium types, including, but not limited to, microprocessors, floppy disks, hard drives, ZIP drives, WORM drives, magnetic tape and optical medium such as CD-ROMs.
  • an analysis of the morphology and density of the bone can be performed, for example using suitable computer programs.
  • This analysis of the object's morphology can occur in two-dimensions or three-dimensions.
  • such analysis of the transmitted x-ray image can be used to measure parameters that are indicative or suggestive of bone loss or metabolic bone disease.
  • Such parameters include all current and future parameters that can be used to evaluate osseous structures.
  • such parameters include, but are not limited to, trabecular spacing, trabecular thickness, trabecular connectivity and intertrabecular space.
  • Information on the morphology or 2D or 3D structure of an anatomic object can be derived more accurately, when image acquisition parameters such as spatial resolution are known. Other parameters such as the degree of cone beam distortion can also be helpful in this setting.
  • an image can be transmitted from a local site into a remote server and the remote server can perform an automated analysis of the image. Further, the remote server or a computer connected to the remote server can then generate a diagnostic report.
  • a computer program e.g., on the remote server or on a computer connected to the remote server
  • the remote server can then transmit the diagnostic report to a physician, typically the physician who ordered the test or who manages the patient.
  • the diagnostic report can also be transmitted to third parties, e.g. health insurance companies.
  • Such transmission of the diagnostic report can occur electronically (e.g. via e-mail), via mail, fax or other means of communication. All or some of the transmitted information (e.g., patient identifying information) can be encrypted to preserve confidentiality of medical records.
  • one exemplary system for analyzing bone morphology or structure in a subject system via a dental x-ray that includes at least a portion of the mandible and/or maxilla of a subject, followed by evaluation or the x-ray image.
  • Dental x-rays are obtained in any conventional method.
  • the x-ray produces an image that can be interpreted (for example, employing a selected algorithm and/or computer program) by an associated system controller to provide a bone mineral density or bone structure evaluation for display.
  • the monitoring system can comprise two or more components, in which a first component comprises an x-ray image and calibration phantom that are used to extract and detect bone-related data on the subject, and a second component that receives the data from the first component, conducts data processing on the data and then displays the processed data.
  • a first component comprises an x-ray image and calibration phantom that are used to extract and detect bone-related data on the subject
  • a second component that receives the data from the first component, conducts data processing on the data and then displays the processed data.
  • Microprocessor functions can be found in one or both components.
  • the second component of the monitoring system can assume many forms
  • correction factors will take into account one or more of a wide variety of influences (e.g.,. soft tissue thickness, region from which the data is extracted and the like) that can alter apparent density or structure information on the image.
  • one or more reference databases can be used for calibration and normalization purposes.
  • image normalization or correction of soft-tissue attenuation can be performed using patient characteristic data such as patient weight, height and body mass index.
  • patient characteristic data such as patient weight, height and body mass index.
  • a higher soft-tissue attenuation can be assumed in high weight and low height subjects; a lower soft-tissue attenuation will be assumed in low weight and high height subjects.
  • a standard calibration curve is applied to x-ray images, whereby said calibration curve can be derived from reference x-rays obtained with use of calibration phantoms.
  • 100 patients can undergo dental x-rays with a calibration phantom and a standard calibration curve can be derived from these images.
  • 100 patients can undergo hip x-rays with a calibration phantom and a standard calibration curve can be derived from these images.
  • Different calibration curves can be generated for different populations, for example, by generating different calibration curves for different ranges in body mass index, body height, sex, race etc.
  • identification of anatomic landmarks of the structure to be analyzed or identification of anatomical landmarks adjacent to the structure to be analyzed with subsequent positioning and computer analysis of the x-ray image relative to these anatomic landmarks or with subsequent positioning and computer analysis of anatomical region of interest (ROI) relative to these anatomic landmarks is performed.
  • the present invention includes also positioning dental or other x-ray detectors, positioning the dental or other x-ray tube, and analyzing the resulting images using landmarks based on either 1) textural information, 2.) structural information, 3.) density information (e.g. density), or 4) 2 or 3 dimensional contour information 5) a combinations thereof of the tissue or structure to be measured and of tissues or structures adjacent to the measurement site.
  • the invention also includes methods and devices that are not necessarily based solely on anatomical landmarks, but in some applications can be combined with anatomical landmark embodiments.
  • many of the embodiments described herein are designed for automated use with a minimum of operator intervene and preferably remote or computer control of such devices.
  • an alignment device may be used to ensure perpendicular or near perpendicular alignment of the dental or other x-ray tube relative to the x-ray film, thereby decreasing geometric distortion resulting from tube angulation.
  • an x-ray film holder is positioned relative to an anatomical landmark, e.g. the posterior wall of the mandible in the incisor region.
  • An anatomical landmark e.g. the posterior wall of the mandible in the incisor region.
  • a side-view of an exemplary alignment system using a dental x-ray film holder is shown in Figure 15.
  • the system includes bite block (100), stainless steel rod (101), film (103), optional calibration phantom (104), Rinn holder (102) typically having a ring or donut shape, and extension tubing (200).
  • the extension tubing is designed to fit within the Rinn holder and may be temporarily or permanently attached.
  • the system can achieve high reproducibility of the film position relative to an anatomical landmark such as the alveolar ridge or the posterior wall of the mandible.
  • the extension tubing allows for alignment of the x-ray tube so that it is near perpendicular to the Rinn instrument and, ultimately, the dental film.
  • a mechanical or electromagnetic device is preferably used in order to achieve perpendicular or near perpendicular alignment between the metal tube anterior to the x- ray tube and the Rinn holder.
  • the metal tube can be physically attached to the Rinn holder with use of one or more VelcroTM straps or it can be aligned using optical aids such as levels, cross-hairs, light sources (points or areas), etc.
  • such physical attachment can be achieved with use of one or more magnets rigidly attached to the dental x-ray system metal tube and the Rinn holder.
  • the magnets on the Rinn holder and the dental x-ray system metal tube will be aligned and brought into physical contact.
  • an extension tube is attached, for example with an adhesive, to the Rinn holder.
  • the extension tubing can also be an integral part of the Rinn holder.
  • the extension tubing can be designed so that its inner diameter is slightly greater than the outer diameter of the dental x-ray system metal tube. The dental x-ray system metal tube is then inserted into the extension tubing attached to the Rinn holder thereby greatly reducing alignment error of the x-ray tube relative to the x-ray film.
  • the extension tubing can be designed so that its outer diameter is slightly smaller than the inner diameter of the dental x-ray system metal tube.
  • the dental x-ray system metal tube is then advanced over the extension tubing attached to the Rinn holder thereby greatly reducing alignment error of the x-ray tube relative to the x-ray film.
  • attachment means can be used for properly aligning the dental x-ray tube with the dental x-ray film. Combinations of attachment mechanisms are also possible.
  • the anatomical landmark that is selected is part of an anatomical region.
  • An anatomical region refers to a site on bone, tooth or other definable biomass that can be identified by an anatomical feature(s) or location.
  • An anatomical region can include the biomass underlying the surface. Usually, such a region will be definable according to standard medical reference methodology, such as that found in Williams et al., Gray's Anatomy, 1980.
  • the anatomical region can be selected from the group consisting of an edge of the mandible, an edge of the maxilla, an edge of a tooth, valleys or grooves in any of these structures or combinations thereof.
  • the dental x-ray image can be readily taken so as to include the anatomical site.
  • Other anatomical regions include but are not limited to the hip, the spine, the forearm, the foot, and the knee.
  • the region of interest is placed between the dental apices and the inferior mandibular cortex.
  • the apices can be found automatically in the following way: for each row of pixels, the gray value profile is examined. While a profile that intersects bone and dental roots in an alternating fashion has several distinct peaks and valleys, a profile that only covers trabecular bone shows irregular changes in the gray values (Fig. 6). The dental apices are located in the transitional region between these two patterns.
  • the measurement techniques to assess trabecular bone structure or macro-anatomical and/or biomechanical parameters are preferably designed to work without user intervention.
  • the ROI has a size of 5.4mm x 5.4mm (128x128 pixels). For other scanning resolutions, the pixel resolution of the ROI can be adjusted accordingly.
  • bone mineral density can be measured in all ROIs that are located on a line that is, for example, 8 mm inferior and parallel to the alveolar ridge.
  • the ROIs can be moved from left to right on a pixel-by- pixel basis.
  • the ROI with the lowest BMD can be chosen for further evaluation of the structural bone parameters. This helps to avoid inclusion of regions on the x-ray where bone mineral density may be overestimated due to projection of the curved parts of the mandible near the canine teeth.
  • the ROI with the median BMD can be used. Other statistical parameters can be employed for this purpose.
  • software or other computational unit can identify the selected anatomic landmark in an interrogated x-ray image and direct analysis of the image using various parameters and analytic functions. Further, such software or other computational analytical unit can be used to identify areas of particular density at a certain distance from the selected landmark. Similarly, manual or computer analysis can be used to identify areas of lowest, highest, median or average density (or structural characteristics) in relation to the selected landmark. [0134] Further, the same landmark may be compared at different times (intra- landmark comparison) or one or more landmarks may be compared (inter-landmark comparison). For instance, an intra-landmark comparison can be used during a single interrogation protocol that entails multiple interrogations of the same region with reference to a particular anatomical landmark. Statistical analysis as described herein and known in the art can be performed.
  • the invention provides for means of assessing bone structure, i.e. the two-dimensional or three-dimensional architectural organization of the trabecular bone including, but not limited to, measurement of trabecular spacing, trabecular thickness, trabecular length and trabecular connectivity. Other examples of measurements of bone structure are provided in TABLE 1. These measurements can be used alone or enhanced with use of calibration phantoms or external standards that can allow a correction or normalization of image intensity and that can in certain embodiments also allow a correction of geometric distortions for example resulting from cone beam geometry of an x-ray beam. [0136] The invention provides for means of assessing macro-anatomical and/or biomechanical parameters.
  • one or more measurements of bone structure or macro-anatomical and/or biomechanical parameters can be used to select a therapy, for example the use of anabolic or antiresorptive agent in the case of bone loss or deterioration.
  • measurements of bone structure and/or one or more macro-anatomical and/or biomechanical parameters are conducted over time to longitudinally monitor a subject's bone health longitudinally over time. Measurements can be performed at different time points T1 , T2, ..., Tn and changes in said bone structure and/or macro-anatomical and/or biomechanical parameters can be registered and used to track a patient's bone health.
  • a physician can be apprised of the measurements and can include a pre-determined cut-off value (e.g., when a bone structure or macro-anatomical and/or biomechanical parameter measured in a patient is more than one or two standard deviations different from a normal, healthy reference population) and use this information to select a therapy.
  • a pre-determined cut-off value e.g., when a bone structure or macro-anatomical and/or biomechanical parameter measured in a patient is more than one or two standard deviations different from a normal, healthy reference population
  • the data obtained and analyzed as described herein can be used to monitor a patient's response to therapy.
  • information regarding bone structural and/or macro-anatomical and/or biomechanical information in a patient receiving an anabolic or antiresorptive drug and be evaluated at different time intervals T1, T2,... , Tn and changes in said bone structure and/or macro-anatomical and/or biomechanical parameters can be used in order to assess therapeutic efficacy.
  • a physician can use this information to adjust the dose of a drug administered (e.g., for treatment of osteoporosis) or to change the drug regimen.
  • Other techniques using x-ray information such as tomosynthesis can also be used for measuring bone structure and for selecting said therapy or monitoring said therapy.
  • Bone structure can be measured using a number of different technical approaches. These include but are not limited to the Hough Transform, analysis of density and size distribution of trabeculae, multidimensional classification schemes, mean pixel intensity, variance of pixel intensity, Fourier spectral analysis, fractal dimension and morphological parameters.
  • the Hough transform (See, e.g., Hough "Machine analysis of bubble chamber pictures" in International Conference on High Energy Accelerators and Instrumentation. 1959. CERN) can be used to detect geometric objects in binary images.
  • the invention includes the use of such methods to analyze direction and length of structures in bone images.
  • the region of interest (ROI) can be blurred with a Gaussian filter.
  • the pixel values of the filtered ROI can then be subtracted from those in the original ROI, and the value 128 can be added at each pixel location. This results in an image with a mean gray value of 128, which is also used as a threshold to yield a binary image in which the trabeculae are represented by the white pixels.
  • p is the perpendicular distance of the line from the origin and ⁇ is the angle between the x-axis and the normal.
  • p is the perpendicular distance of the line from the origin and ⁇ is the angle between the x-axis and the normal.
  • the (p, ⁇ ) plane can be divided into bins, where each bin counts the number of transformed curves that pass through it. This number corresponds to the number of collinear points on a line segment in the original image, and thus the length of this segment. Furthermore, the transformed image provides information on the predominant angles of the line segments in the original image (see Fig, 8).
  • the average length and the variance of the line segments which are calculated for all bins with a count above a certain threshold, can be used as structural parameters for the shape of the bone trabeculae. Average length as well as the variability of the length to decrease in patients with osteoporosis.
  • the threshold has the effect that only segments of a certain minimal length are included in the calculation. Choosing the threshold so that it provides the best discrimination between healthy and diseased individuals can be readily determined by one of skill in the art in view of the teachings herein.
  • each bin is interpreted as an element with a mass equivalent to its count, is a way to measure the predominant angles of the trabecular segments.
  • the angle at cm is measured with respect to the alveolar rim to obtain a standardized value.
  • the variance of the segment angles (again measured after thresholding the bin counts) provides information on the anisotropy of the trabecular structure.
  • Morphological operations such as variations of dilation and erosion and combinations thereof can also be used to detect the size of structures in gray scale or binary images.
  • a skeleton operator can be used to extract and quantify trabeculae of different sizes and directions, which results in a measure of the size distribution of trabecular structures. This skeleton operator is based on the work described in Kumasaka et al. (1997) Dentomaxillofac Rad 26: 161 -168 and works as follows:
  • the 'closing' operation is defined as the minimum search after maximum search:
  • £ 2 is another structuring element that is of circular shape and can be varied in size, and therefore renders the skeleton operator sensitive to the size of the structures in the image.
  • the erosion of f with E 2 erases the structures that are smaller than E 2 and extracts those trabeculae that are at least equal in size. Those structures that are exactly equal in size is reduced to a width of one pixel.
  • the opening step with ⁇ causes all structures that are one pixel wide to disappear (second term in (1)). After subtraction of this term from the first one, only those trabecular structures that exactly match the size of E 2 remain. Finally, the image is thresholded with a level of 1. The effect of this operator is illustrated in Fig. 9.
  • Fig. 10 demonstrates the use of the skeleton operator with the same structural element diameters as in Fig. 9 on a gray scale region of interest from a dental x-ray containing trabecular bone.
  • the number of bright pixels in the binary images resulting from each skeleton operation corresponds to the portion of trabeculae of the particular size in the original image. If the percentage of the bright pixels with respect to the total number of pixels in each skeletonized image is plotted against the diameter of E 2 , the "center of mass" of the curve, i.e. the predominant structure size, can be used as an index to discriminate between osteoporotic and healthy bone.
  • the skeleton operator is preferably optimized and extended to detect structures that are oriented only in a specific direction. This can be achieved by adding erosion operations to the skeleton operator with structural elements in which, for example, only the diagonal pixels are set to 1. [0154] This can be used to calculate an anisotropy index, similar to the one derived from the Hough transform. Both anisotropy indices are tested with respect to their potential to distinguish healthy from osteoporotic bone.
  • the watershed segmentation can be applied to background subtracted gray level structures on x-ray images to characterize the homogeneity of trabecular structures.
  • This process takes into account the gray level contrast between structures to define marrow spaces.
  • the watershed segmentation when applied to background subtracted bone x-ray images, defines regions with lower gray levels (or basins) surrounded by higher gray level structures (or ridges), as marrow space, in accordance to the spatial extend and gray levels of ridges. Therefore, the size and orientation of marrow space segments defined by this procedure can be related to the spacing, relative density and orientation of adjacent trabecular structures.
  • the segments of marrow space generated using the watershed segmentation can be measured for their area, eccentricity, orientation, and the average gray level on the x- ray image within the segment.
  • the statistics for example mean, standard deviation, minimum, maximum, and mode
  • These statistics can be selected to reflect the homogeneity of marrow space and trabecular structures, and can be used to detect presence of abnormal distribution of marrow space and trabecular structures. 3.1.3.0. Multidimensional Classification Schemes [0157]
  • novel approaches that integrate one or more suitable indices can be employed.
  • indices can be optimized and incorporated into a multi-dimensional classification scheme, for example using a nearest neighbor classification.
  • Table 1 provides examples of different analyses and anatomical / physiological correlates of the parameters that can be measured. TABLE 1
  • Mean pixel intensity is a general parameter for the bone mineral density. The degree to which x-rays passing through bone tissue are absorbed depends on the bone's mineral content. Bone with a higher mineral density absorbs a larger portion of x-rays, and therefore appears brighter on an x-ray image.
  • the mean pixel intensity f( ⁇ ,y) in the ROI is calibrated against an aluminum calibration wedge that is included in the image. The log of the average pixel intensity for each thickness level of the calibration wedge is plotted against the thickness, which allows f( ⁇ ,y) to be converted into a standardized aluminum thickness equivalent, which is used as the value for this parameter.
  • the automatic recognition of the different thickness levels of the calibration wedge are made possible by different geometric patterns scribed into the wedge which are shown in the x-ray image and can be localized automatically.
  • the spatial frequency spectrum of a texture provides information about its coarseness. Fine textural structures and edges in an image correspond to high frequencies in the frequency domain, while coarse textures are represented by lower frequencies. Applied to x-ray images of trabecular bone, this means that a region with coarse or little trabeculation should exhibit a Fourier spectral energy concentration at low spatial frequencies, whereas a region of fine trabecular structure should show a spectral energy concentration at high frequencies.
  • These 2-dimensional coefficients are used to determine a 1 -dimensional power spectrum F(u) by averaging all coefficients over circles with radii that correspond to the discrete spatial frequencies u.
  • the mean transform coefficient absolute value and the mean spatial first moment M of the absolute coefficients are determined after exclusion of the first ("DC") coefficient.
  • Mi provides a measure for which frequencies contribute most to the energy of the spectrum, similar to the "center of mass" of a geometric object.
  • Fractals are objects that exhibit certain statistical self-similar or self-affine properties, so that a portion of the object, scaled to its original size, has for example the same surface area (3-d) or the same perimeter (2-d) as the original object.
  • the gray values in a particular texture can be interpreted as an altitude, and the resulting 3-dimensional surface is analyzed (Fig. 11).
  • Fractal dimension (fd) is the rate at which the perimeter or surface area of an object increases as the measurement scale is reduced.
  • osteoporotic trabecular bone in which trabeculae become thinner and lose their continuity, and therefore complexity is increased, should have a higher fractal dimension than healthy bone.
  • ROI's 2-D power spectrum using a fast Fourier transform From the 2-D Fourier coefficients the 1-D power spectrum is produced as described above for the Fourier analysis.
  • FFT fast Fourier transform
  • the 1-D power spectrum is plotted as the logarithm of the power versus the logarithm of the frequency, it must have a negative slope of magnitude b with 1 ⁇ b ⁇ 3 according to fractal theory.
  • f(i,j) is the gray value of pixel (i,j) in the ROI.
  • the fractal dimension is then given by FD 2 -2 -V .
  • the gray scale region of interest is first binarized. As described in White et al. (1999) Oral Surg Oral Med Oral Patholo Oral Radiol Endod 88:628-635, this can be achieved in the following way: The ROI is blurred by means of a Gaussian filter. The blurred ROI is then subtracted from the original ROI, and the value 128 is added at each pixel location. This results in an image with a mean gray value of 128, which is also used as a threshold, resulting in an image, in which trabeculae are white and marrow space is black. [0171] From this binary image, the total number of white pixels represents the trabecular area, which is calculated as a percentage of the total ROI area.
  • the number of pixels on the outer trabecular border measures the peripheral length of the trabeculae. The same parameters can be measured for the marrow space by counting the black pixels. [0172] After skeletonization of the binary image, the total length of the trabeculae is determined by the total number white pixels. Furthermore, counts of the terminal points and of the branch points are expressed as a proportion of trabecular length. An estimate of the average length of the trabeculae is calculated as the ratio of total trabecular length and the sum of terminal points and branch points. 3.1.3.5. Markov Random Fields
  • Markov random fields can be used as models for osteoporosis detection from radiographic images and for fracture risk prediction.
  • osteoporosis is typically manifested in radiographic images by structural changes that can be used for computer-aided detection and characterization.
  • the detection and/or characterization of osteoporosis from radiographic images relies on the measurement and analysis of a feature or set of features relating to the density of the bone or trabecular structures present in an image.
  • Markov random fields can be used to analyze and detect structure density changes by either modeling particular joint feature distributions ( ⁇ F1 , F2, ... , Fn ⁇ ) as they are estimated at each image element or image neighborhood ( Figure 19), or by modeling the actual radiographic manifestation of particular structural definitions (e.g. trabeculae) ( Figure 20).
  • the Markov random field framework is used for a context-based feature analysis/discrimination approach which takes into account local relationships between the features and effectively compensating for space-varying processes (e.g. variable soft tissue or missing or incomplete data due to boundaries) that can affect the relative values of the features taken into account. (Buntine (1994) "Operations for learning with graphical models," J. Artificial Intelligence Res. December: 159-225).
  • This approach can also be used for predicting most likely fracture paths based on the analysis of trabecular structure nodes and their related feature sets by defining the most likely chains of joint feature sets.
  • Bayesian networks for example as described in Heckerman D (1996) "A tutorial on learning with Bayesian networks,” Microsoft Research Technical Report, MSR-TR-95- 06. based on available test case data.
  • Markov random fields can also be used to model the manifestations of the structures in an image in probabilistic terms, (Geman et al. (1984) "Stochastic relaxation, Gibbs distributions, and the Bayesian restoration of images," IEEE Transactions on Pattern Analysis and Machine Intelligence 6:721 -741 ; Besag (1986) “On the statistical analysis of dirty pictures,” Journal of the Royal Statistical Society, 48(3):259-302).
  • probabilistic models P(N) and P(l
  • a common assumption for the noise component in digital/digitized radiographs is to consider Normal or Poisson distributed pixels.
  • T) is such as to reflect that the corresponding probability distribution of the region I is conditional ( expressed by the symbol
  • the analysis tools for such a probabilistic framework are provided by the laws of probability and specifically Bayes' Rule shown in Figure 21.
  • Bayes' rule can be described as the rule according to which our knowledge about the presence of a given characteristic structure in an ROI is updated (a-posteriori information represented by the probability distribution P(T
  • Figure 21 illustrates that simply selecting the structure with the maximum a-posteriori information can be used as a decision criterion.
  • Markov random field modeling may be employed.
  • Markov random fields are specific multidimensional random processes that satisfy what is known as the Markov property.
  • the Markov property simply states that in a random series of events, each event can be predicted and depends only on a limited set of events. This property is convenient and intuitive for the modeling and analysis of structures in images. It basically states that if the distribution of pixels in an ROI can be modeled as having the Markov property, then in order to determine if a pixel belongs to a given structure, only a limited number of neighboring pixels are necessary.
  • cliques are the fundamental elements that can be used to reflect specific spatial distribution properties of a structure of interest, such as for example vertical, horizontal and diagonal geometries. Furthermore, the Markov property is manifested very conveniently as each image pixel can be expressed in terms of the cliques in a local neighborhood:
  • the model parameterization for the families of images characteristic of a particular structural density grade and definition of a priori information can be done either by estimation from available patient data thus defining empirical priors or by implementing physical and stochastic models that are based on the image generation process. 3.1.4.0. Overlapping windows processing
  • two or more overlapping ROIs can also be defined and used to analyze any given image.
  • bone density, microarchitecture , macro- anatomic and/or biomechanical (e.g. derived using finite element modeling) analyses can be applied within a region of predefined size and shape and position. This region of interest may also be referred to as a "window.” Processing can be applied repeatedly within the window at different positions of the image. For example, a field of sampling points may be generated and the analysis performed at these points (Fig. 16). The results of the analyses for each parameter can be stored in a matrix space, e.g., where their position corresponds to the position of the sampling point where the analysis occurred, thereby forming a map of the spatial distribution of the parameter (a parameter map).
  • the sampling field can have regular intervals or irregular intervals with varying density across the image.
  • the amount of overlap between the windows can be determined, for example, using the interval or density of the sampling points (and resolution of the parameter maps).
  • the density of sampling points is set higher in regions where higher resolution is desired and set lower where moderate resolution is sufficient, in order to improve processing efficiency.
  • the size and shape of the window would determine the local specificity of the parameter. Window size is preferably set such that it encloses most of the structure being measured. Oversized windows are generally avoided to help ensure that local specificity is not lost.
  • the shape of the window can be varied to have the same orientation and/or geometry of the local structure being measured to minimize the amount of structure clipping and to maximize local specificity.
  • both 2D and/or 3D windows may be used, depending on the nature of the image and data to be acquired.
  • bone density, microarchitecture, macro-anatomic and/or biomechanical (e.g. derived using finite element modeling) analyses can be applied within a region of predefined size and shape and position.
  • the region is generally selected to include most or all of the anatomic region under investigation and, preferably, the parameters can be assessed on a pixel-by-pixel basis (e.g., in the case of 2D or 3D images) or a voxel-by-voxel basis in the case of cross-sectional or volumetric images (e.g., 3D images obtained using MR and/or CT).
  • the analysis can be applied to clusters of pixels or voxels wherein the size of the clusters is typically selected to represent a compromise between spatial resolution and processing speed.
  • Each type of analysis may yield a parameter map.
  • Parameter maps can be based on measurement of one or more parameters in the image or window; however, parameter maps can also be derived using statistical methods. In one embodiment, such statistical comparisons can include comparison of data to a reference population, e.g. using a z-score or a T-score. Thus, parameter maps can include a display of z-scores or T-scores.
  • the parameter maps can represent individual parameters or combinations of parameters such as density, microarchitecture macro-anatomical parameters or biomechanical parameters, for example derived using finite element modeling, are useful in identifying regions or patches that have similar characteristics. For instance, depending on their position, shape, size, orientation, and extent particular regions or patches that exhibit similar characteristics (e.g., values at high or low ranges of the data set) typically represent regions of bone with different properties, for example areas of stronger or weaker areas. Therefore, parameter maps can be used to generate virtual fracture lines that aid in predicting areas of the bone that may be subject to an increased risk of fracture.
  • One or more parameter maps can be selected by statistical analysis of results from in vitro mechanical loading tests or by other means (e.g.
  • a multivariate regression model can be fitted to generate a composite parameter map derived from 2D or 3D data sets, e.g. x-rays, digital tomosynthesis, CT and MRI, using the techniques described herein and/or statistical methods known to those of skill in the art.
  • a parameter map can be used to predict the overall bone strength or fracture risk or fracture load by analyzing the predicted fracture paths.
  • a predicted fracture path is defined here as the hypothetical path where fracture would most likely to occur, if sufficient forces are applied in one or more particular directions.
  • a watershed segmentation can be applied to the selected or composite parameter map.
  • Watershed segmentation can be applied to 2D images as well as to 3D (cross-sectional or volumetric data obtained, for example, from CT or MR).
  • the boundaries of watershed segmentation generally form along the ridges on the parameter map, i.e., along the peak values.
  • the inverse value of the parameter is used to generate the watershed boundaries so that the boundaries would form along valleys (local minimum) of parameter maps.
  • the nodes of watershed boundaries can be identified and segmented to separate the watershed boundaries into segments ( Figure 17).
  • Each of these segments can be assigned a strength value or fracture load value which is a composite value of one or more parameter maps underlying the segment.
  • the length, orientation, and position of segments can be used as normalizing factors for the strength values.
  • the nodes and segments of the watershed boundaries may be labeled, traced, measured, and recorded in a form of data structure, for example, a graphical structure.
  • the strength values and interconnect relationships are also stored for each segment.
  • a search strategy for example, the depth-first search (Russell S., Norvig, P., Artificial Intelligence: A modern approach. 1995, NJ: Prentice Hall, pp.77), is propagated through the data structure to determine the paths of least resistance from one surface of the bone to another opposite surface restricted by a predefined solid angle.
  • an artificial neural network can be trained to predict fracture paths given the parameter maps as inputs.
  • Fracture risk prediction [0193] Having predicted one or more fracture paths, additional processing may be performed, typically with a new processing grid that has high concentration of nodes along the predicted fracture paths with a different window size and/or shape. Macro- anatomical parameters such as cortical thickness can be evaluated (in two or three dimensional images) with higher resolution at the exits of fracture paths. Parameters that are the best predictors of fracture risk can be evaluated along the predicted fracture paths. These parameters, including density, microarchitecture, macro-anatomical measurements and biomechanical parameters, are selected by statistical analysis of results from in-vitro mechanical loading test or by other means, e.g.
  • osteoporosis subjects in particular those developing fractures, for being highly correlated to the magnitude of one or more mechanical properties of bone, for example in one or more particular loading force directions, or for being highly correlated with fracture risk, incidence of new fractures or fracture loads.
  • the mechanical properties include but are not limited to yielding load, stiffness, and Young's modulus.
  • the values of parameters along the predicted fracture paths may be compared against the statistical distribution of the population.
  • the z-score and T-score of each parameter relates to the risk of fracture occurring in a particular predicted fracture path.
  • a fracture risk score can be assigned to that fracture path.
  • the predicted fracture paths can also be associated with the clinical definition of common fracture types.
  • the overall fracture risk can then be evaluated by weighing fracture risk score of each predicted fracture path with the probability of a particular type of fracture occurring.
  • Figure 18 depicts an exemplary summary of this process.
  • FEM Finite element modeling
  • FEM Finite element modeling
  • Structural finite element analysis a particular subset of FEM, is the calculation of the mechanical behavior (stress and strain) at any point within the structure under specific loading conditions.
  • the foundation of every finite element model is the two-dimensional or three-dimensional data of the object or structure
  • Examples of microarchitecture and micro-anatomical features that can be used as input mesh for finite element analysis include but are not limited to the actual and derivation of image or data structures of trabecular structures, image or data structures of cortical bone, image, data structures of trabecular skeleton or parameter maps derived from overlapping window processing.
  • the input features can be obtained from 2D and/or 3D images.
  • the application of simulated force can be in one or more directions, and is typically associated with the actual force components that would occur in a fracture incident.
  • the finite element analysis provides an estimate of load and direction of fracture for each fracture incident scenario. Fracture risk is estimated by weighing the fracture loads with the probability of each fracture scenario occurring.
  • the fracture paths estimated by finite element analysis can be used as inputs to the analysis of density, micro-architecture, macro- anatomical features.
  • density, micro-architecture, macro-anatomical features can be measured in areas of fracture paths predicted by finite element modeling.
  • finite element analysis can be combined with additional image and clinical data to determine fracture risk by predicting if the bone would fracture, given the force components that would occur in a fracture incident.
  • Bone fracture risk can be evaluated using one or a composite of more than one dependent or independent results of analysis or statistical methods. An example of this combination is the weighted average score of density, microarchitecture, macro-anatomical, finite element analysis and clinical risks factors such as weight, height, history of fracture, family history of fracture, and the like.
  • Finite element modeling can be applied to all of the bony structures included in an image. Preferably, however, finite element modeling is typically applied in selected subregions. In certain embodiments, finite element modeling is applied in areas coinciding with or bordering with the predicted fracture path, for example based on micro-structural or macro-anatomical measurements.
  • biomechanical assessment of bone properties with density micro-architectural and macro-anatomical assessment, the prediction of fracture risk and/or the correlation with fracture load can be improved.
  • regional assessment of biomechanical properties can also improve the accuracy of the fracture path prediction.
  • Biomechanical assessment can also include more traditional approaches estimating levers and forces at the macro-anatomical level, e.g.
  • the macroanatomical parameters that are measured can change depending on the region of interest to be measured. For example, when studying a portion of the spine, the user can combine bone structure measurements with macroanatomical measurements and/or FEA and/or other biomechanical measurements and/or bone mineral density.
  • the actual macroanatomical measurements that are used in the spine can be, for example, the inner pedicle distance, the outer pedicle distance, the vertebral height (either anterior, central, posterior, left, right, or a combination thereof), the vertebral anterior-posterior diameter (taken either in the superior, middle, inferior, or another location), the vertebral right to left diameter (taken in either the superior, middle, inferior or another location), the vertebral diameter (taken in an oblique plane), the vertebral diagonal (using, e.g., internal cortex or external cortex), the thickness of the superior endplate (taken, e.g., anteriorly, centrally, posteriorly, from the left, from the right, or a combination thereof), or using the thickness of the inferior endplate (again taken, e.g., anteriorly, posteriorly, from the left, from the right, or a combination thereof).
  • the user when studying the knee and tibia, can combine bone structure measurements with macroanatomical measurements and/or FEA and/or other biomechanical measurements and/or bone mineral density.
  • the bone structures used for measurements when studying the knee and tibia region change due to changes in anatomy.
  • suitable measurements are taken from, for example, the anterior-posterior diameter of the bone using the inner or outer cortex, or a combination thereof, the medial-lateral diameter of the bone using the inner or outer cortex, or a combination thereof, the cortical thickness in various locations, the standard deviation of cortical thickness, the subchondral bone thickness in various locations, and/or a combination thereof.
  • Variations in soft tissue thickness can be significant in analyzing and evaluating bone density and bone structure, macro-anatomical parameters and biomechanical parameters, e.g. those derived using finite element modeling, in x-rays. Accordingly, the invention also includes methods and devices for correcting for soft tissue in assessment of bone structure or dense tissue, particularly for diagnosing and/or predicting osteoporosis or other bone conditions.
  • the x-ray image is a dental x-ray image and such correction methods involve (a) interrogating at least a portion of a subject's mandible and/or maxilla with an x-ray detector; (b) producing an x-ray image of the interrogated mandible and/or maxilla; (c) obtaining data from the x-ray image regarding bone density or bone structure; (d) interrogating the surrounding soft tissue to determine soft tissue thickness; and (e) correcting the data obtained from the x-ray image by correcting for soft tissue thickness.
  • Such study groups include: non-osteoporotic premenopausal, non-osteoporotic postmenopausal, osteoporotic postmenopausal patients. It will be apparent, although exemplified with respect to dental x-rays, that many of the methods described herein can be applied to other x-ray images, e.g. hip or spine x-ray images.
  • Soft tissue thickness measured in a subject can also be compared to reference soft tissue thickness obtained from a control population (e.g. age-, sex-, race-, or weight-matched normal subjects).
  • Reference soft tissue thickness can be generated by measuring soft tissue thickness in healthy subjects with normal vascular, cardiac, hepatic, or renal function and no other underlying medical condition.
  • Reference soft tissue thickness can be expressed as but are not limited to, mean and standard deviation or standard error.
  • Reference soft tissue thickness can be obtained independently for patients 15-20, 20-30, 30-40, 40-50, 50-60, 60-70, 70-80, and 80 and more years of age and are preferably obtained separately for men and women and for race (e.g. Asian, African, Caucasian, and Hispanic subjects). Additionally, reference soft tissue thickness can be obtained for different subject weights within each age, sex, and racial subgroup.
  • correction factor can be applied.
  • the amount/magnitude of correction factor is influenced by the magnitude of increase in soft tissue thickness that can be influenced by the magnitude of fat, fibrous, and muscle tissue contribution.
  • Clinical study groups can be evaluated to generate databases for further study or to generate more refined correction factors.
  • Such study groups include: non-edematous non-osteoporotic premenopausal, non-edematous non-osteoporotic postmenopausal, non-edematous osteoporotic postmenopausal; edematous non- osteoporotic premenopausal, edematous non-osteoporotic postmenopausal, and edematous osteoporotic postmenopausal patients.
  • DXA dual x-ray absorptiometry
  • QCT quantitative computed tomography
  • correction for soft tissue thickness can also improve the accuracy and discriminatory power in the analysis of x-rays and other x- rays. Such methods can also be used to identify population with an increased or decreased risk of bone conditions such as osteoporosis. 4.0.
  • the measurements of bone m ineral density or trabecular architecture and/or macro-anatomical and/or biomechan cal parameters can be used to derive an assessment of bone health in any subject. Additionally, the analysis and manipulation of data from x- rays allows for the assessment of bone health that in turn can be used to prescribe a suitable treatment regime.
  • Efficacy of a treatment regime can also be assessed using the methods and devices described herein (for example, using measurements of bone mineral density or trabecular architecture and/or macro-anatomical and/or biomechanical parameters in the mandible or the maxilla or the hip or the spine taken at two separate time points T1 and T2 to detect any difference in bone mineral density or trabecular architecture).
  • the methods described herein permit, for example, fully automated assessment of the structural organization and architectural arrangement of trabecular bone and/or macro-anatomical and/or biomechanical parameters on standard hip radiographs as well as improved tools for monitoring progression of osteoporosis and therapeutic response.
  • the methods involve binarizing and skeletonizing trabecular bone using morphological operators with detection of branch points and endpoints of the skeleton network and classification into free-end segments and node-to-node segments.
  • the methods involve measuring trabecular density, trabecular perimeter, trabecular bone pattern factor, segment count, segment length, angle of segment orientation and ratio of node-to-node segments to free-end segments based on the binarized and/or skeletonized images.
  • the methods involve (a) measuring trabecular thickness using a Euclidean distance transform (see, also Example 3); (b) assessing trabecular orientation using a 2D Fast Fourier Transform; and/or (c) creating a bone structure index for diagnosing osteoporosis or for predicting fracture risk combining at least two or more of these structural parameters.
  • the radiograph is of a subject's hip.
  • the methods may include one or more of the following: evaluating the angular dependence of bone structure measurements in the hip, for example by comparing antero-posterior radiographs of the hip joint in healthy to osteoporotic patients (subjects) with the femur radiographs in neutral position and in various degrees of internal and external rotation or by obtaining radiographs of the hip with different degrees of tube angulation.
  • Bone structure and/or macro-anatomical and/or biomechanical measurements can be compared between the different positions to determine which bone structure parameters show the least dependence on radiographic positioning and/or using a foot holder to fix the patients' foot in neutral position in case pair wise coefficients of variation between the results for the 0° neutral position and a 15° internal or external rotation position exceed 10% for the majority of the structural parameters measured.
  • methods of monitoring bone structure and/or macro-anatomical and/or biomechanical parameters over time are also provided, for example to assess progression of osteoporosis and/or response to therapy.
  • the methods involve automated placement of regions of interest (ROI) in the hip joint, for example by creating and using a general model of the proximal femur that includes six defined regions of interest (ROI's).
  • ROI regions of interest
  • kits for obtaining information from images for example for obtaining information regarding bone structure, micro-architecture, macroanatomical and/or biomechanical parameters from an image such as a radiograph.
  • the kit comprises one or more computer (e.g., software) programs, for example for receiving, analyzing and generating reports based on the image(s).
  • the kits can include calibration phantoms, for example calibration phantoms integrated or attachable-to a holder, hygienic cover, x-ray film and/or x-ray film holders.
  • the invention also provides for therapeutic kits, for example for treating osteoporosis or dental disease.
  • kits comprise a calibration phantom for use with one or more x-ray films, a computer software product, a database, a therapeutic drug and, optionally, instructions for use (e.g., instructions regarding positioning the calibration phantom while taking the x-ray, using the software to analyze the x-ray, dosages and the like.
  • the therapeutic drug can be, for example, anti- resorptive or anabolic.
  • kits, methods and/or devices described herein are provided, for example using any of the kits, methods and/or devices described herein. It will be apparent that these methods are applicable to any bone-related disorder including, for example, osteoporosis, bone cancer, and the like, as well as to periodontal disease and implant failure.
  • Osteoporosis alone is a major public health threat for 25 million postmenopausal women and 7 million men. In 1995, national direct expenditures for osteoporosis and related fractures were $13 billion. Changing demographics, with the growth of the elderly population, steadily contribute to increasing numbers of osteoporotic fractures and an incipient and potentially economically unmanageable epidemic of osteoporosis. Projections put the total cost of osteoporosis in the United States alone at more than 240 billion dollars per year in 40 years. [0216] Less than 20% of the pat :iients know they have the disease and many fewer receive physician directed specif i ic therapy.
  • BMD bone mineral density
  • the methods comprise using a computer program to analyze bone mineral density or bone structure and/or macro-anatomical and/or biomechanical parameters of an image (e.g., x-ray image) and comparing the value or measurement obtained from the image with a reference standard or curve, thereby determining if the subject has a bone-related condition such as osteoporosis or thereby determining a subject's fracture risk.
  • the image can also include a calibration phantom, for example a calibration phantom as described herein.
  • measurements of bone structure can be combined or correlated with measurements of macro-anatomical and/or biomechanical parameters (e.g., cortical thickness on a hip x-ray), for example using statistical or mathematical methods, to create an index for the severity of the disease. Subsequently, the index can be used for diagnosing osteoporosis or for predicting fracture risk combining at least two or more of these bone structure or morphological parameters.
  • macro-anatomical and/or biomechanical parameters e.g., cortical thickness on a hip x-ray
  • the methods and devices described herein can also be used to develop an appropriate treatment regime for a subject in need thereof. Additionally, the invention allows for the ongoing analysis of the efficacy of a subject's treatment regime.
  • HRT hormone replacement therapy
  • SERMs selective estrogen receptor modulators
  • diagnosing, predicting, developing treatment regimes, assessing treatment efficacy and the like can be readily accomplished using the methods described herein.
  • these applications will be accomplished using algorithms or decision trees (also known as logic trees or flow charts).
  • One exemplary decision tree is provided in regard to predicting bone problems. It will be readily apparent that such decision trees are equally applicable to other applications (e.g., designing treatment regimes, assessing treatment efficacy, etc.).
  • One exemplary method for predicting bone problems employs a decision tree (also called classification tree) which utilizes a hierarchical evaluation of thresholds (see, for example, J.J. Oliver, et. al, in Proceedings of the 5th Australian Joint Conference on Artificial Intelligence, pages 361-367, A. Adams and L. Sterling, editors, World Scientific, Singapore, 1992; D.J. Hand, et al., Pattern Recognition, 31 (5):641-650, 1998; J.J. Oliver and D.J. Hand, Journal of Classification, 13:281-297, 1996; W. Buntine, Statistics and Computing, 2:63-73, 1992; L.
  • a simple version of such a decision tree is to choose a threshold bone structure and/or macro-anatomical and/or biomechanical or bone mineral density reading at a particular anatomical landmark (e.g., edge of mandible or maxilla, the end of a tooth root, etc.). If a value is equal to or below the threshold bone data value, then more of the image is evaluated. If more of the image is below the threshold value, then a bone problem, periodontal disease or implant failure is predicted.
  • a threshold bone structure and/or macro-anatomical and/or biomechanical or bone mineral density reading at a particular anatomical landmark e.g., edge of mandible or maxilla, the end of a tooth root, etc.
  • a first level decision is made by the algorithm based on the most recent x-ray images obtained and analyzed as described herein is compared to initial thresholds that may indicate an impending or current bone- or periodontal-related event.
  • the next level of the decision tree may be an evaluation of the subject's age and/or gender at time (n) that x-ray is taken, which is compared to a threshold bone measurement for "normal" subjects of that age and/or gender. For example, if the subject's bone measurement is greater than the threshold bone structure level for that particular age and/or gender, then a decision is made by the algorithm to prompt further monitoring in the future. If the information on bone structure is less than or equal to the threshold, then the algorithm continues with the next level of the decision tree.
  • the next level of the decision tree may be, for example, an evaluation of the subject's soft tissue (e.g., gum) thickness (n), which is compared to a threshold measurement. For example, if the soft tissue is significantly below or above the normal range of thickness, then a decision is made by the algorithm to examine more of the x- ray image or to predict a bone-related problem.
  • soft tissue e.g., gum
  • the decision tree could be further elaborated by adding further levels. For example, after a determination that a bone and/or periodontal events are possible, the subject can be x-rayed again to see if values have changed. Again, age, gender, weight, soft tissue thickness and the like can also be tested and considered to confirm the prediction.
  • the most important attribute is typically placed at the root of the decision tree.
  • the root attribute is the current bone structure measurement(s).
  • a predicted bone structure measurement at a future time point may be the root attribute.
  • bone mineral density and/or implant structure could be used as the root attribute.
  • thresholds need not (but can) be established a priori.
  • the algorithm can learn from a database record of an individual subject's readings and measurements.
  • the algorithm can train itself to establish threshold values based on the data in the database record using, for example, a decision tree algorithm.
  • a decision tree may be more complicated than the simple scenario described above. For example, if soft tissue of a particular subject is very thick, the algorithm may set a threshold for the bone measurements that is higher or lower than normal. [0232] By selecting parameters (e.g., current or future bone information, etc.) and allowing the algorithm to train itself based on a database record of these parameters for an individual subject, the algorithm can evaluate each parameter as independent or combined predictors of disease and/or implant failure. Thus, the prediction model is being trained and the algorithm determines what parameters are the most important indicators.
  • a decision tree may be learnt in an automated way from data using an algorithm such as a recursive partitioning algorithm.
  • the recursive partitioning algorithm grows a tree by starting with all the training examples in the root node.
  • the root node may be "split," for example, using a three-step process as follows.
  • the root node may be split on all the attributes available, at all the thresholds available (e.g., in a training database).
  • a criteria is applied (such as, GINI index, entropy of the data, or message length of the data).
  • An attribute (A) and a threshold (T) are selected which optimize the criteria. This results in a decision tree with one split node and two leaves.
  • Each example in the training database is associated with one of these two leaves (based on the measurements of the training example).
  • Each leaf node is then recursively split using the three-step process. Splitting is continued until a stopping criteria is applied.
  • An example of a stopping criteria is if a node has less than 50 examples from the training database that are associated with it.
  • the algorithm software can associate a probability with the decision. The probabilities at each level of decision can be evaluated (e.g., summed) and the cumulative probability can be used to determine whether disease and/or implant failure is predicted.
  • Receiver Operating Characteristic (ROC) curve analysis can be applied to decision tree analysis described above. ROC analysis is another threshold optimization means. It provides a way to determine the optimal true positive fraction, while minimizing the false positive fraction.
  • a ROC analysis can be used to compare two classification schemes, and determine which scheme is a better overall predictor of the selected event (e.g., evidence of osteoporosis); for example, a ROC analysis can be used to compare a simple threshold classifier with a decision tree.
  • ROC software packages typically include procedures for the following: correlated, continuously distributed as well as inherently categorical rating scale data; statistical comparison between two binormal ROC curves; maximum likelihood estimation of binormal ROC curves from set of continuous as well as categorical data; and analysis of statistical power for comparison of ROC curves.
  • ROC Accumetric Corporation, Montreal, Quebec, CA.
  • Related techniques that can be applied to the above analyses include, but are not limited to, Decision Graphs, Decision Rules (also called Rules Induction), Discriminant Analysis (including Stepwise Discriminant Analysis), Logistic Regression, Nearest Neighbor Classification, Neural Networks, and Na ⁇ ve Bayes Classifier.
  • Example 1 In vivo reproducibility and in vivo diagnostic sensitivity
  • A. Dental X-Rays [0237] In order to test in vivo reproducibility of data obtained from dental x-rays, the following experiment was performed. Subjects sat in a dental chair and an x-ray was taken of the area of the incisor teeth and of the molar teeth of the mandible. A calibration phantom step wedge was attached to the dental x-ray film. The dental x-ray film was exposed using standard x-ray imaging techniques for x-rays of the incisor area. The subjects walked around for 15 minutes at which point that test was repeated using the same procedure. [0238] X-ray films were digitized on a commercial flat-bed scanner with transparency option (Acer ScanPremio ST).
  • the regions of interest were placed manually at the same position with respect to the dental roots in all digitized x-rays of the same subject using the NIH Image software program (http://rsb.info.nih.gov/nih- image/Default.html).
  • COV coefficient of variation
  • Table 2 The data are summarized in Table 2.
  • the patient's arms were placed alongside their body. Patient comfort was ensured with a pillow underneath the patient's neck. However, no pillows were used underneath the knees.
  • the x-ray technologist checked that the patient lies straight on the table by looking from the head down towards the feet (which were placed in neutral position with the toes pointing up. The ray was centered onto the hip joint medial and superior to the greater trochanter.
  • Anteroposterior hip radiographs were acquired using the following parameters: Film-focus distance: 100 cm; tube voltage: 65 kVp; exposure: phototimer for automatic exposure or approximately 20 mAs for manual exposure; collimation: limited to the hip joint, including proximal femoral diaphysis; centering: over femoral head (see above); tube angulation: zero degrees.
  • An aluminum step wedge (BioQuest, Tempe, AZ) was included in the images to calibrate gray values before further image analysis. Processing was performed using ImageJ, a Java version of NIH image (http://rsb.info.nih.gov/ij/).
  • sample spine x-ray images will be acquired in more than one patient.
  • the bone structure parameters can be measured in the L1 , L2, L3 and L4 vertebral bodies unless obscured by superimposed ribs, iliac crest or bowel gas.
  • the first patient will provide control data provided the patient has normal bone mineral density in the spine.
  • spine BMD will be measured.
  • Regions of interest will be selected manually at the approximate locations as shown in Figure 22. Trabeculae will be extracted through background subtraction. In a next step, the trabecular bone in the selected regions of interest was skeletonized.
  • FIG. 24 depicts an example of an application of structure extraction and measurement for therapeutic monitoring using spine x-ray. White outline of extracted structure are show in (a) before treatment, and (b) after treatment.
  • sample x-ray images will be acquired in more than one patients.
  • the first patient will provide control data provided the patient has normal bone mineral density or bone structure in the tibia or femur.
  • joint BMD or bone structure will be measured.
  • Regions of interest will be selected manually at the approximate locations as shown in Figure 23.
  • the ROI can, for example, be the region immediately below the tibial plateau subchondral bone. Trabeculae will be extracted through background subtraction. In a next step, the trabecular bone in the selected regions of interest is skeletonized.
  • Knee/Tibial Radiographs - Arthritis To test whether bone texture analysis in knee and tibial x-rays can detect differences between normal patients and patients with arthritis, sample x-ray images will be acquired in more than one patients. The first patient will provide control data provided the patient has normal bone mineral density or bone structure in the tibia or femur. In the second patient and subsequent patients, joint BMD or structure will be measured.
  • Regions of interest will be selected manually at the approximate locations as shown in Figure 23.
  • the ROI can, for example, be the region immediately below the tibial plateau subchondral bone. Trabeculae will be extracted through background subtraction. In a next step, the trabecular bone in the selected regions of interest is skeletonized.
  • Example 2 Image Processing Techniques
  • Techniques to analyze structure of trabeculae in different regions of the femoral head, neck, and proximal shaft are developed in Matlab (The MathWorks, Inc., Natick, MA) on PC's.
  • the following techniques (modules) are developed: algorithms for software analysis of density, length, thickness, and orientation of trabeculae in different regions of interest (ROI) in the radiograph and a technique for automated placement of these ROI.
  • ROI regions of interest
  • the trabeculae in the femur is extracted using the background subtraction method, essentially as described in Geraets et al. (1998) Bone 22:165-173.
  • a copy of the image is blurred with a 15x15 Gaussian filter, and the result represents the non- uniform background.
  • This background image is subtracted from the original image to obtain an image of trabecular structure.
  • This image is then transformed into binary image of trabecular structure by applying a threshold value of 0.
  • An example of the end result is shown in Figure 10.
  • a second step parameters relevant to the geometry and connectivity of trabecular structure are measured on the trabecular skeleton or centerline.
  • the skeletonization is performed using morphological hit-or-miss thinning for example as described in Soille, "Morphological image analysis: principles and application” Springer, 1998: p. 129-154.
  • the branch points and end points of the skeleton network are detected, and the skeleton segments are classified as free-end segments and node-to- node segments.
  • trabecular density trabecular density
  • ratio of trabecular area to total ROI area trabecular perimeter
  • star volume Ikuta et al. (2000) J Bone Miner Res. 18:271- 277; Vesterby (1990) Bone 11 :149-155
  • trabecular bone pattern factor Hahn et al. (1992) Bone 13:327-330
  • Euclidean distance transform assessment of trabecular orientation using Fourier analysis; and orientation-specific trabecular assessment.
  • one or more of the following parameters can be measured in each ROI on the network of skeletonized trabeculae as a whole, all skeleton segments, and each type of segment: segment count; segment length; angle of segment orientation; and Interconnectivity Index (Legrand et al. (2000) J. Bone Miner Res. 15:13-19): normalized ratio of the number of node-to-node segments to free-end segments.
  • each pixel on the binarized trabeculae is assigned a value equal to its Euclidean distance from the structure boundary.
  • thicker trabeculae will have larger distance transform values in the center, thereby estimating trabecular thickness calculates the mean of the distance transform values along the trabecular skeleton (see Figure 11). Further, multiplying this value by 2 provides a measurement of trabecular thickness.
  • predominant trabeculae orientation may be evaluated using the
  • FFT 2D Fast Fourier Transform
  • a rectangular region is selected within each ROI and multiplied with a 2D Kaiser window before applying the transform (see Figure 12, left).
  • the log of the Fourier magnitude is taken to form an image representing the frequency domain of the ROI.
  • the result is then filtered with a 5x5 Gaussian filter to reduce local variation.
  • An example image is shown in Figure 12, center.
  • the Fourier image is subsequently thresholded at a fixed magnitude level. This binary image is resampled to a square image to normalize the length of the vertical and horizontal axes, and the direction and length of its major axis are determined ( Figure 12, right). The angles will be measured with respect to the axes of the femoral neck and shaft.
  • the axes are determined by fitting lines to the two longest segments of the centerline of the binarized femur (see also Figure 14).
  • the ROI's are located such that they include the different groups of compressive and tensile trabeculae in the proximal femur that each can be characterized by a specific direction.
  • a fully automated technique to evaluate the different quantitative structural parameters explained above for those trabeculae in each of the ROI that are oriented in the characteristic direction expected for the particular ROI is developed.
  • each trabecular skeleton segment is found through the gradient of the line fitted to the skeleton points. Based on this orientation information, only those trabeculae are considered in the evaluation of the structure parameters that are approximately oriented in the characteristic direction for a particular ROI.
  • Parameters include, for example, shaft angle, neck angle, diameter of the femur neck, the hip axis length, the largest cross-section of the femur head, the average thickness of the cortical region within a ROI, the standard deviation of cortical thickness within a ROI, or the maximum or minimum thickness of the cortical thickness within a ROI.
  • additional parameters to be considered include, for example, all parameters on vertical structures, all parameters on horizontal structures, vertebral cortical thickness, maximum vertebral height, minimum vertebral height, average vertebral height, anterior vertebral height, medial vertebral height, posterior vertebral height, maximum inter-vertebral height, minimum inter- vertebral height, and average vertebral height.
  • the knee and tibial region can be evaluated using the additional parameters of: average medial joint space width, minimum medial joint space width, maximum medial joint space width, average lateral joint space width, minimum lateral joint space width and maximum lateral joint space width.
  • Example 3 describes a number of parameters that are measured to assess trabecular structure in different regions of the proximal femur. In this Example, the different structural parameters are combined in each section, and a single index is determined over all regions of interest. [0265] A training set of hip x-ray images of a group of subjects are divided into the two categories “osteoporosis” and “no osteoporosis", based on previous DXA results.
  • a single scalar index value is calculated. All index values are combined into one n-dimensional feature vector.
  • the system is trained with the data from clinical validation studies with premenopausal, postmenopausal healthy and postmenopausal osteoporotic subjects.
  • the subject groups are preferably divided into a "fracture” and a "no fracture” category.
  • the feature vectors calculated from the x-ray images are used as prototype patterns.
  • a feature vector is calculated from the x-ray as calculated for the prototype patterns and an individual patient classified as category C if the majority of the k closest prototype patterns is of the category C.
  • the optimum scale for the different parameters is also preferably determined. However, for some parameters differences in the index values between the categories is smaller than for others. Also, the optimum k will be determined. Increasing k is expected to improve the accuracy of the classification, but it has to be smaller than the number of prototypes in each category. The exact percentage value of the majority of the k closest prototype patterns that determines the classification provides a measure for the reliability of the classification. The higher the percentage of prototype patterns from a particular category C, the more significant the information provided by the classification is likely to be.
  • Analysis of x-rays may be facilitated by development of techniques that locate one or more regions of interest (ROI) used for the calculation of the structural parameters of the trabecular bone.
  • ROI regions of interest
  • the general position of the femur can be located using a binary image of the hip radiographs thresholded at the appropriate gray value.
  • the femur is a bright structure extending from the pelvis.
  • Figure 13 By thresholding the digitized radiograph at the typical femur intensity value, a binary image showing the femur is produced.
  • the relatively thin structure of the femoral shaft can be extracted by applying a morphology operation on the binary image.
  • the morphological top-hat filter (opening subtracted from input) with an upright rectangular structuring element segments the femoral shaft.
  • the result is shown in Figure 13 with outline of the binarized femur superimposed on the original radiograph.
  • the region is cropped for further processing, preferably leaving enough room to include the femoral head.
  • a regularized active shape algorithm can be used (Behiels et al. (1999) Proceedings of the 2nd International Conference on Medical Image Computing and Computer-Assisted Intervention - MICCAI'99, Lecture notes in Computer Science 1679:128-137; Cootes (1994) Image and Vision Computing 12:355-366).
  • a general model of the proximal femur is created by manually outlining the shape in a training set of typical hip radiographs to form a mean shape.
  • the six predefined ROI are then embedded into this model. This mean model is scaled down 80%, isometrically along its centerline. This transformation is applied to the predefined ROI as well.
  • the outline of the rescaled model is then used as the initial template and is positioned within the proximal femur in the input image.
  • the control points of the contour are subsequently expanded outwards away from the nearest centerline point.
  • the energy function to be optimized in this iterative process can take into account local features, such as gradient, intensity, deviation from the mean model, and curvature of contour segments.
  • Figure 14 illustrates the propagation of the initial control points towards the femur edge.
  • a deformation field for the model area is calculated. This deformation field is interpolated for the model ROI inside the boundaries of the femur model.
  • the result is a new set of ROI that is adapted to the input image, but similar to the model ROI with respect to anatomical landmarks (see Figure 9).
  • Patients are selected into one of three groups: healthy premenopausal
  • PRE healthy postmenopausal
  • OSTEO osteoporotic postmenopausal women. All groups are studied by: (1) dental x-ray images of the periapical and canine region; (2) quantitative computed tomography of the spine and (3) hip; (4) dual x-ray absorptiometry of the spine and (5) hip; (6) single x-ray absorptiometry of the calcaneus, and (7) ultrasound of the calcaneus using standard techniques.
  • a diagnosis of osteoporosis is made when at least one atraumatic vertebral fracture as determined by a semi-quantitative assessment of morphologic changes of the thoracic and lumbar spine on lateral conventional x-rays is observed.
  • Odds ratios for 1SD change in the measured parameter
  • 95% confidence limits based on the age- adjusted logistic regression are calculated to measure the discriminative ability (for discriminating between the postmenopausal osteoporotic and the normal postmenopausal group) and the risk of osteoporotic fracture associated with the measured parameter.
  • the pairwise comparisons of the discriminative abilities are tested using age-adjusted receiver operating characteristic (ROC) curve analysis.
  • ROC receiver operating characteristic
  • Pairwise comparisons of all techniques are obtained by pooling all subjects (PRE, POST, OSTEO) and using Pearson's correlation coefficients (r), percent standard errors of the estimate (CV), and p-values for testing significance of correlations.
  • a kappa score analysis is performed on the normal postmenopausal women (POST) and the osteoporotic postmenopausal women (OSTEO). This is done by classifying every woman from the postmenopausal groups as osteopenic if her T-score with respect to the reference group (PRE) is less (or in case of structural parameters also greater) than 2.5.
  • the T-score for an individual woman and a particular measurement is defined as the measurement minus the mean measurement of young normals (PRE) divided by the SD of the measurement in the PRE group. Note that the T-score is measuring the position of an individual woman with respect to the PRE group and is different from the Student's t-value.
  • Example 7 Longitudinal Monitoring of Bone Structure
  • Algorithms and software to match follow-up dental x-rays obtained at a time point T 2 relative to baseline x-rays of the mandible obtained at an earlier time point Ti are developed.
  • bone structure parameters have to be measured at the same location of the mandible at different points in time.
  • ROI's regions of interest
  • an elastic matching step is preferably included.
  • the first step is a global affine transformation, for which the mutual information is used as a cost function.
  • Wells et al. (1996) Medical Image Analysis 1:35- 51.
  • the mutual information IM,N of two images M and N is defined as
  • the gray values occurring in the two images are regarded as random variables, and the mutual information provides a measure of the strength of the dependence between these variables, PM and PN are the distributions of M and N respectively, and p MN is the joint distribution of M and N.
  • PM and PN are the distributions of M and N respectively
  • p MN is the joint distribution of M and N.
  • Maintz et al. (1998) SPIE Medical Imaging - Image Processing. These distributions can be approximated from the marginal and joint gray value histograms, more accurately with the use of a Parzen window function.
  • Powell's method can be used as an optimization scheme to find the best affine transformation for N to match it with M. Press et al. ("Numerical Recipes in C.” 2nd edition, 1992, Cambridge University Press. [0280] This global transformation is followed by local elastic adjustments to improve the match.
  • m) are estimated from the joint histogram of the globally registered images.
  • the transformation vector field t(x) is then determined such that N(x-t(x)) is as similar to M(x) as possible by maximizing the local gray value correspondence, which for a fixed value of x is defined as
  • w is a window function whose width determines the size of the region that is used to compute t(x).
  • t a window function whose width determines the size of the region that is used to compute t(x).
  • Standard anteroposterior hip radiographs are obtained with the extremity at 30° internal rotation, 15° internal rotation, 0°, 15° external rotation, and 30° external rotation. These angles are achieved by placing the foot and ankle against a 30° or a 15° degree wedge in either internal or external rotation of the femur. The foot is secured against the wedge using Velcro straps.
  • the effect of positioning is assessed by calculating the pair wise coefficient of variation (CV%) between the results for the 0° position and the other positions for each individual subjects. The angular dependency will be expressed for each of the angles 30° internal rotation, 15° internal rotation, 15° external rotation, and 30° external rotation as the root-mean-square of these CV% values over all subjects. In general, parameters with the least dependency on angular positioning of the femur are selected.
  • a foot holder that fixes the patients' foot in neutral position can be used
  • the foot holder is designed with a base plate extending from the mid to distal thigh to the heel.
  • the base plate preferably sits on the x-ray table.
  • the patients' foot is positioned so that the posterior aspect of the heel is located on top of the base plate.
  • the medial aspect of the foot is placed against a medial guide connected rigidly to the base plate at a 90° angle.
  • a second, lateral guide attached to the base plate at a 90° angle with a sliding mechanism will then be moved toward the lateral aspect of the foot and will be locked in position as soon as it touches the lateral aspect of the foot.
  • the foot will be secured to the medial and lateral guide using Velcro straps. It is expected that the degree of involuntary internal or external rotation can be limited to less than 5° using this approach.
  • Example 9 Influence of X-Ray Tube Angulation on Bone Structural Measurements [0288] The effect(s) of the positioning of the x-ray tube on each parameter of the bone structure assessments is (are) examined. Dental x-rays are obtained in normal postmenopausal women and postmenopausal women with osteoporosis. The diagnosis of osteoporosis is made when at least one atraumatic vertebral fracture as determined by a semi-quantitative assessment of morphologic changes of the thoracic and lumbar spine on lateral conventional radiographs is observed. See, also, Genant et al. (1993) J. Bone Miner Res. 8:1137-1148.
  • Standard anteroposterior dental radiographs are obtained in the incisor region of the mandible.
  • the x-ray tube is aligned with an angle of 0°, 10°, 20°, 30°, and - 10°, - 20°, and -30° relative to the dental x-ray film. These angles are achieved with use of a goniometer applied to the metal tube located in front of the dental x-ray tube.
  • the dental x-ray film is positioned at the posterior mandibular wall in the incisor region.
  • a mechanical alignment system is then applied to the Rinn holder.
  • an extension tubing is attached to the Rinn holder.
  • the extension tubing is designed so that its inner diameter is slightly greater (and fits over) than the outer diameter of the dental x-ray system metal tube (Fig. 15).
  • the dental x-ray system metal tube is then inserted into the extension tubing attached to the Rinn holder that reduces alignment error of the x-ray tube relative to the x-ray film.
  • One group of patients then undergo two x-rays each of the incisor region. The results indicate that the short-term in-vivo reproducibility error of dental bone density and bone structure measurements is reduced with use of the mechanical alignment system by reducing x-ray tube angulation relative to the dental film and the anatomic landmarks in the mandible.
  • An x-ray image of a mandible or a hip or spine or other bone is analyzed using a computer program capable of assessing bone density, bone structure, macro- anatomical parameters, or biomechanical parameters, for example as described above.
  • the computer program derives a measurement of one or more bone density, bone structure, macro-anatomical or biomechanical parameters of the trabecular bone.
  • the measurement of the parameter(s) is compared against a database containing information on said one or more parameters in normal, healthy age-, sex-, and race matched controls. If the patient's measurement differs by more than 2 standard deviations from the age-, sex-, and race matched mean of normal, healthy subjects, a report is sent to the physician who then selects a therapy based on the measurement(s).
  • Example 11 Measurement of Bone Density, Bone Structure, Macro-Anatomical Parameters and Biomechanical Parameters and ⁇ lonitoring Therapy
  • One or more x-ray images are obtained from a patient undergoing therapy for osteoporosis, for example using an anabolic or an antiresorptive drug at two different time points T1 and T2.
  • the x-rays are analyzed using a computer program capable of assessing bone density, bone structure, macro-anatomical parameters, or biomechanical parameters.
  • the computer program derives a measurement of one or more parameters of the bone for both time points T1 and T2.
  • the measurement of the bone density, bone structure, macro-anatomical, or biomechanical parameter(s) at T1 and T2 is compared against a database containing information on said one or more parameters in normal, healthy age-, sex-, and race matched controls for each time point. If the results indicate that the patient has lost 5% or more bone between time points T1 and T2 despite therapy, a physician selects a different, more aggressive therapy.
  • Example 12 Measurement of Macro-anatomical and/or biomechanical Parameters
  • a hip radiograph is obtained using standard techniques and including a calibration phantom as described herein.
  • the reference orientation of hip x-rays is the average orientation of the femoral shaft.
  • a global gray level thresholding is performed using a bi-modal histogram segmentation algorithm on the hip x-ray generates a binary image proximal femur.
  • Edge-detection of the hip x-ray can be used.
  • edge-detection methods are further refined by obtaining breaking edges detected into small segments and characterizing the orientation of each segment, thereby obtaining the outline of proximal femur.
  • Each edge segment is then referenced to a map of expected proximal femur edge orientation and to a map of probability of edge location. Edge segments that do not conform to the expected orientation or are in low probability regions are removed. Morphology operations are applied onto the edge image to connect edge discontinuities.
  • the edge image forms an enclosed boundary of the proximal femur.
  • the region within the boundary is then combined with the binary image from global thresholding to form the final mask of the proximal femur.
  • edge detection is applied.
  • Morphology operations are applied to connect edge discontinuities. Segments are formed within enclosed edges. The area and major axis length of each segments are then measured. The regions are also superimposed on the original gray level image and the average gray level within each region is measured.
  • the cortex is identified as the segments that are connected to the boundary of the proximal femur mask, that has the greatest area, longest major axis length and has a mean gray level above the average gray level of all enclosed segments within the proximal femur mask.
  • the segment identified as cortex is then skeletonized.
  • the orientation of the cortex skeleton is verified to conform to the orientation map of proximal femur edge.
  • Euclidian distance transform is applied to the binary image of the segment.
  • the values of distance transform value along the skeleton are sampled and statistics (average, standard deviation, minimum, maximum and mod) measured.
  • measurements of macroanatomical parameters described here can be applied to hip, spine or knee radiographs with modifications to adapt to the shape, scale and location of macro-anatomical features specific to the anatomical region.

Abstract

The present invention relates to methods and devices for analyzing x-ray images. In particular, devices, methods and algorithms are provided that allow for the accurate and reliable evaluation of bone structure and macro-anatomical parameters from x-ray images.

Description

METHODS FOR THE COMPENSATION OF IMAGING TECHNIQUE IN THE PROCESSING OF RADIOGRAPHIC IMAGES
FIELD OF THE INVENTION
[0001] The present invention is in the field of imaging and analysis thereof. In particular, methods and compositions for accurately analyzing images to determine bone mineral density and/or bone structure are described.
BACKGROUND OF THE INVENTION
[0002] Osteoporosis is a condition that affects millions of Americans.
Osteoporosis refers to a condition characterized by low bone mass and microarchitectural deterioration of bone tissue, with a consequent increase of bone fragility and susceptibility to fracture. Osteoporosis presents commonly with vertebral fractures or hip fractures due to the decrease in bone mineral density and deterioration of structural properties and microarchitecture of bone.
[0003] Imaging techniques are important diagnostic tools, particularly for bone related conditions. Currently available techniques for the noninvasive assessment of the skeleton for the diagnosis of osteoporosis or the evaluation of an increased risk of fracture include dual x-ray absorptiometry (DXA) (Eastell et al. (1998) NewEnglJ. Med 338:736-746); quantitative computed tomography (QCT) (Cann (1988) Radiology 166:509-522); peripheral DXA (pDXA) (Patel et al. (1999) J Clin Densitom 2:397-401); peripheral QCT (pQCT) (Gluer et. al. (1997) Semin Nucl Med 27:229-247); x-ray image absorptiometry (RA) (Gluer et. al. (1997) Semin Nucl Med 27:229-247); and quantitative ultrasound (QUS) (Njeh et al. "Quantitative Ultrasound: Assessment of Osteoporosis and Bone Status" 1999, Martin-Dunitz, London England; U.S. Patent No. 6,077,224, incorporated herein by reference in its entirety). (See, also, WO 9945845; WO 99/08597; and U.S. Patent No. 6,246,745).
[0004] DXA of the spine and hip has established itself as the most widely used method of measuring B D. Tothill, P. and D.W. Pye, (1992) BrJ Radiol 65:807-813. The fundamental principle behind DXA is the measurement of the transmission through the body of x-rays of 2 different photon energy levels. Because of the dependence of the attenuation coefficient on the atomic number and photon energy, measurement of the transmission factors at 2 energy levels enables the area densities (i.e., the mass per unit projected area) of 2 different types of tissue to be inferred. In DXA scans, these are taken to be bone mineral (hydroxyapatite) and soft tissue, respectively. However, it is widely recognized that the accuracy of DXA scans is limited by the variable composition of soft tissue. Because of its higher hydrogen content, the attenuation coefficient of fat is different from that of lean tissue. Differences in the soft tissue composition in the path of the x-ray beam through bone compared with the adjacent soft tissue reference area cause errors in the BMD measurements, according to the results of several studies. Tothill, P. and D.W. Pye, (1992) BrJ Radiol, 65:807-813; Svendsen, O.L., et al., (1995) J Bone Min Res 10:868-873. Moreover, DXA systems are large and expensive, ranging in price between $75,000 and $150,000. [0005] Quantitative computed tomography (QCT) is usually applied to measure the trabecular bone in the vertebral bodies. Cann (1988) Radiology 166:509-522. QCT studies are generally performed using a single kV setting (single-energy QCT), when the principal source of error is the variable composition of the bone marrow. However, a dual-kV scan (dual-energy QCT) is also possible. This reduces the accuracy errors but at the price of poorer precision and higher radiation dose. Like DXA, however, QCT are very expensive and the use of such equipment is currently limited to few research centers.
[0006] Quantitative ultrasound (QUS) is a technique for measuring the peripheral skeleton. Njeh et al. (1997) Osteoporosis Int 7:7-22; Njeh et al. Quantitative Ultrasound: Assessment of Osteoporosis and Bone Status. 1999, London, England: Martin Dunitz. There is a wide variety of equipment available, with most devices using the heel as the measurement site. A sonographic pulse passing through bone is strongly attenuated as the signal is scattered and absorbed by trabeculae. Attenuation increases linearly with frequency, and the slope of the relationship is referred to as broadband ultrasonic attenuation (BUA; units: dB/MHz). BUA is reduced in patients with osteoporosis because there are fewer trabeculae in the calcaneus to attenuate the signal. In addition to BUA, most QUS systems also measure the speed of sound (SOS) in the heel by dividing the distance between the sonographic transducers by the propagation time (units: m/s). SOS values are reduced in patients with osteoporosis because with the loss of mineralized bone, the elastic modulus of the bone is decreased. There remain, however, several limitations to QUS measurements. The success of QUS in predicting fracture risk in younger patients remains uncertain. Another difficulty with QUS measurements is that they are not readily encompassed within the WHO definitions of osteoporosis and osteopenia. Moreover, no intervention thresholds have been developed. Thus, measurements cannot be used for therapeutic decision-making.
[0007] There are also several technical limitations to QUS. Many devices use a foot support that positions the patient's heel between fixed transducers. Thus, the measurement site is not readily adapted to different sizes and shapes of the calcaneus, and the exact anatomic site of the measurement varies from patient to patient. It is generally agreed that the relatively poor precision of QUS measurements makes most devices unsuitable for monitoring patients' response to treatment. Gluer (1997) J Bone Res 12:1280-1288.
[0008] Radiographic absorptiometry (RA) is a technique that was developed many years ago for assessing bone density in the hand, but the technique has recently attracted renewed interest. Gluer et al. (1997) Semin Nucl Med 27:229-247. With this technique, BMD is measured in the phalanges. The principal disadvantage of RA of the hand is the relative lack of high turnover trabecular bone. For this reason, RA of the hand has limited sensitivity in detecting osteoporosis and is not very useful for monitoring therapy-induced changes.
[0009] Peripheral x-ray absorptiometry methods such as those described above are substantially cheaper than DXA and QCT with system prices ranging between $15,000 and $35,000. However, epidemiologic studies have shown that the discriminatory ability of peripheral BMD measurements to predict spine and hip fractures is lower than when spine and hip BMD measurements are used. Cummings et al. (1993) Lancet 341:72- 75; Marshall et al. (1996) BrMedJ 312:1254-1259. The main reason for this is the lack of trabecular bone at the measurement sites used with these techniques. In addition, changes in forearm or hand BMD in response to hormone replacement therapy, bisphosphonates, and selective estrogen receptor modulators are relatively small, making such measurements less suitable than measurements of principally trabecular bone for monitoring response to treatment. Faulkner (1998) J din Densitom 1:279- 285; Hoskings et al. (1998) N Engl J Med 338:485-492. Although attempts to obtain information on bone mineral density from dental x-rays have been attempted (See, e.g., Shrout et al. (2000) J. Periodonol. 71 :335-340; Verhoeven et al. (1998) Clin Oral Implants Res 9(5):333-342), these have not provided accurate and reliable results.
[0010] Furthermore, current methods and devices do not generally take into account bone structure analyses. See, e.g., Ruttimann et al. (1992) Oral Surg Oral Med Oral Pathol 74:98-110; Southard & Southard (1992) Oral Surg Oral Med Oral Pathol 73:751- 9; White & Rudolph, (1999) Oral Surg Oral Med Oral Pathol Oral Radiol Endod 88:628- 35.
[0011] Thus, although a number of devices and methods exist for evaluating bone, there are a number of limitations on such devices and methods. Consequently, the inventors have recognized the need, among other things, to provide methods and compositions that result in the ability to obtain accurate bone mineral density and bone structure information from images (e.g., radiographic images) and data. SUMMARY OF THE INVENTION
[0012] In one aspect, the disclosure provides a method to derive information regarding one or more bone parameters from an image, the method comprising the steps of: (a) obtaining an image comprising bone from a subject; (b) defining two or more regions of interest (ROIs) in the image; and (c) analyzing a plurality of positions in the ROIs to determine one or more parameters selected from the group consisting of bone microarchitecture, bone macro-anatomy, biomechanical parameters and combinations thereof of the ROIs. In certain embodiments, the ROIs are overlapping. The positions analyzed in the ROIs may be at regular intervals relative to one another or, alternatively, may be irregularly spaced relative to each other. Thus, in certain embodiments, the methods involve determining bone micro-architecture, for example by analyzing positions at regular intervals. In other embodiments, the methods involve determining bone macro-anatomy, for example by analyzing positions at irregular intervals in the image.
[0013] In any of the methods described herein, the image can be two- dimensional (2D) or three-dimensional (3D). The images may be x-rays, MRI images, CAT scan images, or any other image including bone. In any of the methods, the image may be an electronic image.
[0014] In any of the methods described herein, the subject can be, for example, an osteoporosis subject. [0015] In another aspect, this disclosure relates to a method of generating a map of one or more bone parameters, the method comprising the steps of (a) obtaining information on bone parameters according to the method of any of methods described herein; and (b) identifying regions of the image that exhibit similar parameter characteristics, thereby creating a parameter map of the image. [0016] In yet another aspect, a method of predicting a fracture path in a subject is provided, the method comprising the steps of: (a) generating multiple parameter maps according to any of the methods of generating parameters maps described herein; (b) generating a composite parameter map from the multiple parameters maps of step (a); and (c) analyzing the composite parameter map to identify possible fracture paths. [0017] In yet another aspect, the invention includes a method of predicting a fracture path in a subject, the method comprising the steps of: (a) analyzing of one or more parameter maps preparing according to any of the methods described herein, wherein the analysis is watershed segmentation analysis or Markov random field analysis; and (c) identifying possible fracture paths based on the analysis of step (a), thereby predicting a fracture path in the subject.
[0018] In another aspect, the invention includes a method of predicting the risk of fracture in a subject, the method comprising the steps of: (a) generating a finite element model from one or more parameter maps obtained according any of the methods described herein; (b) applying simulated force vectors that would occur during a fracture incident to the finite element model generated in step(s); and (c) determining the minimum forces required for fracture to occur, thereby estimating the risk of fracture.
[0019] In a still further aspect, the invention includes a method of determining the risk of fracture in a subject comprising: (a) predicting a fracture path according to any of the methods of predicting fracture path as described herein; (b) evaluating one or more selected bone parameters along the predicted fracture path, thereby estimating the risk of fracture.
[0020] In another aspect, the invention includes a method of treating a subject with bone disease comprising (a) obtaining an image from a subject; (b) analyzing the image obtained in step (a) using any of the methods described herein; (c) diagnosing a bone disease based on the analysis of step (b); and (d) selecting and administering a suitable treatment to said subject based on said diagnosis.
BRIEF DESCRIPTION OF THE DRAWINGS
[0021] FIG. 1 shows an example of a dental x-ray. A calibration phantom 110 is seen. Regions of interest 120 have been placed for measurement of bone mineral density or structure.
[0022] FIG. 2 shows another example of a dental x-ray. A calibration phantom
110 is seen. Regions of interest 120 have been placed for measurement of bone mineral density or structure.
[0023] FIG. 3 shows an example of an analysis report resulting from a measurement of mandibular or maxillary bone mineral density. A subject (X) is more than one standard deviation below the mean of age-matched controls (x-axis age, y- axis arbitrary units BMD).
[0024] FIG. 4 shows an example of a V-shaped calibration phantom 110 mounted on a tooth 120. Gums are also shown 130. [0025] FIG. 5 shows an example of a holder 115 for a calibration phantom 110.
The holder 115 is mounted on a tooth 120. Gums are also shown 130.
[0026] FIG. 6, panels B through E shows gray value profiles along different rows of pixels used for locating dental apices. From top to bottom, the characteristic peaks for the dental roots (shown in dental x-ray panel A) gradually disappear. [0027] FIG. 7 shows a Hough transform (panel A) of a test image (panel B). All collinear points from the same line are transformed into sinusoidal curves that intersect in a single point (circles).
[0028] FIG. 8 shows a Hough transform (panel A) of a skeletonized trabecular bone x-ray image (panel B). The white regions in panel A indicate longer segments and predominant angles.
[0029] FIG. 9 shows the effect of varying size of structuring element E2; calibration phantom image with lines of varying width (1 , 3, 5, 7, 9, 11 , 13 pix) (top left); skeleton operation performed using E2 with a diameter of 3 pix (top right), 7 pix (bottom left), and 11 pix (bottom right), respectively. [0030] FIG. 10 shows the effect of varying size of structuring element E; gray scale image of trabecular bone (top left, panel A); skeleton operation performed using E2 with a diameter of 3 pix (top right, panel B); 7 pix (bottom left, panel C) and 11 pix (bottom right, panel D), respectively.
[0031] FIG. 11 shows gray value surface plot of an anatomical region of interest from a dental x-ray (inset) used for fractal analysis.
[0032] FIG. 12 shows an example of a hygienic cover holder that includes compartments for a calibration phantom and a fluid-filled bolus back. [0033] FIG. 13 shows an example of an anatomical region of interest (black dot), determined relative to the teeth or to the convexity/concavity of the mandible.
[0034] FIG. 14 shows an example of three anatomical region of interests (black dots), determined relative to the teeth or to the convexity/concavity of the mandible. [0035] Fig. 15 is a side view of an exemplary system for minimizing tube angulation as described herein. In the Figure, the system is shown as a dental x-ray system. An extension tubing (200) is attached to a ring-shaped Rinn holder (102). The outer diameter of the extension tubing is slightly smaller than the inner diameter of the tube located in front of the dental x-ray system/dental x-ray tube. The extension tubing can then be inserted into the metal tube thereby reducing tube angulation and resultant errors in bone apparent density and bone structural measurements.
[0036] FIG. 16 depicts an example of a regular interval sampling field for microarchitecture (+) and a higher density sampling field for macro-anatomical features(*) on a femur radiograph. White rectangles are examples of overlapping window positioning.
[0037] FIG. 17 depicts watershed segmentation boundaries superimposed on a parameter map. The two white lines are the actual fracture paths resulted from an in- vitro mechanical loading test.
[0038] FIG. 1 is a flowchart depicting an exemplary process to determine fracture risk using overlapping window processing and fracture paths prediction.
[0039] FIG. 19 depicts a Markov random field analysis by modeling particular joint feature distributions as they are estimated at each image element or image neighborhood.
[0040] FIG. 20 depicts an exemplary model definition for trabecular pattern density characterization in a region of interest (ROI) with a noise model P(N) and characteristic structure pattern given a density level P(l | Ti).
[0041] FIG. 21 depicts exemplary Bayes' Rule analysis. [0042] FIG. 22 depicts an example of a regular interval sampling field for microarchitecture (+) and a higher density sampling field for macro-anatomical features (*) on a spine radiograph. White rectangles are examples of overlapping window positioning. [0043] FIG. 23 depicts an example of a sampling field of varying density for microarchitecture (+, x, diamond) and a regular sampling field for macro-anatomical features(*) on a knee radiograph. White rectangles are examples of overlapping window positioning.
[0044] FIG. 24 depicts an example of an application of structure extraction and measurement for therapeutic monitoring using spine x-ray. White outline of extracted structure are show in (a) before treatment, and (b) after treatment.
DETAILED DESCRIPTION OF THE INVENTION
[0045] The following description is presented to enable any person skilled in the art to make and use the invention. Various modifications to the embodiments described will be readily apparent to those skilled in the art, and the generic principles defined herein can be applied to other embodiments and applications without departing from the spirit and scope of the present invention . Thus, the present invention is not intended to be limited to the embodiments shown, but is to be accorded the widest scope consistent with the principles and features disclosed and shown herein. To the extent necessary to achieve a complete understanding of the invention disclosed, the specification and drawings of all issued patents, patent publications, and patent applications cited in this application are incorporated herein by reference.
[0046] The practice of the present invention employs, unless otherwise indicated, conventional methods of imaging and image processing within the skill of the art. Such techniques are explained fully in the literature. See, e.g., WO 02/22014, incorporated herein in its entirety by reference; X-Ray Structure Determination: A Practical Guide, 2nd Edition, editors Stout and Jensen, 1989, John Wiley & Sons, publisher; Body CT: A Practical Approach, editor Slone, 1999, McGraw-Hill publisher; The Essential Physics of Medical Imaging, editors Bushberg, Seibert, Leidholdt Jr & Boone, 2002, Lippincott, Williams & Wilkins; X-ray Diagnosis: A Physician's Approach, editor Lam, 1998 Springer-Verlag, publisher; and Dental Radiology: Understanding the X-Ray Image, editor Laetitia Brocklebank 1997, Oxford University Press publisher.
[0047] Methods and compositions useful in analyzing images are described. In particular, the invention includes methods of obtaining and/or deriving information about bone mineral density and/or bone structure from an image. Additionally, the present invention relates to the provision of accurate calibration phantoms for use in determining bone structure and methods of using these calibration phantoms. In particular, the present invention recognizes for the first time that errors arising from misplacement of interrogation sites in dental or hip x-rays of bone density and/or bone structure can be corrected by positioning the x-ray tube, the detector and/or the calibration reference with respect to an anatomical landmark (or anatomical region of interest). [0048] Advantages of the present invention include, but are not limited to, (i) providing accessible and reliable means for analyzing x-rays; (ii) providing non-invasive measurements of bone structure and architecture and macro-anatomy; (iii) providing methods of diagnosing bone conditions (e.g., osteoporosis, fracture risk); (iv) providing methods of treating bone conditions; and (iv) providing these methods in cost-effective manner. 1.0. Obtaining Data from Images
[0049] An image can be acquired using well-known techniques from any local site. Non-limiting examples of imaging techniques suitable for acquiring images from which data can be obtained include, ultrasound, CAT scan, MRI and the like. See, also, "Primer of Diagnostic Imaging," 3rd edition, eds. Weissleder et al. (2002), Mosby Press; and International Publication WO 02/22014.
[0050] In certain aspects, 2D planar x-ray imaging techniques are used. 2D planar x-ray imaging is a method that generates an image by transmitting an x-ray beam through a body or structure or material and by measuring the x-ray attenuation on the other side of said body or said structure or said material. 2D planar x-ray imaging is distinguishable from cross-sectional imaging techniques such as computed tomography or magnetic resonance imaging. If the x-ray image was captured using conventional x- ray film, the x-ray can be digitized using any suitable scanning device. Digitized x-ray images can be transmitted over a networked system, e.g. the Internet, into a remote computer or server. It will be readily apparent that x-ray images can also be acquired using digital acquisition techniques, e.g. using photostimulable phosphor detector systems or selenium or silicon detector systems, the x-ray image information is already available in digital format which can be easily transmitted over a network. In other embodiments, 3D images are acquired, for example, using 3D imaging techniques and/or by creating 3D images from 2D images.
[0051] Any images can be used including, but not limited to, digital x-rays and conventional x-ray film (which can be digitized using commercially available flatbed scanners). In certain embodiments, the x-ray is of the hip region, for example performed using standard digital x-ray equipment (Kodak DirectView DR 9000, Kodak, Rochester, NY). Patients are typically positioned on an x-ray table in supine position, parallel to the long axis of the table, with their arms alongside their body. The subject's feet may be placed in neutral position with the toes pointing up or in internal rotation or may be placed in a foot holder such that the foot in a neutral position (0° rotation) or in any desired angle of rotation (e.g., internal or external) relative to neutral (see, also Example 8 below). Foot holders suitable for such purposes may include, for example, a base plate extending from the foot, for example, from the mid to distal thigh to the heel. The base plate preferably sits on the x-ray table. The patients' foot is positioned so that the posterior aspect of the heel is located on top of the base plate. The medial aspect of the foot is placed against a medial guide connected rigidly to the base plate at a 90° angle by any suitable means (e.g., straps, velcro, plastic, tape, etc.). A second, lateral guide attached to the base plate at a 90° angle with a sliding mechanism can then be moved toward the lateral aspect of the foot and be locked in position, for example when it touches the lateral aspect of the foot. The use of a foot holder can help improve the reproducibility of measurements of bone structure parameters or macro-anatomical and/or biomechanical parameters.
[0052] As will be appreciated by those of skill in the art, the patient or subject can be any warm-blooded animal. Typically patients, or subjects, are chosen from the class Mammalia. Thus, for example, patients, or subjects, would include humans and nonhuman primates such as chimpanzees and other apes and monkey species; farm animals such as cattle, sheep, pigs, goats and horses; domestic mammals such as dogs and cats; laboratory animals including rodents such as mice, rats and guinea pigs, and the like. To the extent desirable, other non-mammals can be subjected to the protocols described herein without departing from the scope of the invention,
[0053] Persons of skill in the art will appreciate that macro-anatomical parameters generally describe the shape, size or thickness of bone and/or surrounding structure. Oftentimes the typical parameters are, but need not be, greater than 0.5mm in size in at least one dimension. Generally, in the hip joint, macro-anatomical parameters include thickness of the femoral shaft cortex, thickness of the femoral neck cortex, cortical width, hip axis length, CCD (caput-collum-diaphysis) angle, neck-shaft angle and width of the trochanteric region. In the spine, macro-anatomical parameters include thickness of the superior and inferior endplate, thickness of the anterior, lateral and posterior vertebral walls, diameter and height of the vertebral body, dimensions of the spinal canal and the posterior elements.
[0054] Generally, the ray is centered onto the hip joint medial and superior to the greater trochanter. A calibration phantom, such as an aluminum step wedge may also be included in the images to calibrate gray values before further image analysis. [0055] In other embodiments, dental x-rays are preferred because of the relative ease and lack of expense in obtaining these images. Further, the mandible and maxilla are primarily composed of trabecular bone. Since the metabolic turnover of trabecular bone is approximately eight times greater than that of cortical bone, areas of predominantly trabecular bone such as the vertebral body are preferred sites for measuring bone mineral density. Lang et al. (1991) Radiol Clin North Am 29:49-76. Thus, trabecular bone is clearly visible on the dental x-ray image, thus facilitating quantitative analysis of bone mineral density and structure. Jeffcoat et al. (2000) Periodontol 23:94-102; Southard et al. (2000) J Dent Res 79:964-969. Further, the earliest bone loss in osteoporosis patients occurs in areas of trabecular bone. Multiple dental x-ray images are commonly made in most Americans throughout life. Indeed, there are approximately 750 million U.S. dental visits annually and 150 million of these patients result in more than 1 billion dental x-rays taken each year. Thus, the ability to diagnose osteoporosis on dental x-rays would be extremely valuable since it would create the opportunity for low-cost mass screening of the population.
[0056] Preferably, x-ray imaging is performed using standard x-ray equipment, for instance standard dental x-ray equipment (e.g. General Electric Medical Systems, Milwaukee, Wl). X-rays of the incisor region and canine region are acquired using a standard x-ray imaging technique with 80 kVp and automatic exposure using a phototimer or using a manual technique with 10mA tube current. X-ray images are acquired, for example, on Kodak Ultraspeed film (Kodak, Rochester, NY). X-ray images may be digitized using a commercial flatbed scanner with transparency option (Acer ScanPremio ST). Similarly, other imaging techniques are typically performed using standard equipment, for instance, MRI or CAT equipment. 1.1. Calibration Phantoms
[0057] It is highly preferred that the images include accurate reference markers, for example calibration phantoms for assessing bone mineral density and/or bone structure and/or one or more macro-anatomical and/or biomechanical parameters on any given image. Calibration references (also known as calibration phantoms) for use in imaging technologies have been described. See, e.g., U.S. Patent No. 5,493,601 and U.S. Patent No. 5,235,628. U.S. Patent No. 5,335,260 discloses a calibration phantom representative of human tissue containing variable concentrations of calcium that serves as reference for quantifying calcium, bone mass and bone mineral density in x-ray and CT imaging systems. However, currently available calibration phantoms are not always accurate. Because bone mineral density accounts for considerably less than 100% of fracture risk in osteoporosis (Ouyang et al. (1997) Calif Tissue Int, 60:139-147) some of the methods and devices described herein are designed to assess not only bone mineral density but also bone structure and, in addition, macro-anatomical and/or biomechanical parameters. By assessing two or more of these parameters, more accurate testing and screening can be provided for conditions such as osteoporosis.
[0058] Thus, in certain aspects, the current invention provides for methods and devices that allow accurate quantitative assessment of information contained in an x-ray such as density of an anatomic structure and/or morphology of an anatomic structure.
Any suitable calibration phantom can be used, for example, one that comprises aluminum or other radio-opaque materials. U.S. Patent No. 5,335,260 describes other calibration phantoms suitable for use in assessing bone mineral density in images. Examples of other suitable calibration reference materials can be fluid or fluid-like materials, for example, one or more chambers filled with varying concentrations of calcium chloride or the like.
[0059] Numerous calibration phantoms (or reference calibrations) can be used in the practice of the present invention. Typically, the system used to monitor bone mineral density and/or bone structure and/or one or more macro-anatomical and/or biomechanical parameters in a target organism comprises an image (e.g., a dental or hip radiograph), which provides information on the subject; an assembly including a calibration phantom, which acts as a reference for the data in the image; and at least one data processing system, which evaluates and processes the data from the image and/or from the calibration phantom assembly.
[0060] It will be readily apparent that a calibration phantom can contain a single, known density or structure reference. Furthermore, a gradient in density can be achieved by varying the thickness or the geometry of the calibration phantom along the path of the x-ray beam, for example, by using a V-shape of the calibration phantom of varying thickness (Fig. 4). The calibration phantom can also include angles. For example, the calibration phantom can be "T"-shaped or "L"-shaped thereby including one or more 90 degree angles. [0061] The calibration phantom can contain several different areas of different radio-opacity. For example, the calibration phantom can have a step-like design, whereby changes in local thickness of the wedge result in differences in radio-opacity. Stepwedges using material of varying thickness are frequently used in radiology for quality control testing of x-ray beam properties. By varying the thickness of the steps, the intensity and spectral content of the x-ray beam in the projection image can be varied. Stepwedges are commonly made of aluminum, copper and other convenient and homogeneous materials of known x-ray attenuation properties. Stepwedge-like phantoms can also contain calcium phosphate powder or calcium phosphate powder in molten paraffin. [0062] Alternatively, continuous wedges may be used or the calibration reference may be designed such that the change in radio-opacity is from periphery to center (for example in a round, ellipsoid, rectangular, triangular of other shaped structure). As noted above, the calibration reference can also be constructed as plurality of separate chambers, for example fluid filled chambers, each including a specific concentration of a reference fluid (e.g., calcium chloride). In addition to one or more fluids, a calibration phantom can also contain metal powder, e.g. aluminum or steel powder, embedded within it (for example, embedded in a plastic).
[0063] In certain embodiments, the calibration phantom is specifically designed to serve as a reference for bone structure (e.g., trabecular spacing, thickness and the like). For example, the calibration wedge can contain one or more geometric patterns with known dimensions, e.g. a grid whereby the spacing of a grid, thickness of individual grid elements, etc. are known. This known geometric pattern of radio-opaque elements in the calibration phantom can be used to improve the accuracy of measurements of trabecular bone structure in an x-ray. Such measurements of trabecular bone structure can include, but are not limited to, trabecular spacing, trabecular length and trabecular thickness. Such measurements of trabecular spacing, trabecular length and trabecular thickness can, for example, be performed in a dental or spine or hip x-ray. These calibration phantoms can be made up of a variety of materials include, plastics, metals and combinations thereof. Further, the reference components can be solid, powdered, fluid or combinations thereof. Thus, the calibration wedge can also be used to improve measurements of bone structure.
[0064] In certain embodiments, the calibration phantom is specifically designed to serve as a reference for macro-anatomical parameters (e.g., in the hip joint, thickness of the femoral shaft cortex, thickness of the femoral neck cortex, cortical width, hip axis length, CCD (caput-collum-diaphysis) angle, neck-shaft angle and width of the trochanteric region; and in the spine, thickness of the superior and inferior endplate, thickness of the anterior, lateral and posterior vertebral walls, diameter and height of the vertebral body, dimensions of the spinal canal and the posterior elements). For example, the calibration wedge can contain one or more geometric patterns with known dimensions, e.g. a grid whereby the spacing of a grid, thickness of individual grid elements, etc. are known. This known geometric pattern of radio-opaque elements in the calibration phantom can be used to improve the accuracy of measurements of macro-anatomical and/or biomechanical parameters in an x-ray, for example by aiding in the correction of image magnification. Such measurements of macro-anatomical parameters can, for example, be performed in a dental or spine or hip x-ray. These calibration phantoms can be made up of a variety of materials include, plastics, metals and combinations thereof. Further, the reference components can be solid, powdered, fluid or combinations thereof. Thus, the calibration wedge can also be used to improve measurements of bone structure.
[0065] Since the present invention contemplates analysis of dental x-ray images for information on bone structure, bone mineral density or both structure and density, it will be apparent that calibration phantoms will be selected based on whether structure, density or both are being measured. Thus, one or more calibration phantoms may be present.
[0066] Whatever the overall shape or composition of the calibration phantom, when present, the at least one marker be positioned at a known density and/or structure in the phantom. Furthermore, it is preferred that at least one geometric shape or pattern is included in the calibration phantom. Any shape can be used including, but not limited to, squares, circles, ovals, rectangles, stars, crescents, multiple-sided objects (e.g., octagons), V- or U-shaped, inverted V- or U-shaped, irregular shapes or the like, so long as their position is known to correlate with a particular density of the calibration phantom. In preferred embodiments, the calibration phantoms described herein are used in 2D planar x-ray imaging.
[0067] The calibration phantoms can be imaged before or after the x-ray image is taken. Alternatively, the calibration phantom can be imaged at the same time as the x- ray image. The calibration phantom can be physically connected to an x-ray film and/or film holder. Such physical connection can be achieved using any suitable mechanical or other attachment mechanism, including but not limited to adhesive, a chemical bond, use of screws or nails, welding, a Velcro™ strap or Velcro™ material and the like. Similarly, a calibration phantom can be physically connected to a detector system or a storage plate for digital x-ray imaging using one or more attachment mechanisms (e.g., a mechanical connection device, a Velcro™ strap or other Velcro™ material, a chemical bond, use of screws or nails, welding and an adhesive). The external standard and the film can be connected with use of a holding device, for example using press fit for both film and external standard.
[0068] Additionally, the calibration phantom assembly can be attached to an anatomical structure, for example one or more teeth, mucus membranes, the mandible and/or maxilla. For instance, the calibration phantom can be attached (e.g., via adhesive attachment means) to the epithelium or mucous membrane inside overlying the mandible or the maxilla. Alternatively, the calibration phantom can be placed on or adjacent to a tooth, for example, a V- or U-shaped (in the case of the maxilla) or an inverted V- or U-shaped (in the case of the mandible) calibration phantom can be used. The opening of the V or U will be in contact with the free edge of at least one tooth or possibly several teeth (Fig. 4). [0069] In preferred embodiments, when an x-ray of an anatomic structure or a non-living object is acquired a calibration phantom is included in the field of view. Any suitable calibration phantom can be used, for example, one that comprises aluminum or other radio-opaque materials. U.S. Patent No. 5,335,260 describes other calibration phantoms suitable for use in assessing bone mineral density in images. Examples of other suitable calibration reference materials can be fluid or fluid-like materials, for example, one or more chambers filled with varying concentrations of calcium chloride or the like. In a preferred embodiment, the material of the phantom is stainless steel (e.g., AISI grade 316 comprising carbon (0.08%); manganese (2%); silicon (1%); phosphorus (0.045%); sulphur (0.03%); nickel (10-14%); chromium (16-18%); molybdenum (2-3%); plus iron to make up 100%). The relative percentages of the components may be with respect to weight or volume.
[0070] It will be apparent that calibration phantoms suitable for attachment to an anatomical structure can have different shapes depending on the shape of the anatomical structure (e.g., tooth or teeth) on which or adjacent to which it will be placed including, but not limited to, U-shaped, V-shaped, curved, flat or combinations thereof. For example, U-shaped (or inverted U-shaped) calibration phantoms can be positioned on top of molars while V-shaped (or inverted V-shaped) calibration phantoms can be positioned on top of incisors. Further, it will be apparent that in certain instances (e.g., teeth on the mandible), the calibration phantom can rest on top of the tooth just based on its gravity or it can be attached to the tooth (e.g., using adhesive). In the case of the teeth on the maxilla, the calibration phantom will typically be attached to the tooth, for example with use of an adhesive. [0071] Any of these attachments may be permanent or temporary and the calibration phantom can be integral (e.g., built-in) to the film, film holder and/or detector system or can be attached or positioned permanently or temporarily appropriately after the film and/or film holder is produced. Thus, the calibration phantom can be designed for single-use (e.g., disposable) or for multiple uses with different x-ray images. Thus, in certain embodiments, the calibration phantom is reusable and, additionally, can be sterilized between uses. Integration of a calibration phantom can be achieved by including a material of known x-ray density between two of the physical layers of the x- ray film. Integration can also be achieved by including a material of known x-ray density within one of the physical layers of the x-ray film. Additionally, the calibration phantom can be integrated into the film cover. A calibration phantom or an external standard can also be integrated into a detector system or a storage plate for digital x-ray imaging. For example, integration can be achieved by including a material of known x-ray density between two of the physical layers of the detector system or the storage plate. Integration can also be achieved by including a material of know x-ray density within one of the physical layers of the detector system or the storage plate.
[0072] In certain embodiments, for example those embodiments in which the calibration phantom is temporarily attached to a component of the x-ray assembly system (e.g., x-ray film holder, x-ray film, detector system or the like), cross-hairs, lines or other markers may be placed on the apparatus as indicators for positioning of the calibration phantom. These indicators can help to ensure that the calibration phantom is positioned such that it doesn't project on materials that will alter the apparent density in the resulting image.
[0073] Any of the calibration phantom-containing assemblies described herein can be used in methods of analyzing and/or quantifying bone structure and/or one or more macro-anatomical and/or biomechanical parameters (or bone mineral density) in an x-ray image. The methods generally involve simultaneously imaging or scanning the calibration phantom and another material (e.g., bone tissue from a subject) for the purpose of quantifying the density of the imaged material (e.g., bone mass). In the case of dental radiographs, the calibration phantom, the x-ray tube or dental x-ray film is typically positioned in a manner to ensure inclusion of the calibration phantom and a portion of the mandible and/or maxilla on the dental x-ray image. Preferably, the calibration phantom, the x-ray tube and the dental x-ray film are positioned so that at least a portion of the section of the mandible or maxilla included on the image will contain predominantly trabecular bone rather than cortical bone.
[0074] Thus, under the method of the present invention, the calibration phantom is preferably imaged or scanned simultaneously with the individual subject, although the invention allows for non-simultaneous scanning of the phantom and the subject. Methods of scanning and imaging structures by x-ray imaging technique are well known. By placing the calibration phantom in the x-ray beam with the subject, reference calibration samples allow corrections and calibration of the absorption properties of bone. When the phantom is imaged or scanned simultaneously with each subject, the variation in x-ray beam energy and beam hardening are corrected since the phantom and the subject both see the same x-ray beam spectrum. Each subject, having a different size, thickness, muscle-to-fat ratio, and bone content, attenuate the beam differently and thus change the effective x-ray beam spectrum. It is necessary that the bone-equivalent calibration phantom be present in the same beam spectrum as the subject's bone to allow accurate calibration. [0075] X-ray imaging assemblies that are currently in use do not take into account the position of the calibration phantom in relation to the structures being imaged. Thus, when included in known assemblies, calibration phantom(s) are often positioned such that they project on materials or structures (e.g., bone) that alter apparent density of the calibration phantom in the resulting x-ray image. Clearly, this alteration in apparent density will affect the accuracy of the calibration phantom as a reference for determining bone mineral density, structure or macro-anatomical parameters. Therefore, it is an object of the invention to provide methods in which the calibration phantom projects free of materials or structures that will alter the apparent density of the reference. In the context of dental x-rays, for instance, the methods described herein ensure that the calibration phantom projects free of bone (e.g., teeth, jaw) tissue. This can be accomplished in a variety of ways, for example, positioning the calibration phantom in the x-ray film or in the x-ray film holder such that it will appear between the teeth in the dental x-ray. [0076] The calibration phantom materials and methods of the present invention are preferably configured to be small enough and thin enough to be placed inside the mouth, and the method of the present invention can be used to quantify bone mass using standard dental x-ray systems, for example by including temporary or permanent calibration phantoms in dental x-ray film holders. Further, it is highly desirable that the calibration phantom be positioned so that at least a portion doesn't project on structures or materials that will alter the apparent density or structural characteristics of the calibration phantoms. It is also preferable to position calibration phantom at a defined distance relative to at least one tooth or the mandible or the maxilla whereby a substantial portion of the calibration phantom projects free of said tooth, said mandible or said maxilla on the x-ray image. Any suitable distance can be used, for example between about 1 mm and 5 cm or any value therebetween.
[0077] A cross-calibration phantom can be used to optimize system performance, e.g. x-ray tube settings or film processor settings, or to improve the comparability of different machines or systems, typically located at different sites. For this purpose, a separate image may be obtained which does not include a patient or a body part. The image includes the primary calibration phantom used in patients, e.g. a step-wedge of known density, and the cross-calibration phantom. The apparent density of the primary calibration phantom is then calibrated against the density of the cross-calibration phantom. The resultant cross-calibration of the primary phantom can help to improve the accuracy of measurements of bone density, bone structure and macro-anatomical and/or biomechanical parameters. It can also help improve the overall reproducibility of the measurements. In one embodiment of the invention, an x-ray technologist or a dental hygienist will perform a cross-calibration test once a day, typically early in the morning, prior to the first patient scans. The results of the cross-calibration or the entire cross-calibration study can be transmitted via a network to a central computer. The central computer can then perform adjustments designed to maintain a high level of comparability between different systems. 1.2. Inherent Reference Markers [0078] In certain embodiments of the invention, information inherent in the anatomic structure or the non-living object can be used to estimate the density and/or structure and/or macro-anatomy of selected bone regions of interest within the anatomic structure or the non-living object. For example, since the density of muscle, fat, water (e.g., soft tissue), metal (e.g., dental fillings) and air are typically known, the density of air surrounding an anatomic structure or non-living object, the density of subcutaneous fat, and the density of muscle tissue can be used to estimate the density of a selected region of bone, for example within the distal radius. For instance, a weighted mean can be determined between one or more of the internal standards (e.g., air, water, metal, and/or fat) and used as internal standards to determine bone density in the same x-ray image. Similarly, the density of a tooth or a portion of a tooth can be used to estimate the density of a selected region of bone, e.g. an area in the mandible.
[0079] The information inherent in said anatomic structure can also be combined with information provided by the calibration phantom and the combination can result in an improved accuracy of the calibration phantom. 1.3. Holders and Hygienic Covers
[0080] As noted above, in certain embodiments, a holder can be used to position the calibration phantom. The holder can be U-shaped or V-shaped (Fig. 5) for ease in attachment to a tooth. The attachment can be, for example, with an adhesive. The calibration phantom, in turn, can be attached to the holder. Similarly, the calibration phantom can be attached to holders comprising one or more molds of at least one or more teeth. Additionally, the holder can be used to position both the film and the calibration phantom relative to the osseous structure that will be included in the x-ray image. In another embodiment, a holding device that can hold the x-ray film is integrated in the calibration phantom. This holding device can hold the film in place prior to taking the x-ray. The holding device can be spring-loaded or use other means such as mechanical means of holding and stabilizing the x-ray film.
[0081] In certain embodiments, the holder may comprise a disposable or sterilizeable hygienic cover. See, e.g., WO 99/08598, the disclosure of which is incorporated by reference herein in its entirety. Furthermore, the holder may comprise multiple components, for example, the calibration phantom and a integrated or insertable bolus back that can serve to enhance the accuracy of the calibration phantom by accounting for the effect of soft tissue that may project with the calibration phantom and/or with the bone.
[0082] In certain embodiments, the calibration phantom can be configured so that it stabilizes against the surrounding tissues on its own without the use of an additional holder. The calibration phantom can be protected with a hygienic cover.
[0083] The holder (e.g., hygienic cover) may be comprised of a rigid material, a flexible material or combinations thereof. Furthermore, the holder may include one or more pockets/compartments adapted to receive additional components such as the calibration phantom, a bolus back or the like. Additionally, one or more portions of the holder may be radiolucent. 2.0. Analysis md Manipulation ©f Data [0084] The data obtained from images taken as described above is then preferably analyzed and manipulated. Thus, the systems and assemblies described herein can also include one or more computational units designed, for example, to analyze bone density or bone structure or macro-anatomical and/or biomechanical data in the image; to identify an anatomical landmark in an anatomical region; to correct for soft tissue measurements; and/or to evaluate bone density and structure and macro- anatomy of the image. As will be appreciated by those of skill in the art, the computational unit can include any software, chip or other device used for calculations. Additionally, the computational unit may be designed to control the imaging assembly or detector (as well as other parameters related to the detector(s)). Other applications of the computational unit to the methods and devices described herein will be recognized by those skilled in the art. The computational unit may be used for any other application related to this technology that may be facilitated with use of computer software or hardware. The computational unit can also further comprise a database comprising, for example, reference anatomical maps and the computational unit is further designed to compare the anatomical map with the reference anatomical map. The reference anatomical map may be historic (from the same or another patient, generated as part of an interrogation protocol), or theoretical or any other type of desired reference map. [0085] Any image can be analyzed in order to obtain and manipulate data. Thus, data points, derived data, and data attributes database according to the present invention may comprise the following: (1) the collection of data points, said data points comprising information obtained from an image, for example, bone mineral density information or information on bone structure (architecture); and (2) the association of those data points with relevant data point attributes. The method may further comprise (3) determining derived data points from one or more direct data points and (4) associating those data points with relevant data point attributes. The method may also comprise (5) collection of data points using a remote computer whereby said remote computer operates in a network environment. [0086] In certain preferred embodiments, the information is obtained from a dental x-ray image. As described herein, dental x-ray images can be acquired at a local site using known techniques. If the x-ray image was captured using conventional x-ray film, the data points (information) of the x-ray image can be digitized using a scanning device. The digitized x-ray image information can then be transmitted over the network, e.g. the Internet, into a remote computer or server. If the x-ray image was acquired using digital acquisition techniques, e.g. using phosphorus plate systems or selenium or silicon detector systems, the x-ray image information is already available in digital format. In this case the image can be transmitted directly over the network, e.g. the Internet. The information can also be compressed and/or encrypted prior to transmission. Transmission can also be by other methods such as fax, mail or the like. 2.1. Data Points
[0087] Thus, the methods of and compositions described herein make use of collections of data sets of measurement values, for example measurements of bone structure and/or bone mineral density from x-ray images. Records may be formulated in spreadsheet-like format, for example including data attributes such as date of x-ray, patient age, sex, weight, current medications, geographic location, etc. The database formulations may further comprise the calculation of derived or calculated data points from one or more acquired data points. A variety of derived data points may be useful in providing information about individuals or groups during subsequent database manipulation, and are therefore typically included during database formulation. Derived data points include, but are not limited to the following: (1) maximum bone mineral density, determined for a selected region of bone or in multiple samples from the same or different subjects; (2) minimum bone mineral density, determined for a selected region of bone or in multiple samples from the same or different subjects; (3) mean bone mineral density, determined for a selected region of bone or in multiple samples from the same or different subjects; (4) the number of measurements that are abnormally high or low, determined by comparing a given measurement data point with a selected value; and the like. Other derived data points include, but are not limited to the following: (1) maximum value of a selected bone structure parameter, determined for a selected region of bone or in multiple samples from the same or different subjects; (2) minimum value of a selected bone structure parameter, determined for a selected region of bone or in multiple samples from the same or different subjects; (3) mean value of a selected bone structure parameter, determined for a selected region of bone or in multiple samples from the same or different subjects; (4) the number of bone structure measurements that are abnormally high or low, determined by comparing a given measurement data point with a selected value; and the like. Other derived data points include, but are not limited to the following: (1) maximum value of a selected macro-anatomical and/or biomechanical parameter, determined for a selected region of bone or in multiple samples from the same or different subjects; (2) minimum value of a selected macro-anatomical and/or biomechanical parameter, determined for a selected region of bone or in multiple samples from the same or different subjects; (3) mean value of a selected macro-anatomical and/or biomechanical parameter, determined for a selected region of bone or in multiple samples from the same or different subjects; (4) the number of macro-anatomical and/or biomechanical measurements that are abnormally high or low, determined by comparing a given measurement data point with a selected value; and the like. Other derived data points will be apparent to persons of ordinary skill in the art in light of the teachings of the present specification. The amount of available data and data derived from (or arrived at through analysis of) the original data provide provides an unprecedented amount of information that is very relevant to management of bone related diseases such as osteoporosis. For example, by examining subjects over time, the efficacy of medications can be assessed. [0088] Measurements and derived data points are collected and calculated, respectively, and may be associated with one or more data attributes to form a database. The amount of available data and data derived from (or arrived at through analysis of) the original data provide provides an unprecedented amount of information that is very relevant to management of bone related diseases such as osteoporosis. For example, by examining subjects over time, the efficacy of medications can be assessed.
[0089] Data attributes can be automatically input with the x-ray image and can include, for example, chronological information (e.g., DATE and TIME). Other such attributes may include, but are not limited to, the type of x-ray imager used, scanning information, digitizing information and the like. Alternatively, data attributes can be input by the subject and/or operator, for example subject identifiers, i.e. characteristics associated with a particular subject. These identifiers include but are not limited to the following: (1) a subject code (e.g., a numeric or alpha-numeric sequence); (2) demographic information such as race, gender and age; (3) physical characteristics such as weight, height and body mass index (BMI); (4) selected aspects of the subject's medical history (e.g., disease states or conditions, etc.); and (5) disease-associated characteristics such as the type of bone disorder, if any; the type of medication used by the subject. In the practice of the present invention, each data point would typically be identified with the particular subject, as well as the demographic, etc. characteristic of that subject.
[0090] Other data attributes will be apparent to persons of ordinary skill in the art in light of the teachings of the present specification. 2.2. Storage of Data Sets and Association of Data Points with Relevant Data Attributes
[0091] A number of formats exist for storing data sets and simultaneously associating related attributes, including but not limited to (1) tabular, (2) relational, and (3) dimensional. In general the databases comprise data points, a numeric value which correspond to physical measurement (an "acquired" datum or data point) or to a single numeric result calculated or derived from one or more acquired data points that are obtained using the various methods disclosed herein. The databases can include raw data or can also include additional related information, for example data tags also referred to as "attributes" of a data point. The databases can take a number of different forms or be structured in a variety of ways. [0092] The most familiar format is tabular, commonly referred to as a spreadsheet. A variety of spreadsheet programs are currently in existence, and are typically employed in the practice of the present invention, including but not limited to Microsoft Excel spreadsheet software and Corel Quattro spreadsheet software. In this format, association of data points with related attributes occurs by entering a data point and attributes related to that data point in a unique row at the time the measurement occurs.
[0093] Further, rational, relational (Database Design for Mere Mortals, by Michael
J. Hernandez, 1997, Addison-Wesley Pub. Co., publisher; Database Design for Smarties, by Robert J. Muller, 1999, Morgan Kaufmann Publishers, publisher; Relational Database Design Clearly Explained, by Jan L. Harrington, 1998, Morgan Kaufmann Publishers, publisher) and dimensional (Data-Parallel Computing, by V.B. Muchnick, et al., 1996, International Thomson Publishing, publisher; Understanding Fourth Dimensions, by David Graves, 1993, Computerized Pricing Systems, publisher) database systems and management may be employed as well.
[0094] Relational databases typically support a set of operations defined by relational algebra. Such databases typically include tables composed of columns and rows for the data included in the database. Each table of the database has a primary key, which can be any column or set of columns, the values for which uniquely identify the rows in a table. The tables in the database can also include a foreign key that is a column or set of columns, the values of which match the primary key values of another table. Typically, relational databases also support a set of operations (e.g., select, join and combine) that form the basis of the relational algebra governing relations within the database.
[0095] Such relational databases can be implemented in various ways. For instance, in Sybase® (Sybase Systems, Emeryville, CA) databases, the tables can be physically segregated into different databases. With Oracle® (Oracle Inc., Redwood Shores, CA) databases, in contrast, the various tables are not physically separated, because there is one instance of work space with different ownership specified for different tables. In some configurations, databases are all located in a single database (e.g., a data warehouse) on a single computer. In other instances, various databases are split between different computers.
[0096] It should be understood, of course, that the databases are not limited to the foregoing arrangements or structures. A variety of other arrangements will be apparent to those of skill in the art. 2.3. Data Manipulation [0097] Data obtained from x-ray images as described herein can be manipulated, for example, using a variety of statistical analyses, to produce useful information. The databases of the present invention may be generated, for example, from data collected for an individual or from a selected group of individuals over a defined period of time (e.g., days, months or years), from derived data, and from data attributes.
[0098] For example, data may be aggregated, sorted, selected, sifted, clustered and segregated by means of the attributes associated with the data points. A number of data mining software programs exist which may be used to perform the desired manipulations. [0099] Relationships in various data can be directly queried and/or the data analyzed by statistical methods to evaluate the information obtained from manipulating the database.
[0100] For example, a distribution curve can be established for a selected data set, and the mean, median and mode calculated therefor. Further, data spread characteristics, e.g. variability, quartiles and standard deviations can be calculated.
[0101] The nature of the relationship between any variables of interest can be examined by calculating correlation coefficients. Useful methods for doing so include but are not limited to the following: Pearson Product Moment Correlation and Spearman Rank Order Correlation. [0102] Analysis of variance permits testing of differences among sample groups to determine whether a selected variable has a discernible effect on the parameter being measured.
[0103] Non-parametric tests may be used as a means of testing whether variations between empirical data and experimental expectancies are attributable merely to chance or to the variable or variables being examined. These include the Chi Square test, the Chi Square Goodness of Fit, the 2 x 2 Contingency Table, the Sign Test, and the Phi Correlation Coefficient. [0104] There are numerous tools and analyses available in standard data mining software that can be applied to the analysis of the databases of the present invention. Such tools and analyses include, but are not limited to, cluster analysis, factor analysis, decision trees, neural networks, rule induction, data driven modeling, and data visualization. Some of the more complex methods of data mining techniques are used to discover relationships that are more empirical and data-driven, as opposed to theory- driven, relationships.
[0105] Exemplary data mining software that can be used in analysis and/or generation of the databases of the present invention includes, but is not limited to: Link Analysis (e.g., Associations analysis, Sequential Patterns, Sequential time patterns and Bayes Networks); Classification (e.g., Neural Networks Classification, Bayesian Classification, k-nearest neighbors classification, linear discriminant analysis, Memory based Reasoning, and Classification by Associations); Clustering (e.g., k-Means Clustering, demographic clustering, relational analysis, and Neural Networks Clustering); Statistical methods (e.g., Means, Std dev, Frequencies, Linear Regression, non-linear regression, t-tests, F-test, Chi2 tests, Principal Component Analysis, and Factor Analysis); Prediction (e.g., Neural Networks Prediction Models, Radial Based Functions predictions, Fuzzy logic predictions, Times Series Analysis, and Memory based Reasoning); Operating Systems; and Others (e.g., Parallel Scalability, Simple Query Language functions, and C++ objects generated for applications). Companies that provide such software include, for example, the following: Adaptative Methods Group at UTS (UTS City Campus, Sydney, NSW 2000), CSI®, Inc., (Computer Science Innovations, Inc. Melbourne, Florida), IBM® (International Business Machines Corporation, Armonk, NY), Oracle® (Oracle Inc., Redwood Shores, CA) and SAS® (SAS Institute Inc., Gary, NC).
[0106] These methods and processes may be applied to the data obtained using the methods described herein, for example, databases comprising, x-ray image data sets, derived data, and data attributes. [0107] In certain embodiments, data (e.g., bone structural information or macro- anatomical and/or biomechanical information or bone mineral density information) is obtained from normal control subjects using the methods described herein. These databases are typically referred to as "reference databases" and can be used to aid analysis of any given subject's x-ray image, for example, by comparing the information obtained from the subject to the reference database. Generally, the information obtained from the normal control subjects will be averaged or otherwise statistically manipulated to provide a range of "normal" (reference) measurements. Suitable statistical manipulations and/or evaluations will be apparent to those of skill in the art in view of the teachings herein.
[0108] ADD z-score, T-scores here; other statistical measurements that you deem important. The comparison of the subject's x-ray information to the reference database can be used to determine if the subject's bone information falls outside the normal range found in the reference database or is statistically significantly different from a normal control. Data comparison and statistical significance can be readily determined by those of skill in the art using for example the z-test or t-test statistics for continuous variables, the chi-square test or Fisher's exact test for categorical data and the rank-sum test or Kruskal-Wallis test for ranked data. The use of reference databases in the analysis of x-ray images facilitates that diagnosis, treatment and monitoring of bone conditions such as osteoporosis.
[0109] For a general discussion of statistical methods applied to data analysis, see Applied Statistics for Science and Industry, by A. Romano, 1977, Allyn and Bacon, publisher.
[0110] The data is preferably stored and manipulated using one or more computer programs or computer systems. These systems will typically have data storage capability (e.g., disk drives, tape storage, CD-ROMs, etc.). Further, the computer systems may be networked or may be stand-alone systems. If networked, the computer system would be able to transfer data to any device connected to the networked computer system for example a medical doctor or medical care facility using standard e-mail software, a central database using database query and update software (e.g., a data warehouse of data points, derived data, and data attributes obtained from a large number of subjects). Alternatively, a user could access from a doctor's office or medical facility, using any computer system with Internet access, to review historical data that may be useful for determining treatment.
[0111] If the networked computer system includes a World Wide Web application, the application includes the executable code required to generate database language statements, for example, SQL statements. Such executables typically include embedded SQL statements. The application further includes a configuration file that contains pointers and addresses to the various software entities that are located on the database server in addition to the different external and internal databases that are accessed in response to a user request. The configuration file also directs requests for database server resources to the appropriate hardware, as may be necessary if the database server is distributed over two or more different computers.
[0112] Usually each networked computer system includes a World Wide Web browser that provides a user interface to the networked database server. The networked computer system is able to construct search requests for retrieving information from a database via a Web browser. With access to a Web browser users can typically point and click to user interface elements such as buttons, pull down menus, and other graphical user interface elements to prepare and submit a query that extracts the relevant information from the database. Requests formulated in this manner are subsequently transmitted to the Web application that formats the requests to produce a query that can be used to extract the relevant information from the database.
[0113] When Web-based applications are utilized, the Web application accesses data from a database by constructing a query in a database language such as Sybase or Oracle SQL which is then transferred to a relational database management system that in turn processes the query to obtain the pertinent information from the database.
[0114] Accordingly, in one aspect the present invention describes a method of providing data obtained from x-ray images on a network, for example the Internet, and methods of using this connection to provide real-time and delayed data analysis. The central network can also allow access by the physician to a subject's data. Similarly, an alert could be sent to the physician if a subject's readings are out of a predetermined range, etc. The physician can then send advice back to the patient via e-mail or a message on a web page interface. Further, access to the entire database of data from all subjects may be useful for statistical or research purposes. Appropriate network security features (e.g., for data transfer, inquiries, device updates, etc.) are of course employed.
[0115] Further, a remote computer can be used to analyze the x-ray that has been transmitted over the network automatically. For example, x-ray density information or structural information about an object can be generated in this fashion. X-ray density information can, for example, be bone mineral density. If used in this fashion, the test can be used to diagnose bone-related conditions such as osteoporosis.
2.4. Graphical User Interface
[0116] In certain of the computer systems, an interface such as an interface screen that includes a suite of functions is included to enable users to easily access the information they seek from the methods and databases of the invention. Such interfaces usually include a main menu page from which a user can initiate a variety of different types of analyses. For example, the main menu page for the databases generally include buttons for accessing certain types of information, including, but not limited to, project information, inter-project comparisons, times of day, events, dates, times, ranges of values, etc.
2.5. Computer Program Products [0117] A variety of computer program products can be utilized for conducting the various methods and analyses disclosed herein. In general, the computer program products comprise a computer-readable medium and the code necessary to perform the methods set forth supra. The computer-readable medium on which the program instructions are encoded can be any of a variety of known medium types, including, but not limited to, microprocessors, floppy disks, hard drives, ZIP drives, WORM drives, magnetic tape and optical medium such as CD-ROMs.
[0118] For example, once an image or data from that image is transmitted via a local or long-distance computer network and the data received by a remote computer or a computer connected to the remote network computer, an analysis of the morphology and density of the bone can be performed, for example using suitable computer programs. This analysis of the object's morphology can occur in two-dimensions or three-dimensions. For example, in imaging osseous structures, such analysis of the transmitted x-ray image can be used to measure parameters that are indicative or suggestive of bone loss or metabolic bone disease. Such parameters include all current and future parameters that can be used to evaluate osseous structures. For example, such parameters include, but are not limited to, trabecular spacing, trabecular thickness, trabecular connectivity and intertrabecular space.
[0119] Information on the morphology or 2D or 3D structure of an anatomic object can be derived more accurately, when image acquisition parameters such as spatial resolution are known. Other parameters such as the degree of cone beam distortion can also be helpful in this setting.
[0120] As noted above, an image can be transmitted from a local site into a remote server and the remote server can perform an automated analysis of the image. Further, the remote server or a computer connected to the remote server can then generate a diagnostic report. Thus, in certain embodiments, a computer program (e.g., on the remote server or on a computer connected to the remote server) can generate charges for the diagnostic report. The remote server can then transmit the diagnostic report to a physician, typically the physician who ordered the test or who manages the patient. The diagnostic report can also be transmitted to third parties, e.g. health insurance companies. Such transmission of the diagnostic report can occur electronically (e.g. via e-mail), via mail, fax or other means of communication. All or some of the transmitted information (e.g., patient identifying information) can be encrypted to preserve confidentiality of medical records.
[0121] Thus, one exemplary system is described herein for analyzing bone morphology or structure in a subject system via a dental x-ray that includes at least a portion of the mandible and/or maxilla of a subject, followed by evaluation or the x-ray image. Dental x-rays are obtained in any conventional method. The x-ray produces an image that can be interpreted (for example, employing a selected algorithm and/or computer program) by an associated system controller to provide a bone mineral density or bone structure evaluation for display.
[0122] In a further aspect of the present invention, the monitoring system can comprise two or more components, in which a first component comprises an x-ray image and calibration phantom that are used to extract and detect bone-related data on the subject, and a second component that receives the data from the first component, conducts data processing on the data and then displays the processed data. Microprocessor functions can be found in one or both components. The second component of the monitoring system can assume many forms
3.0.0.0 Correction Factors
[0123] Although the presence of calibration phantoms greatly aids in increasing the accuracy of data obtained from images such as dental, hip or spine x-rays, the present inventors also recognize that, in certain instances, there may be a need to apply one or more correction factors to further enhance accuracy of the data obtained from any given x-ray image. Such correction factors will take into account one or more of a wide variety of influences (e.g.,. soft tissue thickness, region from which the data is extracted and the like) that can alter apparent density or structure information on the image.
[0124] In this regard, one or more reference databases can be used for calibration and normalization purposes. For example, image normalization or correction of soft-tissue attenuation can be performed using patient characteristic data such as patient weight, height and body mass index. In one example, a higher soft-tissue attenuation can be assumed in high weight and low height subjects; a lower soft-tissue attenuation will be assumed in low weight and high height subjects.
[0125] In another embodiment, a standard calibration curve is applied to x-ray images, whereby said calibration curve can be derived from reference x-rays obtained with use of calibration phantoms. For example, 100 patients can undergo dental x-rays with a calibration phantom and a standard calibration curve can be derived from these images. Similarly, 100 patients can undergo hip x-rays with a calibration phantom and a standard calibration curve can be derived from these images. Different calibration curves can be generated for different populations, for example, by generating different calibration curves for different ranges in body mass index, body height, sex, race etc.
3.1.0.0. Anatomical Landmarks
[0126] In one embodiment, identification of anatomic landmarks of the structure to be analyzed or identification of anatomical landmarks adjacent to the structure to be analyzed with subsequent positioning and computer analysis of the x-ray image relative to these anatomic landmarks or with subsequent positioning and computer analysis of anatomical region of interest (ROI) relative to these anatomic landmarks is performed. The present invention includes also positioning dental or other x-ray detectors, positioning the dental or other x-ray tube, and analyzing the resulting images using landmarks based on either 1) textural information, 2.) structural information, 3.) density information (e.g. density), or 4) 2 or 3 dimensional contour information 5) a combinations thereof of the tissue or structure to be measured and of tissues or structures adjacent to the measurement site. The invention also includes methods and devices that are not necessarily based solely on anatomical landmarks, but in some applications can be combined with anatomical landmark embodiments. Preferably, many of the embodiments described herein are designed for automated use with a minimum of operator intervene and preferably remote or computer control of such devices.
[0127] In one embodiment, an alignment device may be used to ensure perpendicular or near perpendicular alignment of the dental or other x-ray tube relative to the x-ray film, thereby decreasing geometric distortion resulting from tube angulation. For example, an x-ray film holder is positioned relative to an anatomical landmark, e.g. the posterior wall of the mandible in the incisor region. A side-view of an exemplary alignment system using a dental x-ray film holder is shown in Figure 15. The system includes bite block (100), stainless steel rod (101), film (103), optional calibration phantom (104), Rinn holder (102) typically having a ring or donut shape, and extension tubing (200). The extension tubing is designed to fit within the Rinn holder and may be temporarily or permanently attached. The system can achieve high reproducibility of the film position relative to an anatomical landmark such as the alveolar ridge or the posterior wall of the mandible. The extension tubing allows for alignment of the x-ray tube so that it is near perpendicular to the Rinn instrument and, ultimately, the dental film. [0128] Since manual alignment of the dental x-ray tube, namely the tube (e.g., metal) located in front of the dental x-ray tube for pointing and alignment purposes, is often not very accurate with alignment errors of 3, 5 or even more degrees, a mechanical or electromagnetic device is preferably used in order to achieve perpendicular or near perpendicular alignment between the metal tube anterior to the x- ray tube and the Rinn holder. For example, the metal tube can be physically attached to the Rinn holder with use of one or more Velcro™ straps or it can be aligned using optical aids such as levels, cross-hairs, light sources (points or areas), etc. Alternatively, such physical attachment can be achieved with use of one or more magnets rigidly attached to the dental x-ray system metal tube and the Rinn holder. In this embodiment, the magnets on the Rinn holder and the dental x-ray system metal tube will be aligned and brought into physical contact. In another embodiment, an extension tube is attached, for example with an adhesive, to the Rinn holder. The extension tubing can also be an integral part of the Rinn holder. The extension tubing can be designed so that its inner diameter is slightly greater than the outer diameter of the dental x-ray system metal tube. The dental x-ray system metal tube is then inserted into the extension tubing attached to the Rinn holder thereby greatly reducing alignment error of the x-ray tube relative to the x-ray film. Alternatively, the extension tubing can be designed so that its outer diameter is slightly smaller than the inner diameter of the dental x-ray system metal tube. The dental x-ray system metal tube is then advanced over the extension tubing attached to the Rinn holder thereby greatly reducing alignment error of the x-ray tube relative to the x-ray film. One of skill in the art will easily recognize in view of the teachings herein that many other attachment means can be used for properly aligning the dental x-ray tube with the dental x-ray film. Combinations of attachment mechanisms are also possible.
[0129] The anatomical landmark that is selected is part of an anatomical region.
An anatomical region refers to a site on bone, tooth or other definable biomass that can be identified by an anatomical feature(s) or location. An anatomical region can include the biomass underlying the surface. Usually, such a region will be definable according to standard medical reference methodology, such as that found in Williams et al., Gray's Anatomy, 1980. The anatomical region can be selected from the group consisting of an edge of the mandible, an edge of the maxilla, an edge of a tooth, valleys or grooves in any of these structures or combinations thereof. The dental x-ray image can be readily taken so as to include the anatomical site. Other anatomical regions include but are not limited to the hip, the spine, the forearm, the foot, and the knee.
[0130] For example, the region of interest is placed between the dental apices and the inferior mandibular cortex. The apices can be found automatically in the following way: for each row of pixels, the gray value profile is examined. While a profile that intersects bone and dental roots in an alternating fashion has several distinct peaks and valleys, a profile that only covers trabecular bone shows irregular changes in the gray values (Fig. 6). The dental apices are located in the transitional region between these two patterns.
[0131] The measurement techniques to assess trabecular bone structure or macro-anatomical and/or biomechanical parameters are preferably designed to work without user intervention. In order to fully automate the process of analyzing dental x- rays, it is necessary to develop a technique to locate the regions of interest (ROIs) that are used for the calculation of the structural parameters of the trabecular bone. If the profile for a particular row of pixels contains distinct peaks, their number, width and height can be determined. Next, the rows below these lines can be evaluated until the peaks have disappeared. This line determines the boundary, 5 mm below which the ROI can be placed in the center between the longitudinal axes of the roots, which can also be determined from the row profiles (Fig. 6). At a pixel size of 0.042mm x 0.042mm, which corresponds to a resolution of 600dpi, the ROI has a size of 5.4mm x 5.4mm (128x128 pixels). For other scanning resolutions, the pixel resolution of the ROI can be adjusted accordingly.
[0132] In the case of an edentulous patient, bone mineral density can be measured in all ROIs that are located on a line that is, for example, 8 mm inferior and parallel to the alveolar ridge. The ROIs can be moved from left to right on a pixel-by- pixel basis. Eventually, the ROI with the lowest BMD can be chosen for further evaluation of the structural bone parameters. This helps to avoid inclusion of regions on the x-ray where bone mineral density may be overestimated due to projection of the curved parts of the mandible near the canine teeth. Alternatively, the ROI with the median BMD can be used. Other statistical parameters can be employed for this purpose.
[0133] Thus, software or other computational unit can identify the selected anatomic landmark in an interrogated x-ray image and direct analysis of the image using various parameters and analytic functions. Further, such software or other computational analytical unit can be used to identify areas of particular density at a certain distance from the selected landmark. Similarly, manual or computer analysis can be used to identify areas of lowest, highest, median or average density (or structural characteristics) in relation to the selected landmark. [0134] Further, the same landmark may be compared at different times (intra- landmark comparison) or one or more landmarks may be compared (inter-landmark comparison). For instance, an intra-landmark comparison can be used during a single interrogation protocol that entails multiple interrogations of the same region with reference to a particular anatomical landmark. Statistical analysis as described herein and known in the art can be performed.
[0135] Thus, the invention provides for means of assessing bone structure, i.e. the two-dimensional or three-dimensional architectural organization of the trabecular bone including, but not limited to, measurement of trabecular spacing, trabecular thickness, trabecular length and trabecular connectivity. Other examples of measurements of bone structure are provided in TABLE 1. These measurements can be used alone or enhanced with use of calibration phantoms or external standards that can allow a correction or normalization of image intensity and that can in certain embodiments also allow a correction of geometric distortions for example resulting from cone beam geometry of an x-ray beam. [0136] The invention provides for means of assessing macro-anatomical and/or biomechanical parameters. These measurements can be used alone or enhanced with use of calibration phantoms or external standards that can allow a correction or normalization of image intensity and that can in certain embodiments also allow a correction of geometric distortions including magnification, for example resulting from cone beam geometry of an x-ray beam.
[0137] As described herein, one or more measurements of bone structure or macro-anatomical and/or biomechanical parameters can be used to select a therapy, for example the use of anabolic or antiresorptive agent in the case of bone loss or deterioration. In certain embodiments, measurements of bone structure and/or one or more macro-anatomical and/or biomechanical parameters are conducted over time to longitudinally monitor a subject's bone health longitudinally over time. Measurements can be performed at different time points T1 , T2, ..., Tn and changes in said bone structure and/or macro-anatomical and/or biomechanical parameters can be registered and used to track a patient's bone health. In either single or longitudinally measurements, a physician can be apprised of the measurements and can include a pre-determined cut-off value (e.g., when a bone structure or macro-anatomical and/or biomechanical parameter measured in a patient is more than one or two standard deviations different from a normal, healthy reference population) and use this information to select a therapy.
[0138] The data obtained and analyzed as described herein can be used to monitor a patient's response to therapy. For example, information regarding bone structural and/or macro-anatomical and/or biomechanical information in a patient receiving an anabolic or antiresorptive drug and be evaluated at different time intervals T1, T2,... , Tn and changes in said bone structure and/or macro-anatomical and/or biomechanical parameters can be used in order to assess therapeutic efficacy. A physician can use this information to adjust the dose of a drug administered (e.g., for treatment of osteoporosis) or to change the drug regimen. [0139] Other techniques using x-ray information such as tomosynthesis can also be used for measuring bone structure and for selecting said therapy or monitoring said therapy.
[0140] Bone structure can be measured using a number of different technical approaches. These include but are not limited to the Hough Transform, analysis of density and size distribution of trabeculae, multidimensional classification schemes, mean pixel intensity, variance of pixel intensity, Fourier spectral analysis, fractal dimension and morphological parameters.
3.1.1.0. Hough Transform [0141] The Hough transform (See, e.g., Hough "Machine analysis of bubble chamber pictures" in International Conference on High Energy Accelerators and Instrumentation. 1959. CERN) can be used to detect geometric objects in binary images. As an entirely new approach to assessing bone structure and/or macro- anatomical, the invention includes the use of such methods to analyze direction and length of structures in bone images. For this purpose, the region of interest (ROI) can be blurred with a Gaussian filter. The pixel values of the filtered ROI can then be subtracted from those in the original ROI, and the value 128 can be added at each pixel location. This results in an image with a mean gray value of 128, which is also used as a threshold to yield a binary image in which the trabeculae are represented by the white pixels.
[0142] After a skeletonization step, a Hough transform with the line parameterization p = x cos θ + y sin θ can be applied to the binary image in order to find straight line segments. Here p is the perpendicular distance of the line from the origin and θ is the angle between the x-axis and the normal. Each point (x,y) in the original image is transformed into a sinusoidal curve p = xcosθ + ysinθ in the (p,θ) plane of the transformed image (see Fig. 7)). Ideally, the curves from collinear points in the original image intersect in a single point in the transformed image. However, the (p,θ) plane can be divided into bins, where each bin counts the number of transformed curves that pass through it. This number corresponds to the number of collinear points on a line segment in the original image, and thus the length of this segment. Furthermore, the transformed image provides information on the predominant angles of the line segments in the original image (see Fig, 8).
[0143] The average length and the variance of the line segments, which are calculated for all bins with a count above a certain threshold, can be used as structural parameters for the shape of the bone trabeculae. Average length as well as the variability of the length to decrease in patients with osteoporosis. The threshold has the effect that only segments of a certain minimal length are included in the calculation. Choosing the threshold so that it provides the best discrimination between healthy and diseased individuals can be readily determined by one of skill in the art in view of the teachings herein.
[0144] The "center of mass" of the transformed image h, given as:
Figure imgf000045_0001
in which each bin is interpreted as an element with a mass equivalent to its count, is a way to measure the predominant angles of the trabecular segments. The angle at cm is measured with respect to the alveolar rim to obtain a standardized value. More importantly, the variance of the segment angles (again measured after thresholding the bin counts) provides information on the anisotropy of the trabecular structure.
Histomorphological studies of osteoporotic vertebrae have shown that the variability of trabecular orientations decreases with the disease.
3.1.2.0. Analysis of Density and Size Distribution of Trabeculae
[0145] Morphological operations such as variations of dilation and erosion and combinations thereof can also be used to detect the size of structures in gray scale or binary images. For example, a skeleton operator can be used to extract and quantify trabeculae of different sizes and directions, which results in a measure of the size distribution of trabecular structures. This skeleton operator is based on the work described in Kumasaka et al. (1997) Dentomaxillofac Rad 26: 161 -168 and works as follows:
[0146] Let a two-dimensional structuring element e be a function over the window
- m ≤ ij ≤m (m>0) with E(i,j) e {o,i} . The dilation operator sets a pixel value f(x,y) in a gray scale image f to the maximum of those values within the window of size m, for which e(/,y)=1 : {f(x + i,y + j)\E(i, j) - 1}
Figure imgf000046_0001
[0147] The erosion operator is defined accordingly, using the minimum instead of the maximum:
Figure imgf000046_0002
[0148] 'Opening' is the operation of maximum search after minimum search: fE =(f ® E) ® E
[0149] Accordingly, the 'closing' operation is defined as the minimum search after maximum search:
fE = (f Φ E) ® E [0150] If a fixed structuring element E\ is given as Eι(/ )=1 for -ι < /, <l , the skeleton operation is then defined as
ST beculae (f) = (f ® EZ ) ' (f ® El)E (V
[0151] £2 is another structuring element that is of circular shape and can be varied in size, and therefore renders the skeleton operator sensitive to the size of the structures in the image. The erosion of f with E2 erases the structures that are smaller than E2 and extracts those trabeculae that are at least equal in size. Those structures that are exactly equal in size is reduced to a width of one pixel. The opening step with \ causes all structures that are one pixel wide to disappear (second term in (1)). After subtraction of this term from the first one, only those trabecular structures that exactly match the size of E2 remain. Finally, the image is thresholded with a level of 1. The effect of this operator is illustrated in Fig. 9.
[0152] Fig. 10 demonstrates the use of the skeleton operator with the same structural element diameters as in Fig. 9 on a gray scale region of interest from a dental x-ray containing trabecular bone. The number of bright pixels in the binary images resulting from each skeleton operation corresponds to the portion of trabeculae of the particular size in the original image. If the percentage of the bright pixels with respect to the total number of pixels in each skeletonized image is plotted against the diameter of E2, the "center of mass" of the curve, i.e. the predominant structure size, can be used as an index to discriminate between osteoporotic and healthy bone.
[0153] Furthermore, the skeleton operator is preferably optimized and extended to detect structures that are oriented only in a specific direction. This can be achieved by adding erosion operations to the skeleton operator with structural elements in which, for example, only the diagonal pixels are set to 1. [0154] This can be used to calculate an anisotropy index, similar to the one derived from the Hough transform. Both anisotropy indices are tested with respect to their potential to distinguish healthy from osteoporotic bone.
[0155] In a similar manner the sizes of the marrow spaces can be examined. The skeleton operator is then defined as
SMamw{f) = <f ® E2y (f ® E2)El
[0156] In addition, the watershed segmentation can be applied to background subtracted gray level structures on x-ray images to characterize the homogeneity of trabecular structures. This process takes into account the gray level contrast between structures to define marrow spaces. The watershed segmentation, when applied to background subtracted bone x-ray images, defines regions with lower gray levels (or basins) surrounded by higher gray level structures (or ridges), as marrow space, in accordance to the spatial extend and gray levels of ridges. Therefore, the size and orientation of marrow space segments defined by this procedure can be related to the spacing, relative density and orientation of adjacent trabecular structures. The segments of marrow space generated using the watershed segmentation can be measured for their area, eccentricity, orientation, and the average gray level on the x- ray image within the segment. The statistics (for example mean, standard deviation, minimum, maximum, and mode) foreach of the segment characteristics can be measured. These statistics can be selected to reflect the homogeneity of marrow space and trabecular structures, and can be used to detect presence of abnormal distribution of marrow space and trabecular structures. 3.1.3.0. Multidimensional Classification Schemes [0157] In certain embodiments, it is preferred to use multiple indices to measure bone structure and/or macro-anatomical parameters. Thus, novel approaches that integrate one or more suitable indices can be employed. The indices can be optimized and incorporated into a multi-dimensional classification scheme, for example using a nearest neighbor classification. Cover et al. (1967) IEEE Trans Inform Theory 13(1 ):21 - 7. (See, Example 3). [0158] Table 1 provides examples of different analyses and anatomical / physiological correlates of the parameters that can be measured. TABLE 1
Figure imgf000048_0001
3.1.3.1 MEAN PIXEL INTENSITY [0159] Mean pixel intensity is a general parameter for the bone mineral density. The degree to which x-rays passing through bone tissue are absorbed depends on the bone's mineral content. Bone with a higher mineral density absorbs a larger portion of x-rays, and therefore appears brighter on an x-ray image. [0160] The mean pixel intensity f(χ,y) in the ROI is calibrated against an aluminum calibration wedge that is included in the image. The log of the average pixel intensity for each thickness level of the calibration wedge is plotted against the thickness, which allows f(χ,y) to be converted into a standardized aluminum thickness equivalent, which is used as the value for this parameter. The automatic recognition of the different thickness levels of the calibration wedge are made possible by different geometric patterns scribed into the wedge which are shown in the x-ray image and can be localized automatically.
3.1.3.2. Variance of pixel intensity [0161] The variance of the pixel gray values in the ROI, var f(x,y), describes the variability of the pixel intensities and can therefore be a measure of the degree of trabeculation. A loss of trabecular bone is predicted to be reflected by a decreased var f(x,y). Southard & Southard (1992) Oral Surg Oral Med Oral Pathol 74: 111 -117.
3.1.3.3. Fourier spectral analysis [0162] The spatial frequency spectrum of a texture provides information about its coarseness. Fine textural structures and edges in an image correspond to high frequencies in the frequency domain, while coarse textures are represented by lower frequencies. Applied to x-ray images of trabecular bone, this means that a region with coarse or little trabeculation should exhibit a Fourier spectral energy concentration at low spatial frequencies, whereas a region of fine trabecular structure should show a spectral energy concentration at high frequencies.
[0163] Typically, the 2-dimensional Fourier coefficients for the selected ROI.
These 2-dimensional coefficients are used to determine a 1 -dimensional power spectrum F(u) by averaging all coefficients over circles with radii that correspond to the discrete spatial frequencies u. The mean transform coefficient absolute value
Figure imgf000049_0001
and the mean spatial first moment M of the absolute coefficients are
Figure imgf000050_0001
determined after exclusion of the first ("DC") coefficient. Mi provides a measure for which frequencies contribute most to the energy of the spectrum, similar to the "center of mass" of a geometric object.
3.1.3.4. Fractal dimension
[0164] A different approach to analyze the texture in an image is by fractal analysis. Fractals are objects that exhibit certain statistical self-similar or self-affine properties, so that a portion of the object, scaled to its original size, has for example the same surface area (3-d) or the same perimeter (2-d) as the original object. In the context of fractal analysis, the gray values in a particular texture can be interpreted as an altitude, and the resulting 3-dimensional surface is analyzed (Fig. 11).
[0165] Fractal dimension (fd) is the rate at which the perimeter or surface area of an object increases as the measurement scale is reduced. Russ "The Image
Processing Handbook," Third edition ed. 1999, Boca Raton: CRC press. It is a measure for the complexity of a boundary or surface and corresponds to the intuitive notion of an object's roughness. Without being bound by one theory, it is postulated that osteoporotic trabecular bone, in which trabeculae become thinner and lose their continuity, and therefore complexity is increased, should have a higher fractal dimension than healthy bone.
[0166] The results from the several ways in which FD can be measured are not comparable. Thus, various methods can be tested to determine which one (or combination) provides the best discrimination between normal and osteoporotic subjects. [0167] The first method is applied in the frequency domain after calculation of the
ROI's 2-D power spectrum using a fast Fourier transform (FFT). From the 2-D Fourier coefficients the 1-D power spectrum is produced as described above for the Fourier analysis. When this 1-D power spectrum is plotted as the logarithm of the power versus the logarithm of the frequency, it must have a negative slope of magnitude b with 1 < b < 3 according to fractal theory. The FD value is then calculated as FDX =3.5 -b/2.
[0168] Another approach, the Minkowski method, measures the difference
(summed over the ROI) between an upper and lower envelope fitted to the surface as a function of the size of the neighborhood used. Peleg et al. (1984) Anal Mach Intell 6(4):518-523. If δ (δ=1 ,2,3,...) is the distance between the envelopes and the surface, then the upper envelope uδ and the lower envelope lδ are given by
Figure imgf000051_0001
where f(i,j) is the gray value of pixel (i,j) in the ROI. The log of the area A(δ), plotted against log(δ), yields a line with a negative slope of magnitude b'. The fractal dimension is then given by FD2 -2 -V . The area is calculated as A(β) = — — — with
vδ = ∑ (^ A) -^O', J)) - (ιj)eROI
3.1.3.5. Morphological Parameters
[0169] While the previous features and parameters provide rather general information on trabecular bone structure, the following examples describe more detailed aspects.
[0170] The gray scale region of interest is first binarized. As described in White et al. (1999) Oral Surg Oral Med Oral Patholo Oral Radiol Endod 88:628-635, this can be achieved in the following way: The ROI is blurred by means of a Gaussian filter. The blurred ROI is then subtracted from the original ROI, and the value 128 is added at each pixel location. This results in an image with a mean gray value of 128, which is also used as a threshold, resulting in an image, in which trabeculae are white and marrow space is black. [0171] From this binary image, the total number of white pixels represents the trabecular area, which is calculated as a percentage of the total ROI area. The number of pixels on the outer trabecular border measures the peripheral length of the trabeculae. The same parameters can be measured for the marrow space by counting the black pixels. [0172] After skeletonization of the binary image, the total length of the trabeculae is determined by the total number white pixels. Furthermore, counts of the terminal points and of the branch points are expressed as a proportion of trabecular length. An estimate of the average length of the trabeculae is calculated as the ratio of total trabecular length and the sum of terminal points and branch points. 3.1.3.5. Markov Random Fields
[0173] In certain embodiments, Markov random fields can be used as models for osteoporosis detection from radiographic images and for fracture risk prediction. As noted herein, osteoporosis is typically manifested in radiographic images by structural changes that can be used for computer-aided detection and characterization. Thus, the detection and/or characterization of osteoporosis from radiographic images relies on the measurement and analysis of a feature or set of features relating to the density of the bone or trabecular structures present in an image.
[0174] Markov random fields can be used to analyze and detect structure density changes by either modeling particular joint feature distributions ({F1 , F2, ... , Fn}) as they are estimated at each image element or image neighborhood (Figure 19), or by modeling the actual radiographic manifestation of particular structural definitions (e.g. trabeculae) (Figure 20). [0175] In the first case of estimation at each image element or image neighborhood, the Markov random field framework is used for a context-based feature analysis/discrimination approach which takes into account local relationships between the features and effectively compensating for space-varying processes (e.g. variable soft tissue or missing or incomplete data due to boundaries) that can affect the relative values of the features taken into account. (Buntine (1994) "Operations for learning with graphical models," J. Artificial Intelligence Res. December: 159-225).
[0176] This approach can also be used for predicting most likely fracture paths based on the analysis of trabecular structure nodes and their related feature sets by defining the most likely chains of joint feature sets. The analysis framework can be a « P( ROI I Normal ) Λ- =
Likehood Ratio approach: P( OI I Abnormal )) where P( ROI | . ) is given by the corresponding Markov random field model.
[0177] Another analysis approach is through the implementation and training of
Bayesian networks, for example as described in Heckerman D (1996) "A tutorial on learning with Bayesian networks," Microsoft Research Technical Report, MSR-TR-95- 06. based on available test case data.
[0178] Markov random fields can also be used to model the manifestations of the structures in an image in probabilistic terms, (Geman et al. (1984) "Stochastic relaxation, Gibbs distributions, and the Bayesian restoration of images," IEEE Transactions on Pattern Analysis and Machine Intelligence 6:721 -741 ; Besag (1986) "On the statistical analysis of dirty pictures," Journal of the Royal Statistical Society, 48(3):259-302). As depicted in Figure 20, each of the image components (noise and characteristic structure) has associated probabilistic models, P(N) and P(l | T) respectively, that describe the spatial distribution of the gray-level intensity. For example, a common assumption for the noise component in digital/digitized radiographs is to consider Normal or Poisson distributed pixels. The nomenclature for the distribution of the characteristic texture P(l | T) is such as to reflect that the corresponding probability distribution of the region I is conditional ( expressed by the symbol | )on the characteristic structure present T. The analysis tools for such a probabilistic framework are provided by the laws of probability and specifically Bayes' Rule shown in Figure 21. Bayes' rule can be described as the rule according to which our knowledge about the presence of a given characteristic structure in an ROI is updated (a-posteriori information represented by the probability distribution P(T | ROI) ), based on experience of how often (or likely) each characteristic structure is present (a priori information represented by the probability distribution P(T) ) and knowledge of how the sources of noise and variability change the manifestation of the corresponding characteristic structure (knowledge of the likelihood, thus also called the likelihood function P( ROI | T ), of the ROI image given the possible characteristic structures and overlapping degrading components). Figure 21 illustrates that simply selecting the structure with the maximum a-posteriori information can be used as a decision criterion.
[0179] To define the likelihood function P(ROI | T), Markov random field modeling may be employed. Markov random fields are specific multidimensional random processes that satisfy what is known as the Markov property. The Markov property simply states that in a random series of events, each event can be predicted and depends only on a limited set of events. This property is convenient and intuitive for the modeling and analysis of structures in images. It basically states that if the distribution of pixels in an ROI can be modeled as having the Markov property, then in order to determine if a pixel belongs to a given structure, only a limited number of neighboring pixels are necessary.
[0180] Random fields having the Markov property confer the additional benefit of having an associated Gibbs probability distribution given by the following equation:
U(s, '2 ' -) P( ROI = s1 , s2 J ... ,sm )= [0181] Where the function:
U(s1 l 82 l ...) = ∑ V(sI) + ∑ V(sI I sJ ) + ∑ V(s| f sJ I sk) + ...
depends on functions V (called potentials) of local neighboring elements called cliques: {si} . {si , Sj} ) {si , Sj l Sk} ) ... e C
[0182] The significance of cliques is that they are the fundamental elements that can be used to reflect specific spatial distribution properties of a structure of interest, such as for example vertical, horizontal and diagonal geometries. Furthermore, the Markov property is manifested very conveniently as each image pixel can be expressed in terms of the cliques in a local neighborhood:
-V N, , P(S. sκ e N. ) = u © rW Sc [0183] The model parameterization for the families of images characteristic of a particular structural density grade and definition of a priori information can be done either by estimation from available patient data thus defining empirical priors or by implementing physical and stochastic models that are based on the image generation process. 3.1.4.0. Overlapping windows processing
[0184] Furthermore, two or more overlapping ROIs can also be defined and used to analyze any given image. In other words, bone density, microarchitecture , macro- anatomic and/or biomechanical (e.g. derived using finite element modeling) analyses can be applied within a region of predefined size and shape and position. This region of interest may also be referred to as a "window." Processing can be applied repeatedly within the window at different positions of the image. For example, a field of sampling points may be generated and the analysis performed at these points (Fig. 16). The results of the analyses for each parameter can be stored in a matrix space, e.g., where their position corresponds to the position of the sampling point where the analysis occurred, thereby forming a map of the spatial distribution of the parameter (a parameter map). The sampling field can have regular intervals or irregular intervals with varying density across the image.
[0185] The amount of overlap between the windows can be determined, for example, using the interval or density of the sampling points (and resolution of the parameter maps). Thus, the density of sampling points is set higher in regions where higher resolution is desired and set lower where moderate resolution is sufficient, in order to improve processing efficiency. The size and shape of the window would determine the local specificity of the parameter. Window size is preferably set such that it encloses most of the structure being measured. Oversized windows are generally avoided to help ensure that local specificity is not lost.
[0186] The shape of the window can be varied to have the same orientation and/or geometry of the local structure being measured to minimize the amount of structure clipping and to maximize local specificity. Thus, both 2D and/or 3D windows may be used, depending on the nature of the image and data to be acquired. [0187] In another embodiment, bone density, microarchitecture, macro-anatomic and/or biomechanical (e.g. derived using finite element modeling) analyses can be applied within a region of predefined size and shape and position. The region is generally selected to include most or all of the anatomic region under investigation and, preferably, the parameters can be assessed on a pixel-by-pixel basis (e.g., in the case of 2D or 3D images) or a voxel-by-voxel basis in the case of cross-sectional or volumetric images (e.g., 3D images obtained using MR and/or CT). Alternatively, the analysis can be applied to clusters of pixels or voxels wherein the size of the clusters is typically selected to represent a compromise between spatial resolution and processing speed. Each type of analysis may yield a parameter map.
[0188] Parameter maps can be based on measurement of one or more parameters in the image or window; however, parameter maps can also be derived using statistical methods. In one embodiment, such statistical comparisons can include comparison of data to a reference population, e.g. using a z-score or a T-score. Thus, parameter maps can include a display of z-scores or T-scores.
3.1.4.1. Analysis and selection of parameter maps
[0189] The parameter maps can represent individual parameters or combinations of parameters such as density, microarchitecture macro-anatomical parameters or biomechanical parameters, for example derived using finite element modeling, are useful in identifying regions or patches that have similar characteristics. For instance, depending on their position, shape, size, orientation, and extent particular regions or patches that exhibit similar characteristics (e.g., values at high or low ranges of the data set) typically represent regions of bone with different properties, for example areas of stronger or weaker areas. Therefore, parameter maps can be used to generate virtual fracture lines that aid in predicting areas of the bone that may be subject to an increased risk of fracture. One or more parameter maps can be selected by statistical analysis of results from in vitro mechanical loading tests or by other means (e.g. from cross-sectional or longitudinal studies in osteoporosis subjects, in particular those developing fractures). Selection can be based, for example, on patch location, shape, size, orientation and extent that best correlates with location of actual fracture lines and/or for having parameter values that are best correlated with fracture risk, the incidence of osteoporotic fractures or fracture loads. 3.1.4.2. Fracture path prediction
[0190] When there are multiple parameter maps that correlate well with fracture line, a multivariate regression model can be fitted to generate a composite parameter map derived from 2D or 3D data sets, e.g. x-rays, digital tomosynthesis, CT and MRI, using the techniques described herein and/or statistical methods known to those of skill in the art. A parameter map can be used to predict the overall bone strength or fracture risk or fracture load by analyzing the predicted fracture paths. A predicted fracture path is defined here as the hypothetical path where fracture would most likely to occur, if sufficient forces are applied in one or more particular directions.
[0191] In certain embodiments, a watershed segmentation can be applied to the selected or composite parameter map. Watershed segmentation can be applied to 2D images as well as to 3D (cross-sectional or volumetric data obtained, for example, from CT or MR). The boundaries of watershed segmentation generally form along the ridges on the parameter map, i.e., along the peak values. For a parameter that is positively correlated to bone strength or fracture load, i.e., higher values correspond to stronger bones, the inverse value of the parameter is used to generate the watershed boundaries so that the boundaries would form along valleys (local minimum) of parameter maps. The nodes of watershed boundaries can be identified and segmented to separate the watershed boundaries into segments (Figure 17). Each of these segments can be assigned a strength value or fracture load value which is a composite value of one or more parameter maps underlying the segment. The length, orientation, and position of segments can be used as normalizing factors for the strength values.
[0192] The nodes and segments of the watershed boundaries may be labeled, traced, measured, and recorded in a form of data structure, for example, a graphical structure. The strength values and interconnect relationships are also stored for each segment. To identify the most likely fracture paths, a search strategy, for example, the depth-first search (Russell S., Norvig, P., Artificial Intelligence: A modern approach. 1995, NJ: Prentice Hall, pp.77), is propagated through the data structure to determine the paths of least resistance from one surface of the bone to another opposite surface restricted by a predefined solid angle. Alternatively, an artificial neural network can be trained to predict fracture paths given the parameter maps as inputs.
3.1.4.3. Fracture risk prediction [0193] Having predicted one or more fracture paths, additional processing may be performed, typically with a new processing grid that has high concentration of nodes along the predicted fracture paths with a different window size and/or shape. Macro- anatomical parameters such as cortical thickness can be evaluated (in two or three dimensional images) with higher resolution at the exits of fracture paths. Parameters that are the best predictors of fracture risk can be evaluated along the predicted fracture paths. These parameters, including density, microarchitecture, macro-anatomical measurements and biomechanical parameters, are selected by statistical analysis of results from in-vitro mechanical loading test or by other means, e.g. using cross- sectional or longitudinal studies in osteoporosis subjects, in particular those developing fractures, for being highly correlated to the magnitude of one or more mechanical properties of bone, for example in one or more particular loading force directions, or for being highly correlated with fracture risk, incidence of new fractures or fracture loads. The mechanical properties include but are not limited to yielding load, stiffness, and Young's modulus.
[0194] The values of parameters along the predicted fracture paths may be compared against the statistical distribution of the population. The z-score and T-score of each parameter relates to the risk of fracture occurring in a particular predicted fracture path. Thus, a fracture risk score can be assigned to that fracture path. The predicted fracture paths can also be associated with the clinical definition of common fracture types. The overall fracture risk can then be evaluated by weighing fracture risk score of each predicted fracture path with the probability of a particular type of fracture occurring. Figure 18 depicts an exemplary summary of this process.
3.1.5.0. Biomechanical assessment [0195] The features and values extracted from the processing of density, microarchitecture, macro-anatomical parameters can be used as the inputs for biomechanical modeling, for instance modeling using finite element analysis. Finite element modeling (FEM) can be used as a surrogate for the physical mechanical properties of bone or composite of bone and implants. Briefly, FEM involves the division of a structure or object into discrete shaped elements, where the mechanical behavior of each element can be described by precise mathematical equations. Structural finite element analysis (FEA), a particular subset of FEM, is the calculation of the mechanical behavior (stress and strain) at any point within the structure under specific loading conditions. The foundation of every finite element model is the two-dimensional or three-dimensional data of the object or structure
[0196] Examples of microarchitecture and micro-anatomical features that can be used as input mesh for finite element analysis include but are not limited to the actual and derivation of image or data structures of trabecular structures, image or data structures of cortical bone, image, data structures of trabecular skeleton or parameter maps derived from overlapping window processing. As described herein, the input features can be obtained from 2D and/or 3D images. The application of simulated force can be in one or more directions, and is typically associated with the actual force components that would occur in a fracture incident. The finite element analysis provides an estimate of load and direction of fracture for each fracture incident scenario. Fracture risk is estimated by weighing the fracture loads with the probability of each fracture scenario occurring. Further, the fracture paths estimated by finite element analysis can be used as inputs to the analysis of density, micro-architecture, macro- anatomical features. For example, density, micro-architecture, macro-anatomical features can be measured in areas of fracture paths predicted by finite element modeling. Conversely, finite element analysis can be combined with additional image and clinical data to determine fracture risk by predicting if the bone would fracture, given the force components that would occur in a fracture incident. [0197] Bone fracture risk can be evaluated using one or a composite of more than one dependent or independent results of analysis or statistical methods. An example of this combination is the weighted average score of density, microarchitecture, macro-anatomical, finite element analysis and clinical risks factors such as weight, height, history of fracture, family history of fracture, and the like. [0198] Finite element modeling can be applied to all of the bony structures included in an image. Preferably, however, finite element modeling is typically applied in selected subregions. In certain embodiments, finite element modeling is applied in areas coinciding with or bordering with the predicted fracture path, for example based on micro-structural or macro-anatomical measurements. By combining biomechanical assessment of bone properties with density, micro-architectural and macro-anatomical assessment, the prediction of fracture risk and/or the correlation with fracture load can be improved. Finally, regional assessment of biomechanical properties can also improve the accuracy of the fracture path prediction. [0199] Biomechanical assessment can also include more traditional approaches estimating levers and forces at the macro-anatomical level, e.g. measurement of moments, shear and compressive forces based on macro-geometric parameters of the bone and anticipated loads or stresses. These more traditional approaches can be combined with finite element modeling, measurements of density, bone structure, and macroanatomical parameters, e.g. cortical thickness, thereby improving assessment of bone strength and fracture risk and improving the correlation with fracture loads and, ultimately, incident new fractures.
[0200] As will be appreciated by those of skill in the art, the macroanatomical parameters that are measured can change depending on the region of interest to be measured. For example, when studying a portion of the spine, the user can combine bone structure measurements with macroanatomical measurements and/or FEA and/or other biomechanical measurements and/or bone mineral density. The actual macroanatomical measurements that are used in the spine can be, for example, the inner pedicle distance, the outer pedicle distance, the vertebral height (either anterior, central, posterior, left, right, or a combination thereof), the vertebral anterior-posterior diameter (taken either in the superior, middle, inferior, or another location), the vertebral right to left diameter (taken in either the superior, middle, inferior or another location), the vertebral diameter (taken in an oblique plane), the vertebral diagonal (using, e.g., internal cortex or external cortex), the thickness of the superior endplate (taken, e.g., anteriorly, centrally, posteriorly, from the left, from the right, or a combination thereof), or using the thickness of the inferior endplate (again taken, e.g., anteriorly, posteriorly, from the left, from the right, or a combination thereof).
[0201] Similarly, when studying the knee and tibia, the user can combine bone structure measurements with macroanatomical measurements and/or FEA and/or other biomechanical measurements and/or bone mineral density. However, as will be appreciated by those of skill in the art, the bone structures used for measurements when studying the knee and tibia region change due to changes in anatomy. Thus, in studying the knee and tibia region, suitable measurements are taken from, for example, the anterior-posterior diameter of the bone using the inner or outer cortex, or a combination thereof, the medial-lateral diameter of the bone using the inner or outer cortex, or a combination thereof, the cortical thickness in various locations, the standard deviation of cortical thickness, the subchondral bone thickness in various locations, and/or a combination thereof. [0202] Cases may arise where the macroanatomical measurements are used to normalize bone structure or bone density measurements. For example, in the tibia, bone structure and/or bone density measurements could be altered if the patient has a tibia that is thick in the anterior-posterior dimension (e.g., thicker than average). The macroanatomical measurements are then used to normalize the tibial measurement by, for example, forming a ratio between the thick tibial measurement in the anterior- posterior direction and another measurement.
3.2.0.0. Soft Tissue
[0203] Variations in soft tissue thickness can be significant in analyzing and evaluating bone density and bone structure, macro-anatomical parameters and biomechanical parameters, e.g. those derived using finite element modeling, in x-rays. Accordingly, the invention also includes methods and devices for correcting for soft tissue in assessment of bone structure or dense tissue, particularly for diagnosing and/or predicting osteoporosis or other bone conditions. [0204] In certain embodiments, the x-ray image is a dental x-ray image and such correction methods involve (a) interrogating at least a portion of a subject's mandible and/or maxilla with an x-ray detector; (b) producing an x-ray image of the interrogated mandible and/or maxilla; (c) obtaining data from the x-ray image regarding bone density or bone structure; (d) interrogating the surrounding soft tissue to determine soft tissue thickness; and (e) correcting the data obtained from the x-ray image by correcting for soft tissue thickness. Such study groups include: non-osteoporotic premenopausal, non-osteoporotic postmenopausal, osteoporotic postmenopausal patients. It will be apparent, although exemplified with respect to dental x-rays, that many of the methods described herein can be applied to other x-ray images, e.g. hip or spine x-ray images.
[0205] Soft tissue thickness measured in a subject can also be compared to reference soft tissue thickness obtained from a control population (e.g. age-, sex-, race-, or weight-matched normal subjects). Reference soft tissue thickness can be generated by measuring soft tissue thickness in healthy subjects with normal vascular, cardiac, hepatic, or renal function and no other underlying medical condition. Reference soft tissue thickness can be expressed as but are not limited to, mean and standard deviation or standard error. Reference soft tissue thickness can be obtained independently for patients 15-20, 20-30, 30-40, 40-50, 50-60, 60-70, 70-80, and 80 and more years of age and are preferably obtained separately for men and women and for race (e.g. Asian, African, Caucasian, and Hispanic subjects). Additionally, reference soft tissue thickness can be obtained for different subject weights within each age, sex, and racial subgroup.
[0206] Individual patients can be compared to reference soft tissue thickness. If patient's soft tissue thickness is elevated, a correction factor can be applied. The amount/magnitude of correction factor is influenced by the magnitude of increase in soft tissue thickness that can be influenced by the magnitude of fat, fibrous, and muscle tissue contribution. Clinical study groups can be evaluated to generate databases for further study or to generate more refined correction factors. Such study groups include: non-edematous non-osteoporotic premenopausal, non-edematous non-osteoporotic postmenopausal, non-edematous osteoporotic postmenopausal; edematous non- osteoporotic premenopausal, edematous non-osteoporotic postmenopausal, and edematous osteoporotic postmenopausal patients. In each study group the following procedures can be performed for comparison: dual x-ray absorptiometry ("DXA") of the spine, hip, or calcaneus, along with SOS and BUA measurements or quantitative computed tomography ("QCT"). Thus, correction for soft tissue thickness can also improve the accuracy and discriminatory power in the analysis of x-rays and other x- rays. Such methods can also be used to identify population with an increased or decreased risk of bone conditions such as osteoporosis. 4.0. Applications
[0207] The measurements of bone m ineral density or trabecular architecture and/or macro-anatomical and/or biomechan cal parameters, for example in the mandible or maxilla or in the hip or in the sp ne, can be used to derive an assessment of bone health in any subject. Additionally, the analysis and manipulation of data from x- rays allows for the assessment of bone health that in turn can be used to prescribe a suitable treatment regime. Efficacy of a treatment regime can also be assessed using the methods and devices described herein (for example, using measurements of bone mineral density or trabecular architecture and/or macro-anatomical and/or biomechanical parameters in the mandible or the maxilla or the hip or the spine taken at two separate time points T1 and T2 to detect any difference in bone mineral density or trabecular architecture).
[0208] In addition, the methods described herein permit, for example, fully automated assessment of the structural organization and architectural arrangement of trabecular bone and/or macro-anatomical and/or biomechanical parameters on standard hip radiographs as well as improved tools for monitoring progression of osteoporosis and therapeutic response. In certain embodiments, the methods involve binarizing and skeletonizing trabecular bone using morphological operators with detection of branch points and endpoints of the skeleton network and classification into free-end segments and node-to-node segments. In other embodiments, the methods involve measuring trabecular density, trabecular perimeter, trabecular bone pattern factor, segment count, segment length, angle of segment orientation and ratio of node-to-node segments to free-end segments based on the binarized and/or skeletonized images. In still further embodiments, the methods involve (a) measuring trabecular thickness using a Euclidean distance transform (see, also Example 3); (b) assessing trabecular orientation using a 2D Fast Fourier Transform; and/or (c) creating a bone structure index for diagnosing osteoporosis or for predicting fracture risk combining at least two or more of these structural parameters. [0209] In certain embodiments, the radiograph is of a subject's hip. Furthermore, to help control the influence of radiographic positioning on the accuracy of bone structure and/or macro-anatomical and/or biomechanical measurements, the methods may include one or more of the following: evaluating the angular dependence of bone structure measurements in the hip, for example by comparing antero-posterior radiographs of the hip joint in healthy to osteoporotic patients (subjects) with the femur radiographs in neutral position and in various degrees of internal and external rotation or by obtaining radiographs of the hip with different degrees of tube angulation. Bone structure and/or macro-anatomical and/or biomechanical measurements can be compared between the different positions to determine which bone structure parameters show the least dependence on radiographic positioning and/or using a foot holder to fix the patients' foot in neutral position in case pair wise coefficients of variation between the results for the 0° neutral position and a 15° internal or external rotation position exceed 10% for the majority of the structural parameters measured.
[0210] In other embodiments, methods of monitoring bone structure and/or macro-anatomical and/or biomechanical parameters over time (e.g., longitudinally) are also provided, for example to assess progression of osteoporosis and/or response to therapy. In certain embodiments, the methods involve automated placement of regions of interest (ROI) in the hip joint, for example by creating and using a general model of the proximal femur that includes six defined regions of interest (ROI's).
[0211] The methods described herein, which allow, in part, for the measurement of bone structure are useful in both the diagnosis and treatment of osteoporosis. Ultimately, these techniques could help screen large numbers of women at risk for osteoporosis in a highly cost-effective and accurate manner using standard, widely available radiographic equipment without the need for expensive dedicated capital equipment. It is clear that a program of this type would be powerfully enabling for therapeutic intervention with new anabolic or anti-resorptive drugs that are needed to prevent the expected pandemic of osteoporotic fractures.
4.1. Kits
[0212] The invention also provides kits for obtaining information from images, for example for obtaining information regarding bone structure, micro-architecture, macroanatomical and/or biomechanical parameters from an image such as a radiograph. In certain embodiments, the kit comprises one or more computer (e.g., software) programs, for example for receiving, analyzing and generating reports based on the image(s). In further embodiments, the kits can include calibration phantoms, for example calibration phantoms integrated or attachable-to a holder, hygienic cover, x-ray film and/or x-ray film holders. [0213] The invention also provides for therapeutic kits, for example for treating osteoporosis or dental disease. In certain embodiments, the kits comprise a calibration phantom for use with one or more x-ray films, a computer software product, a database, a therapeutic drug and, optionally, instructions for use (e.g., instructions regarding positioning the calibration phantom while taking the x-ray, using the software to analyze the x-ray, dosages and the like. The therapeutic drug can be, for example, anti- resorptive or anabolic.
4.2. Diagnosis and Prediction [0214] In yet another aspect, methods of diagnosing or predicting bone-related disorders (e.g., osteoporosis, Paget's Disease, osteogenesis imperfecta, bone cancers), periodontal disease or oral implant failure in a subject are provided, for example using any of the kits, methods and/or devices described herein. It will be apparent that these methods are applicable to any bone-related disorder including, for example, osteoporosis, bone cancer, and the like, as well as to periodontal disease and implant failure.
[0215] Osteoporosis alone is a major public health threat for 25 million postmenopausal women and 7 million men. In 1995, national direct expenditures for osteoporosis and related fractures were $13 billion. Changing demographics, with the growth of the elderly population, steadily contribute to increasing numbers of osteoporotic fractures and an incipient and potentially economically unmanageable epidemic of osteoporosis. Projections put the total cost of osteoporosis in the United States alone at more than 240 billion dollars per year in 40 years. [0216] Less than 20% of the pat :iients know they have the disease and many fewer receive physician directed specif i ic therapy. A major impediment in successfully dealing with the impending osteoporos is epidemic is not a lack of treatment modalities but the inability to identify persons at risk and who require treatment. The limited access to osteoporosis testing is largely the result of the high cost of the currently available systems resulting in a small installed base limited to hospitals and specialty clinics.
[0217] The devices and methods described herein address these and other issues by providing inexpensive and reliable bone structural analysis screens and resulting diagnosis of bone condition and/or presence of disease. Indeed, while measurements of bone mineral density (BMD) are technically relatively easy to perform, low BMD accounts for considerably less than 100% of fracture risk although it is well established that progressive disruption of trabecular structure and architecture contribute in a major way to fracture risk in older individuals. [0218] Thus, in certain embodiments, the methods comprise using a computer program to analyze bone mineral density or bone structure and/or macro-anatomical and/or biomechanical parameters of an image (e.g., x-ray image) and comparing the value or measurement obtained from the image with a reference standard or curve, thereby determining if the subject has a bone-related condition such as osteoporosis or thereby determining a subject's fracture risk. The image can also include a calibration phantom, for example a calibration phantom as described herein.
[0219] In certain embodiments, measurements of bone structure can be combined or correlated with measurements of macro-anatomical and/or biomechanical parameters (e.g., cortical thickness on a hip x-ray), for example using statistical or mathematical methods, to create an index for the severity of the disease. Subsequently, the index can be used for diagnosing osteoporosis or for predicting fracture risk combining at least two or more of these bone structure or morphological parameters. 4.3. Treatment
[0220] The methods and devices described herein can also be used to develop an appropriate treatment regime for a subject in need thereof. Additionally, the invention allows for the ongoing analysis of the efficacy of a subject's treatment regime.
[0221] Although estrogen deficiency after menopause is one of the most well documented causes of osteoporosis that can be prevented by hormone replacement therapy (HRT), HRT may also cause an increase (approximately 35%) in the risk of breast cancer in long-term users. Lancet (1997)350:1047-1059. Consequently, much effort has been devoted to developing alternative treatments for osteoporosis. Among those treatments, bisphosphonates are becoming increasingly recognized as the treatment of choice. Lin (1996) Bone 18:75-85; Liberman et al. (1995) N Engl J Med 333:1437-1443; Mortensen et al. (1998) J Clin Endocrinol Metab 83:396-402. Another new class of therapeutic agents recently introduced is the selective estrogen receptor modulators (SERMs). Delmas et al. (1997) N Engl J Med 337: 1641 -1647; Lufkin et al. (1998) J Bone Min Res 13:1747-1754. Anabolic therapies such as parathyroid hormone have also been suggested for treatment of osteoporosis. Roe et al. (1999) J Bone Miner Res 14(suppl1):S137, Abst#1019; Lane et al. (1998) J Clin Invest 102:1627-33.
[0222] The combined results of these and other studies suggest that effective treatments for osteoporosis can be developed once the condition is diagnosed. For instance, using any of the methods, kits, and/or devices described herein, the presence of osteoporosis in a subject can be diagnosed and that subject provided with appropriate therapy (e.g., one or more anti-resorptive agents and/or one or more anabolic agents). Periodontal disease can be similarly diagnosed and treatments ranging from oral hygiene practices to surgery can be recommended. Over time, the methods and compositions described herein can be used to assess the efficacy of the selected treatment and the treatment regime altered as necessary. For example, a subject can be given a one-time or ongoing therapy and images evaluated after such therapy to monitor its effectiveness. Thus, in certain embodiments, treatment or monitoring of treatment of bone related disorders are provided.
4.4. Decision Trees
[0223] Thus, diagnosing, predicting, developing treatment regimes, assessing treatment efficacy and the like can be readily accomplished using the methods described herein. In certain aspects, these applications will be accomplished using algorithms or decision trees (also known as logic trees or flow charts). One exemplary decision tree is provided in regard to predicting bone problems. It will be readily apparent that such decision trees are equally applicable to other applications (e.g., designing treatment regimes, assessing treatment efficacy, etc.).
[0224] One exemplary method for predicting bone problems (e.g., osteoporoses, etc.), periodontal disease or oral implant failure employs a decision tree (also called classification tree) which utilizes a hierarchical evaluation of thresholds (see, for example, J.J. Oliver, et. al, in Proceedings of the 5th Australian Joint Conference on Artificial Intelligence, pages 361-367, A. Adams and L. Sterling, editors, World Scientific, Singapore, 1992; D.J. Hand, et al., Pattern Recognition, 31 (5):641-650, 1998; J.J. Oliver and D.J. Hand, Journal of Classification, 13:281-297, 1996; W. Buntine, Statistics and Computing, 2:63-73, 1992; L. Breiman, et al., "Classification and Regression Trees" Wadsworth, Belmont, CA, 1984; C4.5: Programs for Machine Learning, J. Ross Quinlan, The Morgan Kaufmann Series in Machine Learning, Pat Langley, Series Editor, October 1992, ISBN 1-55860-238-0). Commercial software for structuring and execution of decision trees is available (e.g., CART (5), Salford Systems, San Diego, CA; C4.5 (6), RuleQuest Research Pty Ltd., St Ives NSW Australia) and may be used in the methods of the present invention in view of the teachings of the present specification. A simple version of such a decision tree is to choose a threshold bone structure and/or macro-anatomical and/or biomechanical or bone mineral density reading at a particular anatomical landmark (e.g., edge of mandible or maxilla, the end of a tooth root, etc.). If a value is equal to or below the threshold bone data value, then more of the image is evaluated. If more of the image is below the threshold value, then a bone problem, periodontal disease or implant failure is predicted.
[0225] For example, a first level decision is made by the algorithm based on the most recent x-ray images obtained and analyzed as described herein is compared to initial thresholds that may indicate an impending or current bone- or periodontal-related event. For example, the algorithm may compare the current bone structure measurements (time=n) or a predicted bone structure measurement (time=n+1) to a threshold value. If the bone structure measurement is greater than the threshold value then a decision is made by the algorithm to suggest further future x-rays. If the bone structure measurement is less than or equal to the threshold level(s) then the algorithm continues with the next level of the decision tree. [0226] The next level of the decision tree may be an evaluation of the subject's age and/or gender at time (n) that x-ray is taken, which is compared to a threshold bone measurement for "normal" subjects of that age and/or gender. For example, if the subject's bone measurement is greater than the threshold bone structure level for that particular age and/or gender, then a decision is made by the algorithm to prompt further monitoring in the future. If the information on bone structure is less than or equal to the threshold, then the algorithm continues with the next level of the decision tree.
[0227] The next level of the decision tree may be, for example, an evaluation of the subject's soft tissue (e.g., gum) thickness (n), which is compared to a threshold measurement. For example, if the soft tissue is significantly below or above the normal range of thickness, then a decision is made by the algorithm to examine more of the x- ray image or to predict a bone-related problem.
[0228] The decision tree could be further elaborated by adding further levels. For example, after a determination that a bone and/or periodontal events are possible, the subject can be x-rayed again to see if values have changed. Again, age, gender, weight, soft tissue thickness and the like can also be tested and considered to confirm the prediction.
[0229] In such decision trees, the most important attribute is typically placed at the root of the decision tree. In one embodiment of the present invention the root attribute is the current bone structure measurement(s). In another embodiment, a predicted bone structure measurement at a future time point may be the root attribute. Alternatively, bone mineral density and/or implant structure could be used as the root attribute.
[0230] Further, thresholds need not (but can) be established a priori. The algorithm can learn from a database record of an individual subject's readings and measurements. The algorithm can train itself to establish threshold values based on the data in the database record using, for example, a decision tree algorithm.
[0231] Further, a decision tree may be more complicated than the simple scenario described above. For example, if soft tissue of a particular subject is very thick, the algorithm may set a threshold for the bone measurements that is higher or lower than normal. [0232] By selecting parameters (e.g., current or future bone information, etc.) and allowing the algorithm to train itself based on a database record of these parameters for an individual subject, the algorithm can evaluate each parameter as independent or combined predictors of disease and/or implant failure. Thus, the prediction model is being trained and the algorithm determines what parameters are the most important indicators. A decision tree may be learnt in an automated way from data using an algorithm such as a recursive partitioning algorithm. The recursive partitioning algorithm grows a tree by starting with all the training examples in the root node. The root node may be "split," for example, using a three-step process as follows. (1) The root node may be split on all the attributes available, at all the thresholds available (e.g., in a training database). To each considered split a criteria is applied (such as, GINI index, entropy of the data, or message length of the data). (2) An attribute (A) and a threshold (T) are selected which optimize the criteria. This results in a decision tree with one split node and two leaves. (3) Each example in the training database is associated with one of these two leaves (based on the measurements of the training example). Each leaf node is then recursively split using the three-step process. Splitting is continued until a stopping criteria is applied. An example of a stopping criteria is if a node has less than 50 examples from the training database that are associated with it. [0233] In a further embodiment, at each level of the decision in the decision tree, the algorithm software can associate a probability with the decision. The probabilities at each level of decision can be evaluated (e.g., summed) and the cumulative probability can be used to determine whether disease and/or implant failure is predicted. Receiver Operating Characteristic (ROC) curve analysis can be applied to decision tree analysis described above. ROC analysis is another threshold optimization means. It provides a way to determine the optimal true positive fraction, while minimizing the false positive fraction. A ROC analysis can be used to compare two classification schemes, and determine which scheme is a better overall predictor of the selected event (e.g., evidence of osteoporosis); for example, a ROC analysis can be used to compare a simple threshold classifier with a decision tree. ROC software packages typically include procedures for the following: correlated, continuously distributed as well as inherently categorical rating scale data; statistical comparison between two binormal ROC curves; maximum likelihood estimation of binormal ROC curves from set of continuous as well as categorical data; and analysis of statistical power for comparison of ROC curves. Commercial software for structuring and execution of ROC is available (e.g., Analyse-lt for Microsoft Excel, Analyse-lt Software, Ltd., Leeds LS12 5XA, England, UK; MedCalc®, MedCalc Software, Mariakerke, Belgium; AccuROC, Accumetric Corporation, Montreal, Quebec, CA). [0234] Related techniques that can be applied to the above analyses include, but are not limited to, Decision Graphs, Decision Rules (also called Rules Induction), Discriminant Analysis (including Stepwise Discriminant Analysis), Logistic Regression, Nearest Neighbor Classification, Neural Networks, and Naϊve Bayes Classifier.
[0235] All of these aspects of the invention can be practiced separately or in combination. Typically, the use of combinations of the embodiments listed above is more advantageous. Further, although preferred embodiments of the subject invention have been described in some detail, it is understood that obvious variations can be made without departing from the spirit and the scope of the invention.
EXPERIMENTAL [0236] Below are examples of specific embodiments for carrying out the present invention. The examples are offered for illustrative purposes only, and are not intended to limit the scope of the present invention in any way.
Example 1: In vivo reproducibility and in vivo diagnostic sensitivity
A. Dental X-Rays [0237] In order to test in vivo reproducibility of data obtained from dental x-rays, the following experiment was performed. Subjects sat in a dental chair and an x-ray was taken of the area of the incisor teeth and of the molar teeth of the mandible. A calibration phantom step wedge was attached to the dental x-ray film. The dental x-ray film was exposed using standard x-ray imaging techniques for x-rays of the incisor area. The subjects walked around for 15 minutes at which point that test was repeated using the same procedure. [0238] X-ray films were digitized on a commercial flat-bed scanner with transparency option (Acer ScanPremio ST). The regions of interest (ROIs) were placed manually at the same position with respect to the dental roots in all digitized x-rays of the same subject using the NIH Image software program (http://rsb.info.nih.gov/nih- image/Default.html). The reproducibility of the measurement of the average gray values inside the ROIs was determined as the coefficient of variation (COV=standard deviation of measurements/mean of measurements). Overall results are given as root mean square (RMS = "xf jn ) over both subjects. The data are summarized in Table 2.
TABLE 2: Reproducibility of measurements of average gray values in di itized dental x-ra s
Figure imgf000074_0001
[0239] The data show that reproducibility is achieved that is already comparable with that of many ultrasound systems to diagnose osteoporosis. B. Hip Radiographs [0240] To test whether bone texture analysis in hip x-rays can detect differences between normal and osteoporotic bone, sample hip x-ray images were acquired in two patients with a Fuji FCR 5000 computed radiography system (Fuji Medical Systems, Stemford, CT). The first patient had normal bone mineral density in the hip as measured by DXA. In the second patient, femoral neck BMD measured by DXA was one standard deviation below normal. [0241] For x-ray imaging, patients were positioned on the x-ray table in supine position, parallel to the long axis of the table. The patient's arms were placed alongside their body. Patient comfort was ensured with a pillow underneath the patient's neck. However, no pillows were used underneath the knees. The x-ray technologist checked that the patient lies straight on the table by looking from the head down towards the feet (which were placed in neutral position with the toes pointing up. The ray was centered onto the hip joint medial and superior to the greater trochanter.
[0242] Anteroposterior hip radiographs were acquired using the following parameters: Film-focus distance: 100 cm; tube voltage: 65 kVp; exposure: phototimer for automatic exposure or approximately 20 mAs for manual exposure; collimation: limited to the hip joint, including proximal femoral diaphysis; centering: over femoral head (see above); tube angulation: zero degrees. An aluminum step wedge (BioQuest, Tempe, AZ) was included in the images to calibrate gray values before further image analysis. Processing was performed using ImageJ, a Java version of NIH image (http://rsb.info.nih.gov/ij/).
[0243] Six regions of interest were selected manually at the approximate locations as shown in Figure 9. Trabeculae were extracted through background subtraction. The resulting binarized images are shown in the Figures. In a next step, the trabecular bone in the selected regions of interest was skeletonized. [0244] The binarized ROI's in the normal and the osteopenic patient were used to determine the trabecular density ratio (trabecular area vs. ROI area). The following bone structure measurements were obtained from the skeletonized ROI's; mean segment length, total skeleton length (normalized by ROI area), skeleton segment count (normalized by ROI area), and skeleton node count (normalized by ROI area). Results are shown in Tables 3 through 7. TABLE 3; Trabecular Density Ratio (Trabecular Area / ROI Area)
Figure imgf000075_0001
Figure imgf000076_0001
These results demonstrate that the evaluation of trabecular structure reveals significant differences between normal and osteopenic bone and that selective analysis of trabeculae oriented in certain directions in the different ROI allows for the assessment of structures critical for biomechanical stability of the proximal femur. C. Spine Radiographs
[0245] To test whether bone texture analysis in spine x-rays can detect differences between normal and osteoporotic bone, sample spine x-ray images will be acquired in more than one patient. In the spine, the bone structure parameters can be measured in the L1 , L2, L3 and L4 vertebral bodies unless obscured by superimposed ribs, iliac crest or bowel gas. The first patient will provide control data provided the patient has normal bone mineral density in the spine. In the second patient and subsequent patients, spine BMD will be measured.
[0246] Regions of interest will be selected manually at the approximate locations as shown in Figure 22. Trabeculae will be extracted through background subtraction. In a next step, the trabecular bone in the selected regions of interest was skeletonized. FIG. 24 depicts an example of an application of structure extraction and measurement for therapeutic monitoring using spine x-ray. White outline of extracted structure are show in (a) before treatment, and (b) after treatment.
D. Knee/Tibial Radiographs - Osteoporosis
[0247] To test whether bone texture analysis in knee and tibial x-rays can detect differences between normal and osteoporotic bone, sample x-ray images will be acquired in more than one patients. The first patient will provide control data provided the patient has normal bone mineral density or bone structure in the tibia or femur. In the second patient and subsequent patients, joint BMD or bone structure will be measured.
[0248] Regions of interest will be selected manually at the approximate locations as shown in Figure 23. The ROI can, for example, be the region immediately below the tibial plateau subchondral bone. Trabeculae will be extracted through background subtraction. In a next step, the trabecular bone in the selected regions of interest is skeletonized.
D. Knee/Tibial Radiographs - Arthritis [0249] To test whether bone texture analysis in knee and tibial x-rays can detect differences between normal patients and patients with arthritis, sample x-ray images will be acquired in more than one patients. The first patient will provide control data provided the patient has normal bone mineral density or bone structure in the tibia or femur. In the second patient and subsequent patients, joint BMD or structure will be measured.
[0250] Regions of interest will be selected manually at the approximate locations as shown in Figure 23. The ROI can, for example, be the region immediately below the tibial plateau subchondral bone. Trabeculae will be extracted through background subtraction. In a next step, the trabecular bone in the selected regions of interest is skeletonized.
Example 2: Image Processing Techniques [0251] Techniques to analyze structure of trabeculae in different regions of the femoral head, neck, and proximal shaft are developed in Matlab (The MathWorks, Inc., Natick, MA) on PC's. The following techniques (modules) are developed: algorithms for software analysis of density, length, thickness, and orientation of trabeculae in different regions of interest (ROI) in the radiograph and a technique for automated placement of these ROI.
[0252] Six regions of interest are selected in the proximal femur for bone microstructure evaluation. The size and shape of these ROI are designed to capture the local changes of trabecular density and structure (see, e.g., Figure 9), and may reflect the location of the different compressive and tensile groups of trabeculae. Singh et al. (1970) J Bone Joint Surg Am. 1970. 52:457-467. Thus, a classification scheme based on statistical convergence of multiple parameters that would provide a high precision index for predicting hip fractures is developed.
Example 3: Bone Structure Analysis of Hip Radiographs
[0253] The trabeculae in the femur is extracted using the background subtraction method, essentially as described in Geraets et al. (1998) Bone 22:165-173. A copy of the image is blurred with a 15x15 Gaussian filter, and the result represents the non- uniform background. This background image is subtracted from the original image to obtain an image of trabecular structure. This image is then transformed into binary image of trabecular structure by applying a threshold value of 0. An example of the end result is shown in Figure 10.
[0254] In a second step, parameters relevant to the geometry and connectivity of trabecular structure are measured on the trabecular skeleton or centerline. The skeletonization is performed using morphological hit-or-miss thinning for example as described in Soille, "Morphological image analysis: principles and application" Springer, 1998: p. 129-154. The branch points and end points of the skeleton network are detected, and the skeleton segments are classified as free-end segments and node-to- node segments.
[0255] One or more of the following parameters from the binarized and from the skeletonized ROI's are used: trabecular density; ratio of trabecular area to total ROI area; trabecular perimeter; star volume (Ikuta et al. (2000) J Bone Miner Res. 18:271- 277; Vesterby (1990) Bone 11 :149-155); trabecular bone pattern factor (Hahn et al. (1992) Bone 13:327-330); Euclidean distance transform; assessment of trabecular orientation using Fourier analysis; and orientation-specific trabecular assessment. Further, one or more of the following parameters can be measured in each ROI on the network of skeletonized trabeculae as a whole, all skeleton segments, and each type of segment: segment count; segment length; angle of segment orientation; and Interconnectivity Index (Legrand et al. (2000) J. Bone Miner Res. 15:13-19): normalized ratio of the number of node-to-node segments to free-end segments.
[0256] For example, in Euclidean Distance Transform each pixel on the binarized trabeculae is assigned a value equal to its Euclidean distance from the structure boundary. Thus, thicker trabeculae will have larger distance transform values in the center, thereby estimating trabecular thickness calculates the mean of the distance transform values along the trabecular skeleton (see Figure 11). Further, multiplying this value by 2 provides a measurement of trabecular thickness. [0257] Similarly, predominant trabeculae orientation may be evaluated using the
2D Fast Fourier Transform (FFT). A rectangular region is selected within each ROI and multiplied with a 2D Kaiser window before applying the transform (see Figure 12, left). The log of the Fourier magnitude is taken to form an image representing the frequency domain of the ROI. The result is then filtered with a 5x5 Gaussian filter to reduce local variation. An example image is shown in Figure 12, center. The Fourier image is subsequently thresholded at a fixed magnitude level. This binary image is resampled to a square image to normalize the length of the vertical and horizontal axes, and the direction and length of its major axis are determined (Figure 12, right). The angles will be measured with respect to the axes of the femoral neck and shaft. The axes are determined by fitting lines to the two longest segments of the centerline of the binarized femur (see also Figure 14). The ROI's are located such that they include the different groups of compressive and tensile trabeculae in the proximal femur that each can be characterized by a specific direction. A fully automated technique to evaluate the different quantitative structural parameters explained above for those trabeculae in each of the ROI that are oriented in the characteristic direction expected for the particular ROI is developed.
[0258] The orientation of each trabecular skeleton segment is found through the gradient of the line fitted to the skeleton points. Based on this orientation information, only those trabeculae are considered in the evaluation of the structure parameters that are approximately oriented in the characteristic direction for a particular ROI.
[0259] As will be appreciated by those of skill in the art, all measurements can be constrained by one or more desired orientation by measuring only segments within specified angle ranges. The statistics of watershed segments include: number of segments, total area of segments, average area of segments, standard deviation of segment area, smallest segment area, and largest segment area. These segments are, however, general in nature. [0260] When evaluating the hip, additional parameters can be considered.
Parameters include, for example, shaft angle, neck angle, diameter of the femur neck, the hip axis length, the largest cross-section of the femur head, the average thickness of the cortical region within a ROI, the standard deviation of cortical thickness within a ROI, or the maximum or minimum thickness of the cortical thickness within a ROI.
[0261] In contrast, when evaluating the spine, additional parameters to be considered include, for example, all parameters on vertical structures, all parameters on horizontal structures, vertebral cortical thickness, maximum vertebral height, minimum vertebral height, average vertebral height, anterior vertebral height, medial vertebral height, posterior vertebral height, maximum inter-vertebral height, minimum inter- vertebral height, and average vertebral height.
[0262] The knee and tibial region can be evaluated using the additional parameters of: average medial joint space width, minimum medial joint space width, maximum medial joint space width, average lateral joint space width, minimum lateral joint space width and maximum lateral joint space width.
[0263] As will be appreciated by those of skill in the art, the additional parameters listed for these exemplar anatomies above can include other parameters. Additionally, parameters can be evaluated for other anatomies not specifically set forth without departing from the scope of the invention. Example 4: (Multidimensional Classification
[0264] Example 3 describes a number of parameters that are measured to assess trabecular structure in different regions of the proximal femur. In this Example, the different structural parameters are combined in each section, and a single index is determined over all regions of interest. [0265] A training set of hip x-ray images of a group of subjects are divided into the two categories "osteoporosis" and "no osteoporosis", based on previous DXA results. Subsequently, for all x-rays in the training set, the parameters listed in Example 3 are calculated for all regions of interest placed as described in Example 3, resulting in a set of m-dimensionai prototype feature vectors ft = (fn,...,fim)τ for the training set
Figure imgf000082_0001
[0266] For each parameter a single scalar index value is calculated. All index values are combined into one n-dimensional feature vector. In one step, the system is trained with the data from clinical validation studies with premenopausal, postmenopausal healthy and postmenopausal osteoporotic subjects. The subject groups are preferably divided into a "fracture" and a "no fracture" category. The feature vectors calculated from the x-ray images are used as prototype patterns. [0267] For each patient, a feature vector is calculated from the x-ray as calculated for the prototype patterns and an individual patient classified as category C if the majority of the k closest prototype patterns is of the category C. The distance d between the patient's feature vector f = (f f2,...,fn)τ and a prototype pattern p = (p p2,...,pn)τ s defined by the Euclidean norm L2: d f,p) = L2(f,p) = ∑ιffl - PiY2
[0268] The optimum scale for the different parameters is also preferably determined. However, for some parameters differences in the index values between the categories is smaller than for others. Also, the optimum k will be determined. Increasing k is expected to improve the accuracy of the classification, but it has to be smaller than the number of prototypes in each category. The exact percentage value of the majority of the k closest prototype patterns that determines the classification provides a measure for the reliability of the classification. The higher the percentage of prototype patterns from a particular category C, the more significant the information provided by the classification is likely to be. [0269] This classification approach is validated with a series of leave-one-out experiments using the 0° neutral position images of the femoral position study (see Example 8) and the baseline hip x-rays of the short-term in vivo reproducibility study. For these experiments, each subject is preferably used as a test case once. The training set for the system consists of the patterns calculated for all or most of the remaining subjects. The test case is correctly classified using this training set, and the diagnostic sensitivity and specificity of the combination of bone structure parameters is determined.
[0270] In addition to the measurements described above (which provide index values for the parameters "length of trabeculae", "direction of trabeculae and anisotropy", and "trabecular thickness"), additional measurements for other parameters in the classification system that have been explored in the past to study bone density and structure from x-ray, CT, and MR images such as: (1) mean pixel intensity; (2) variance of pixel intensity; (3) Fourier spectral analysis; (4) fractal dimension; (5) morphological parameters such as the trabecular area, trabecular periphery, total trabecular length, number of terminal and branch points, as well as similar parameters for the bone marrow can be used.
Example 5: Automated Placement of Region of Interest (ROI)
[0271] Analysis of x-rays (e.g., hip radiographs) may be facilitated by development of techniques that locate one or more regions of interest (ROI) used for the calculation of the structural parameters of the trabecular bone. For example, the general position of the femur can be located using a binary image of the hip radiographs thresholded at the appropriate gray value. In a typical hip radiograph, the femur is a bright structure extending from the pelvis. (Figure 13). By thresholding the digitized radiograph at the typical femur intensity value, a binary image showing the femur is produced. The relatively thin structure of the femoral shaft can be extracted by applying a morphology operation on the binary image. The morphological top-hat filter (opening subtracted from input) with an upright rectangular structuring element segments the femoral shaft. The result is shown in Figure 13 with outline of the binarized femur superimposed on the original radiograph. The region is cropped for further processing, preferably leaving enough room to include the femoral head.
[0272] To position the set of predetermined ROI, a regularized active shape algorithm can be used (Behiels et al. (1999) Proceedings of the 2nd International Conference on Medical Image Computing and Computer-Assisted Intervention - MICCAI'99, Lecture notes in Computer Science 1679:128-137; Cootes (1994) Image and Vision Computing 12:355-366). A general model of the proximal femur is created by manually outlining the shape in a training set of typical hip radiographs to form a mean shape. The six predefined ROI are then embedded into this model. This mean model is scaled down 80%, isometrically along its centerline. This transformation is applied to the predefined ROI as well. The outline of the rescaled model is then used as the initial template and is positioned within the proximal femur in the input image. The control points of the contour are subsequently expanded outwards away from the nearest centerline point. The energy function to be optimized in this iterative process can take into account local features, such as gradient, intensity, deviation from the mean model, and curvature of contour segments. Figure 14 illustrates the propagation of the initial control points towards the femur edge. When the iteration is completed, a deformation field for the model area is calculated. This deformation field is interpolated for the model ROI inside the boundaries of the femur model. The result is a new set of ROI that is adapted to the input image, but similar to the model ROI with respect to anatomical landmarks (see Figure 9).
Example 6: Data Analysis
[0273] Patients are selected into one of three groups: healthy premenopausal
(PRE); healthy postmenopausal (POST), and osteoporotic postmenopausal (OSTEO) women. All groups are studied by: (1) dental x-ray images of the periapical and canine region; (2) quantitative computed tomography of the spine and (3) hip; (4) dual x-ray absorptiometry of the spine and (5) hip; (6) single x-ray absorptiometry of the calcaneus, and (7) ultrasound of the calcaneus using standard techniques. A diagnosis of osteoporosis is made when at least one atraumatic vertebral fracture as determined by a semi-quantitative assessment of morphologic changes of the thoracic and lumbar spine on lateral conventional x-rays is observed.
[0274] The means and standard deviations of the different bone structure measurements (see above) and bone mineral density measurements (mandibular BMD, QCT spine, QCT hip, DXA spine, DXA hip, SXA calcaneus, ultrasound calcaneus) are calculated for each patient group. The Student's t-test (t-values and p-values) and percent decrement are used for comparing the different measurements for reflecting intergroup differences. Annual, age-related changes are expressed as percent changes relative to the predicted values at age 30 and as fractional standard deviation (SD) of PRE. Correlations with age along with p-values are also be reported. Odds ratios (for 1SD change in the measured parameter) and 95% confidence limits based on the age- adjusted logistic regression are calculated to measure the discriminative ability (for discriminating between the postmenopausal osteoporotic and the normal postmenopausal group) and the risk of osteoporotic fracture associated with the measured parameter. The pairwise comparisons of the discriminative abilities are tested using age-adjusted receiver operating characteristic (ROC) curve analysis.
[0275] Pairwise comparisons of all techniques are obtained by pooling all subjects (PRE, POST, OSTEO) and using Pearson's correlation coefficients (r), percent standard errors of the estimate (CV), and p-values for testing significance of correlations.
[0276] To compare measurements for their diagnostic ability, a kappa score analysis is performed on the normal postmenopausal women (POST) and the osteoporotic postmenopausal women (OSTEO). This is done by classifying every woman from the postmenopausal groups as osteopenic if her T-score with respect to the reference group (PRE) is less (or in case of structural parameters also greater) than 2.5. The T-score for an individual woman and a particular measurement is defined as the measurement minus the mean measurement of young normals (PRE) divided by the SD of the measurement in the PRE group. Note that the T-score is measuring the position of an individual woman with respect to the PRE group and is different from the Student's t-value.
Example 7: Longitudinal Monitoring of Bone Structure [0277] Algorithms and software to match follow-up dental x-rays obtained at a time point T2 relative to baseline x-rays of the mandible obtained at an earlier time point Ti are developed. For purposes of monitoring of therapeutic response, bone structure parameters have to be measured at the same location of the mandible at different points in time. Thus, in order to compensate for differences in patient positioning and in order to find corresponding regions of interest (ROI's) for comparison of the results between baseline and follow-up examinations, it is desirable to register two dental x-ray images.
[0278] Due to possible slight differences in the projection angle of the x-ray beam on the film in the two images to be registered, an elastic matching step is preferably included. The first step, however, is a global affine transformation, for which the mutual information is used as a cost function. Wells et al. (1996) Medical Image Analysis 1:35- 51. The mutual information IM,N of two images M and N is defined as
Figure imgf000086_0001
[0279] Here, the gray values occurring in the two images are regarded as random variables, and the mutual information provides a measure of the strength of the dependence between these variables, PM and PN are the distributions of M and N respectively, and pMN is the joint distribution of M and N. Maintz et al. (1998) SPIE Medical Imaging - Image Processing. These distributions can be approximated from the marginal and joint gray value histograms, more accurately with the use of a Parzen window function. Powell's method can be used as an optimization scheme to find the best affine transformation for N to match it with M. Press et al. ("Numerical Recipes in C." 2nd edition, 1992, Cambridge University Press. [0280] This global transformation is followed by local elastic adjustments to improve the match. To achieve this, the conditional probability densities p(n|m) are estimated from the joint histogram of the globally registered images. The transformation vector field t(x) is then determined such that N(x-t(x)) is as similar to M(x) as possible by maximizing the local gray value correspondence, which for a fixed value of x is defined as
(t) = jw(*' - x)p(N(x' - 1) I M(x'))dx' .
[0281] Here, w is a window function whose width determines the size of the region that is used to compute t(x). To determine the window function, an approach similar to the one described in Warfield et al. "Brain Warping" 1999, Academic Press, p:67-84 is used. A number of successively wider window functions Wj are combined into a single window w = ^wlwl , where the weights Wj are given as
W, - x)VN(x')VN7 '(x')dx' .
Figure imgf000087_0001
[0282] The exact location of the ROI after automatic placement in the baseline image for a particular patient is kept in a database. When the patient returns for a follow-up exam, the new image is registered with the baseline image, and thus transformed into the coordinate system of the baseline image. The bone structure in the registered follow-up x-ray can then be measured at exactly the same position as in the baseline image. Example 8: Influence of Positioning of the Femur on Bone Structural Measurements
[0283] The effect(s) of the positioning of the femur on each parameter of the bone structure assessments is (are) examined. Hip x-rays are obtained in normal postmenopausal women and postmenopausal women with osteoporosis in neutral position and in various degrees of internal and external rotation. [0284] The diagnosis of osteoporosis is made when at least one atraumatic vertebral fracture as determined by a semi-quantitative assessment of morphologic changes of the thoracic and lumbar spine on lateral conventional radiographs is observed. See, also, Genant et al. (1993) J. Bone Miner Res. 8:1137-1148. [0285] Standard anteroposterior hip radiographs are obtained with the extremity at 30° internal rotation, 15° internal rotation, 0°, 15° external rotation, and 30° external rotation. These angles are achieved by placing the foot and ankle against a 30° or a 15° degree wedge in either internal or external rotation of the femur. The foot is secured against the wedge using Velcro straps. [0286] The effect of positioning is assessed by calculating the pair wise coefficient of variation (CV%) between the results for the 0° position and the other positions for each individual subjects. The angular dependency will be expressed for each of the angles 30° internal rotation, 15° internal rotation, 15° external rotation, and 30° external rotation as the root-mean-square of these CV% values over all subjects. In general, parameters with the least dependency on angular positioning of the femur are selected.
[0287] If the pair wise coefficient of variation between the results for the 0° neutral position and the 15° internal or external rotation position exceed 10% for the majority of the structural parameters measured, a foot holder that fixes the patients' foot in neutral position can be used The foot holder is designed with a base plate extending from the mid to distal thigh to the heel. The base plate preferably sits on the x-ray table. The patients' foot is positioned so that the posterior aspect of the heel is located on top of the base plate. The medial aspect of the foot is placed against a medial guide connected rigidly to the base plate at a 90° angle. A second, lateral guide attached to the base plate at a 90° angle with a sliding mechanism will then be moved toward the lateral aspect of the foot and will be locked in position as soon as it touches the lateral aspect of the foot. The foot will be secured to the medial and lateral guide using Velcro straps. It is expected that the degree of involuntary internal or external rotation can be limited to less than 5° using this approach.
Example 9: Influence of X-Ray Tube Angulation on Bone Structural Measurements [0288] The effect(s) of the positioning of the x-ray tube on each parameter of the bone structure assessments is (are) examined. Dental x-rays are obtained in normal postmenopausal women and postmenopausal women with osteoporosis. The diagnosis of osteoporosis is made when at least one atraumatic vertebral fracture as determined by a semi-quantitative assessment of morphologic changes of the thoracic and lumbar spine on lateral conventional radiographs is observed. See, also, Genant et al. (1993) J. Bone Miner Res. 8:1137-1148.
[0289] Standard anteroposterior dental radiographs are obtained in the incisor region of the mandible. The x-ray tube is aligned with an angle of 0°, 10°, 20°, 30°, and - 10°, - 20°, and -30° relative to the dental x-ray film. These angles are achieved with use of a goniometer applied to the metal tube located in front of the dental x-ray tube. The dental x-ray film is positioned at the posterior mandibular wall in the incisor region.
[0290] The effect of positioning is assessed by calculating the pair wise coefficient of variation (CV%) between the results for the 0° position and the other tube positions for each individual subject. The angular dependency will be expressed for each of the angles as the root-mean-square of these CV% values over all subjects.
[0291] The results indicate that a 10 degree tube angulation can result in a 12% error in apparent density.
[0292] A mechanical alignment system is then applied to the Rinn holder. For this purpose, an extension tubing is attached to the Rinn holder. The extension tubing is designed so that its inner diameter is slightly greater (and fits over) than the outer diameter of the dental x-ray system metal tube (Fig. 15). The dental x-ray system metal tube is then inserted into the extension tubing attached to the Rinn holder that reduces alignment error of the x-ray tube relative to the x-ray film. One group of patients then undergo two x-rays each of the incisor region. The results indicate that the short-term in-vivo reproducibility error of dental bone density and bone structure measurements is reduced with use of the mechanical alignment system by reducing x-ray tube angulation relative to the dental film and the anatomic landmarks in the mandible.
Example 10: Measurement of Bone Density, Bone Structure, Macro-Anatomical Parameters and Biomechanical Parameters and Selecting Therapy
[0293] An x-ray image of a mandible or a hip or spine or other bone is analyzed using a computer program capable of assessing bone density, bone structure, macro- anatomical parameters, or biomechanical parameters, for example as described above. The computer program derives a measurement of one or more bone density, bone structure, macro-anatomical or biomechanical parameters of the trabecular bone. The measurement of the parameter(s) is compared against a database containing information on said one or more parameters in normal, healthy age-, sex-, and race matched controls. If the patient's measurement differs by more than 2 standard deviations from the age-, sex-, and race matched mean of normal, healthy subjects, a report is sent to the physician who then selects a therapy based on the measurement(s).
Example 11 : Measurement of Bone Density, Bone Structure, Macro-Anatomical Parameters and Biomechanical Parameters and ϊlonitoring Therapy
[0294] One or more x-ray images (mandible, hip or spine or other bone) are obtained from a patient undergoing therapy for osteoporosis, for example using an anabolic or an antiresorptive drug at two different time points T1 and T2. The x-rays are analyzed using a computer program capable of assessing bone density, bone structure, macro-anatomical parameters, or biomechanical parameters. The computer program derives a measurement of one or more parameters of the bone for both time points T1 and T2. The measurement of the bone density, bone structure, macro-anatomical, or biomechanical parameter(s) at T1 and T2 is compared against a database containing information on said one or more parameters in normal, healthy age-, sex-, and race matched controls for each time point. If the results indicate that the patient has lost 5% or more bone between time points T1 and T2 despite therapy, a physician selects a different, more aggressive therapy. Example 12: Measurement of Macro-anatomical and/or biomechanical Parameters
[0295] A hip radiograph is obtained using standard techniques and including a calibration phantom as described herein. The reference orientation of hip x-rays is the average orientation of the femoral shaft.
A. Edge-detection [0296] A global gray level thresholding is performed using a bi-modal histogram segmentation algorithm on the hip x-ray generates a binary image proximal femur. Edge-detection of the hip x-ray can be used. Optionally, edge-detection methods are further refined by obtaining breaking edges detected into small segments and characterizing the orientation of each segment, thereby obtaining the outline of proximal femur. Each edge segment is then referenced to a map of expected proximal femur edge orientation and to a map of probability of edge location. Edge segments that do not conform to the expected orientation or are in low probability regions are removed. Morphology operations are applied onto the edge image to connect edge discontinuities. The edge image forms an enclosed boundary of the proximal femur. The region within the boundary is then combined with the binary image from global thresholding to form the final mask of the proximal femur.
[0297] Within a selected region of interest, edge detection is applied.
Morphology operations are applied to connect edge discontinuities. Segments are formed within enclosed edges. The area and major axis length of each segments are then measured. The regions are also superimposed on the original gray level image and the average gray level within each region is measured. The cortex is identified as the segments that are connected to the boundary of the proximal femur mask, that has the greatest area, longest major axis length and has a mean gray level above the average gray level of all enclosed segments within the proximal femur mask.
The segment identified as cortex is then skeletonized. The orientation of the cortex skeleton is verified to conform to the orientation map of proximal femur edge. Euclidian distance transform is applied to the binary image of the segment. The values of distance transform value along the skeleton are sampled and statistics (average, standard deviation, minimum, maximum and mod) measured.
[0298] As will be appreciated by those of skill in the art, measurements of macroanatomical parameters described here can be applied to hip, spine or knee radiographs with modifications to adapt to the shape, scale and location of macro-anatomical features specific to the anatomical region.

Claims

CLAIMS 1. A method to derive information regarding one or more bone parameters from an image comprising: (a) obtaining an image comprising bone from a subject; (b) defining two or more regions of interest (ROIs) in the image; and (c) analyzing a plurality of positions in the ROIs to determine one or more parameters selected from the group consisting of bone microarchitecture, bone macro- anatomy, biomechanical parameters and combinations thereof of the ROIs.
2. The method of claim 1 , wherein the ROIs are overlapping.
3. The method of claim 1 , wherein the positions analyzed in the ROIs are at regular intervals in the image.
4. The method of claim 1, wherein the positions analyzed in the ROIs are at irregular intervals in the image.
5. The method of claim 1, wherein the parameter is bone micro-architecture and the positions analyzed are at regular intervals.
6. The method of claim 1 , wherein the parameter is bone macro-anatomy and the positions are analyzed are at irregular intervals in the image.
7. The method of any of claims 1-6, wherein the image is two-dimensional.
8. The method of claim 7, wherein image is an x-ray image.
9. The method of any of claims 1-6, wherein the image is three-dimensional.
10. The method of any of the preceding claims wherein the image is an electronic image.
11. The method of claim 1 , wherein the subject is an osteoporosis subject.
12. A method of generating a map of one or more bone parameters, comprising (a) obtaining information on bone parameters according to the method of any of claims 1-11 ; and (b) identifying regions of the image that exhibit similar parameter characteristics, thereby creating a parameter map of the image.
13. A method of predicting a fracture path in a subject, comprising: (a) generating multiple parameter maps according to the method of claim 12; (b) generating a composite parameter map from the multiple parameters maps of step (a); and (c) analyzing the composite parameter map to identify possible fracture paths.
14. A method of predicting a fracture path in a subject comprising: (a) analyzing one or more parameter maps preparing according to the method of claim 12, wherein the analysis is watershed segmentation analysis or Markov random field analysis; and (c) identifying possible fracture paths based on the analysis of step (a), thereby predicting a fracture path in the subject.
15. A method of predicting the risk of fracture in a subject comprising: (a) generating a finite element model from one or more parameter maps obtained according to the method of claim 12; (b) applying simulated force vectors that would occur during a fracture incident to the finite element model generated in step(s); and (c) determining the minimum forces required for fracture to occur, thereby estimating the risk of fracture.
16. A method of determining the risk of fracture in a subject comprising: (a) predicting a fracture path according to the method of claim 13 or claim 14; (b) evaluating one or more selected bone parameters along the predicted fracture path, thereby estimating the risk of fracture.
17. A method of treating a subject with bone disease comprising (a) obtaining an image from a subject; (b) analyzing the image obtained in step (a) as described in any of claims 1-11; (c) diagnosing a bone disease based on the analysis of step (b); and (d) selecting and administering a suitable treatment to said subject based on said diagnosis.
PCT/US2004/009165 2003-03-25 2004-03-25 Methods for the compensation of imaging technique in the processing of radiographic images WO2004086972A2 (en)

Priority Applications (3)

Application Number Priority Date Filing Date Title
CA002519187A CA2519187A1 (en) 2003-03-25 2004-03-25 Methods for the compensation of imaging technique in the processing of radiographic images
EP04758337A EP1605824A2 (en) 2003-03-25 2004-03-25 Methods for the compensation of imaging technique in the processing of radiographic images
JP2006509289A JP2007524438A (en) 2003-03-25 2004-03-25 Compensation method in radiological image processing technology

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
US45759903P 2003-03-25 2003-03-25
US60/457,599 2003-03-25
US47845403P 2003-06-13 2003-06-13
US60/478,454 2003-06-13

Publications (3)

Publication Number Publication Date
WO2004086972A2 WO2004086972A2 (en) 2004-10-14
WO2004086972A9 true WO2004086972A9 (en) 2005-01-20
WO2004086972A3 WO2004086972A3 (en) 2005-04-28

Family

ID=33135059

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2004/009165 WO2004086972A2 (en) 2003-03-25 2004-03-25 Methods for the compensation of imaging technique in the processing of radiographic images

Country Status (5)

Country Link
US (6) US7664298B2 (en)
EP (1) EP1605824A2 (en)
JP (1) JP2007524438A (en)
CA (1) CA2519187A1 (en)
WO (1) WO2004086972A2 (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8913818B2 (en) 2000-10-11 2014-12-16 Imatx, Inc. Methods and devices for evaluating and treating a bone condition based on X-ray image analysis
US8939917B2 (en) 2009-02-13 2015-01-27 Imatx, Inc. Methods and devices for quantitative analysis of bone and cartilage
US8965075B2 (en) 2002-09-16 2015-02-24 Imatx, Inc. System and method for predicting future fractures
US8965087B2 (en) 2004-09-16 2015-02-24 Imatx, Inc. System and method of predicting future fractures
US9155501B2 (en) 2003-03-25 2015-10-13 Imatx, Inc. Methods for the compensation of imaging technique in the processing of radiographic images
US9267955B2 (en) 2001-05-25 2016-02-23 Imatx, Inc. Methods to diagnose treat and prevent bone loss
JP7187714B2 (en) 2019-04-12 2022-12-12 フラウンホーファー-ゲゼルシャフト・ツール・フェルデルング・デル・アンゲヴァンテン・フォルシュング・アインゲトラーゲネル・フェライン A method for calculating the deformation of an object in a time-resolved form and a computer program for the calculation

Families Citing this family (231)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090222103A1 (en) * 2001-05-25 2009-09-03 Conformis, Inc. Articular Implants Providing Lower Adjacent Cartilage Wear
US8771365B2 (en) * 2009-02-25 2014-07-08 Conformis, Inc. Patient-adapted and improved orthopedic implants, designs, and related tools
US8545569B2 (en) * 2001-05-25 2013-10-01 Conformis, Inc. Patient selectable knee arthroplasty devices
US8480754B2 (en) 2001-05-25 2013-07-09 Conformis, Inc. Patient-adapted and improved articular implants, designs and related guide tools
US8234097B2 (en) * 2001-05-25 2012-07-31 Conformis, Inc. Automated systems for manufacturing patient-specific orthopedic implants and instrumentation
US9603711B2 (en) * 2001-05-25 2017-03-28 Conformis, Inc. Patient-adapted and improved articular implants, designs and related guide tools
US20070100462A1 (en) * 2001-05-25 2007-05-03 Conformis, Inc Joint Arthroplasty Devices
US8882847B2 (en) * 2001-05-25 2014-11-11 Conformis, Inc. Patient selectable knee joint arthroplasty devices
US20070233269A1 (en) * 2001-05-25 2007-10-04 Conformis, Inc. Interpositional Joint Implant
US8556983B2 (en) 2001-05-25 2013-10-15 Conformis, Inc. Patient-adapted and improved orthopedic implants, designs and related tools
US8735773B2 (en) 2007-02-14 2014-05-27 Conformis, Inc. Implant device and method for manufacture
US20110071802A1 (en) * 2009-02-25 2011-03-24 Ray Bojarski Patient-adapted and improved articular implants, designs and related guide tools
US8617242B2 (en) 2001-05-25 2013-12-31 Conformis, Inc. Implant device and method for manufacture
US20110071645A1 (en) * 2009-02-25 2011-03-24 Ray Bojarski Patient-adapted and improved articular implants, designs and related guide tools
US10085839B2 (en) * 2004-01-05 2018-10-02 Conformis, Inc. Patient-specific and patient-engineered orthopedic implants
US9289153B2 (en) * 1998-09-14 2016-03-22 The Board Of Trustees Of The Leland Stanford Junior University Joint and cartilage diagnosis, assessment and modeling
JP2002532126A (en) 1998-09-14 2002-10-02 スタンフォード ユニバーシティ Joint condition evaluation and damage prevention device
US7239908B1 (en) 1998-09-14 2007-07-03 The Board Of Trustees Of The Leland Stanford Junior University Assessing the condition of a joint and devising treatment
US7467892B2 (en) 2000-08-29 2008-12-23 Imaging Therapeutics, Inc. Calibration devices and methods of use thereof
US6904123B2 (en) * 2000-08-29 2005-06-07 Imaging Therapeutics, Inc. Methods and devices for quantitative analysis of x-ray images
US20020186818A1 (en) * 2000-08-29 2002-12-12 Osteonet, Inc. System and method for building and manipulating a centralized measurement value database
US7050534B2 (en) * 2000-08-29 2006-05-23 Imaging Therapeutics, Inc. Methods and devices for quantitative analysis of x-ray images
WO2002017789A2 (en) * 2000-08-29 2002-03-07 Imaging Therapeutics Methods and devices for quantitative analysis of x-ray images
EP1322224B1 (en) 2000-09-14 2008-11-05 The Board Of Trustees Of The Leland Stanford Junior University Assessing condition of a joint and cartilage loss
WO2002022014A1 (en) 2000-09-14 2002-03-21 The Board Of Trustees Of The Leland Stanford Junior University Assessing the condition of a joint and devising treatment
US20070047794A1 (en) * 2000-10-11 2007-03-01 Philipp Lang Methods and devices for analysis of x-ray images
US7660453B2 (en) 2000-10-11 2010-02-09 Imaging Therapeutics, Inc. Methods and devices for analysis of x-ray images
US7477770B2 (en) * 2001-12-05 2009-01-13 The Trustees Of The University Of Pennsylvania Virtual bone biopsy
AU2002310193B8 (en) 2001-05-25 2007-05-17 Conformis, Inc. Methods and compositions for articular resurfacing
WO2004025541A1 (en) * 2002-09-16 2004-03-25 Imaging Therapeutics, Inc. Imaging markers in musculoskeletal disease
US7840247B2 (en) 2002-09-16 2010-11-23 Imatx, Inc. Methods of predicting musculoskeletal disease
ATE497740T1 (en) * 2002-10-07 2011-02-15 Conformis Inc MINIMALLY INVASIVE JOINT IMPLANT WITH A THREE-DIMENSIONAL GEOMETRY ADAPTED TO THE JOINT SURFACES
EP3075356B1 (en) * 2002-11-07 2023-07-05 ConforMIS, Inc. Method of selecting a meniscal implant
US7765319B1 (en) * 2003-07-30 2010-07-27 Gorman Sean P System and method for analyzing the structure of logical networks
US8290564B2 (en) * 2003-09-19 2012-10-16 Imatx, Inc. Method for bone structure prognosis and simulated bone remodeling
US8073521B2 (en) * 2003-09-19 2011-12-06 Imatx, Inc. Method for bone structure prognosis and simulated bone remodeling
GB0325523D0 (en) * 2003-10-31 2003-12-03 Univ Aberdeen Apparatus for predicting bone fracture risk
US7536044B2 (en) * 2003-11-19 2009-05-19 Siemens Medical Solutions Usa, Inc. System and method for detecting and matching anatomical structures using appearance and shape
US8064660B2 (en) * 2004-02-27 2011-11-22 National University Of Singapore Method and system for detection of bone fractures
KR100686289B1 (en) * 2004-04-01 2007-02-23 주식회사 메디슨 Apparatus and method for forming 3d ultrasound image using volume data in the contour of a target object image
JP5036534B2 (en) * 2004-04-26 2012-09-26 ヤンケレヴィッツ,デヴィット,エフ. Medical imaging system for precise measurement and evaluation of changes in target lesions
US7529195B2 (en) 2004-07-30 2009-05-05 Fortiusone, Inc. System and method of mapping and analyzing vulnerabilities in networks
US7283654B2 (en) * 2004-08-26 2007-10-16 Lumeniq, Inc. Dynamic contrast visualization (DCV)
US7684643B2 (en) * 2004-10-26 2010-03-23 Siemens Medical Solutions Usa, Inc. Mutual information regularized Bayesian framework for multiple image restoration
EP1831843A2 (en) 2004-12-21 2007-09-12 Koninklijke Philips Electronics N.V. Handling of datasets
US8055487B2 (en) * 2005-02-22 2011-11-08 Smith & Nephew, Inc. Interactive orthopaedic biomechanics system
US20070053491A1 (en) * 2005-09-07 2007-03-08 Eastman Kodak Company Adaptive radiation therapy method with target detection
US7471761B2 (en) * 2005-09-15 2008-12-30 Schick Technologies, Inc. System and method for computing oral bone mineral density with a panoramic x-ray system
DE102005047539A1 (en) * 2005-09-30 2007-04-05 Siemens Ag Medical diagnostic device e.g. digital flat-panel detector, windowing and/or dose control determining and setting method, involves selecting region of interest image data units of object area, and evaluating image data units
US7920730B2 (en) * 2005-10-07 2011-04-05 Siemens Medical Solutions Usa, Inc. Automatic bone detection in MRI images
US7697827B2 (en) 2005-10-17 2010-04-13 Konicek Jeffrey C User-friendlier interfaces for a camera
WO2007058918A2 (en) * 2005-11-11 2007-05-24 Hologic Inc. Estimating risk of future bone fracture utilizing three-dimensional bone density model
CN101384230A (en) * 2005-11-21 2009-03-11 福特真公司 Devices and methods for treating facet joints, uncovertebral joints, costovertebral joints and other joints
US8417010B1 (en) 2006-01-12 2013-04-09 Diagnoscan, LLC Digital x-ray diagnosis and evaluation of dental disease
US7986827B2 (en) * 2006-02-07 2011-07-26 Siemens Medical Solutions Usa, Inc. System and method for multiple instance learning for computer aided detection
US20070192301A1 (en) * 2006-02-15 2007-08-16 Encirq Corporation Systems and methods for indexing and searching data records based on distance metrics
FR2901043B1 (en) * 2006-05-12 2008-07-18 Rech S De L Ecole Nationale Su METHOD FOR MAKING A MODEL WITH FINISHED ELEMENTS OF A SINGLE BODY OF COMPLEX SHAPE AND STRUCTURE
JP2008009523A (en) * 2006-06-27 2008-01-17 Fujitsu Ltd Design support apparatus, design support method and design support program
US7572232B2 (en) * 2006-07-24 2009-08-11 Cardiac Pacemakers, Inc. Cardiac signal display and event detection using multiresolution Z-score transform
US20080077001A1 (en) * 2006-08-18 2008-03-27 Eastman Kodak Company Medical information system for intensive care unit
US20080119719A1 (en) * 2006-08-21 2008-05-22 The Regents Of The University Of California Templates for assessing bone quality and methods of use thereof
US9147272B2 (en) * 2006-09-08 2015-09-29 Christopher Allen Ingrassia Methods and systems for providing mapping, data management, and analysis
WO2008034101A2 (en) * 2006-09-15 2008-03-20 Imaging Therapeutics, Inc. Method and system for providing fracture/no fracture classification
US8090166B2 (en) * 2006-09-21 2012-01-03 Surgix Ltd. Medical image analysis
EP1935340B1 (en) * 2006-12-19 2017-11-01 Agfa HealthCare NV Method for neutralizing image artifacts prior to the determination of the Signal-to-noise ratio in CR/DR radiography systems
EP1946702B1 (en) * 2007-01-22 2012-03-07 BrainLAB AG Illustration of anatomic structure
AU2008216368A1 (en) * 2007-02-13 2008-08-21 Fortiusone, Inc. A method and system for integrating a social network and data repository to enable map creation
JP2008200114A (en) * 2007-02-16 2008-09-04 Konica Minolta Medical & Graphic Inc Bone trabecula evaluation system
KR100845474B1 (en) * 2007-03-13 2008-07-10 아시아나아이디티 주식회사 Tire built in rfid tag
US20100111395A1 (en) * 2007-04-12 2010-05-06 Konica Minolta Medical & Graphic, Inc. X-ray image analyzing system and program
JP5077345B2 (en) * 2007-04-12 2012-11-21 コニカミノルタエムジー株式会社 Bone disease evaluation system
JP5280647B2 (en) * 2007-05-29 2013-09-04 古野電気株式会社 Bone strength diagnostic device using ultrasound and method of operating bone strength diagnostic device using ultrasound
IL184151A0 (en) 2007-06-21 2007-10-31 Diagnostica Imaging Software Ltd X-ray measurement method
US7959742B2 (en) * 2007-07-11 2011-06-14 Whirlpool Corporation Outer support body for a drawer-type dishwasher
US20090067700A1 (en) * 2007-09-10 2009-03-12 Riverain Medical Group, Llc Presentation of computer-aided detection/diagnosis (CAD) results
JP2011500217A (en) * 2007-10-19 2011-01-06 ボストン サイエンティフィック サイムド,インコーポレイテッド Display of classifier output and reliability measure in image
US8065166B2 (en) * 2007-10-30 2011-11-22 Onemednet Corporation Methods, systems, and devices for managing medical images and records
US9171344B2 (en) 2007-10-30 2015-10-27 Onemednet Corporation Methods, systems, and devices for managing medical images and records
JP5416132B2 (en) * 2007-12-21 2014-02-12 スリーエム イノベイティブ プロパティズ カンパニー Orthodontic treatment monitoring based on reduced images
JP5312807B2 (en) * 2008-01-08 2013-10-09 オリンパス株式会社 Image processing apparatus and image processing program
US8189885B2 (en) * 2008-02-15 2012-05-29 The University Of Iowa Research Foundation Apparatus and method for computing regional statistical distribution over a mean anatomic space
WO2009111626A2 (en) 2008-03-05 2009-09-11 Conformis, Inc. Implants for altering wear patterns of articular surfaces
WO2009111656A1 (en) * 2008-03-05 2009-09-11 Conformis, Inc. Edge-matched articular implant
JP5353876B2 (en) * 2008-03-06 2013-11-27 コニカミノルタ株式会社 Image processing device
US8852128B2 (en) * 2008-03-12 2014-10-07 University Of Cincinnati Computer system and method for assessing dynamic bone quality
JP2011519713A (en) * 2008-05-12 2011-07-14 コンフォーミス・インコーポレイテッド Devices and methods for treatment of facet joints and other joints
ITMI20080967A1 (en) * 2008-05-23 2009-11-24 Salvatore Longoni METHOD OF DESIGN AND / OR SELECTION OF A PLANT AND / OR IMPLANTABLE MATERIAL IN FABRICS OF THE HUMAN OR ANIMAL BODY AND DEVICE AND / OR MATERIAL SO IT IS OBTAINED
US9109998B2 (en) * 2008-06-18 2015-08-18 Orthopedic Navigation Ltd. Method and system for stitching multiple images into a panoramic image
US8811696B2 (en) * 2008-08-12 2014-08-19 Wyeth Pharmaceuticals, Inc. Morphometry of the human hip joint and prediction of osteoarthritis
US8582843B2 (en) 2008-08-12 2013-11-12 Wyeth Pharmaceuticals, Inc. Morphometry of the human knee joint and prediction for osteoarthritis
EP2328476B1 (en) * 2008-09-19 2014-04-16 Duke University Systems and methods for generating an osteoarthritis progression predictor and systems and methods for using the predictor
KR101009782B1 (en) * 2008-10-28 2011-01-19 (주)메디슨 Ultrasound system and method providing wide image mode
US8649577B1 (en) * 2008-11-30 2014-02-11 Image Analysis, Inc. Automatic method and system for measurements of bone density and structure of the hip from 3-D X-ray imaging devices
KR101087137B1 (en) * 2008-12-04 2011-11-25 한국전자통신연구원 Method and support device for jawbone mineral density measurement
JP5373470B2 (en) * 2009-04-28 2013-12-18 ジーイー・メディカル・システムズ・グローバル・テクノロジー・カンパニー・エルエルシー Modeling apparatus, magnetic resonance imaging apparatus, modeling method, and program
WO2010126797A1 (en) 2009-04-29 2010-11-04 Onemednet Corporation Methods, systems, and devices for managing medical images and records
US8608481B2 (en) 2009-05-13 2013-12-17 Medtronic Navigation, Inc. Method and apparatus for identifying an instrument location based on measuring a characteristic
KR101069026B1 (en) * 2009-05-22 2011-09-29 경희대학교 산학협력단 Image Transform System For Analysis Of Dental Caries
DE102009022834A1 (en) * 2009-05-27 2010-12-09 Siemens Aktiengesellschaft Method for automatic analysis of image data of a structure
GB0910316D0 (en) 2009-06-16 2009-07-29 Univ Manchester Image analysis method
US8768016B2 (en) * 2009-06-19 2014-07-01 Carestream Health, Inc. Method for quantifying caries
AU2010292991B2 (en) * 2009-09-11 2015-03-26 Curvebeam Ai Limited Method and system for image analysis of selected tissue structures
US20110081054A1 (en) * 2009-10-02 2011-04-07 Harris Corporation Medical image analysis system for displaying anatomical images subject to deformation and related methods
US20110081055A1 (en) * 2009-10-02 2011-04-07 Harris Corporation, Corporation Of The State Of Delaware Medical image analysis system using n-way belief propagation for anatomical images subject to deformation and related methods
US20110081061A1 (en) * 2009-10-02 2011-04-07 Harris Corporation Medical image analysis system for anatomical images subject to deformation and related methods
GB0917524D0 (en) 2009-10-07 2009-11-25 Cambridge Entpr Ltd Image data processing systems
US8687859B2 (en) 2009-10-14 2014-04-01 Carestream Health, Inc. Method for identifying a tooth region
US9235901B2 (en) * 2009-10-14 2016-01-12 Carestream Health, Inc. Method for locating an interproximal tooth region
US8908936B2 (en) * 2009-10-14 2014-12-09 Carestream Health, Inc. Method for extracting a carious lesion area
KR101107513B1 (en) * 2009-10-14 2012-01-31 (주) 케이앤아이테크놀로지 Method of calibrating a dual x-ray imaging system and method of obtaining a three-dimensional position of a postoperative articulation using the calibrating method
US8644608B2 (en) * 2009-10-30 2014-02-04 Eiffel Medtech Inc. Bone imagery segmentation method and apparatus
US9014440B2 (en) * 2009-11-11 2015-04-21 Thiagarajar College Of Engineering Dental cysts detector
US20110110575A1 (en) * 2009-11-11 2011-05-12 Thiagarajar College Of Engineering Dental caries detector
JP5538862B2 (en) * 2009-12-18 2014-07-02 キヤノン株式会社 Image processing apparatus, image processing system, image processing method, and program
US9058665B2 (en) * 2009-12-30 2015-06-16 General Electric Company Systems and methods for identifying bone marrow in medical images
JP5671059B2 (en) * 2010-01-12 2015-02-18 クロメック リミテッド Data set calibration
EP2544583B1 (en) 2010-03-08 2016-03-02 Bruce Adams System, method and article for normalization and enhancement of tissue images
US9865050B2 (en) 2010-03-23 2018-01-09 Hologic, Inc. Measuring intramuscular fat
JP6077993B2 (en) 2010-04-30 2017-02-08 アイキャド インクiCAD, INC. Image data processing method, system, and program for identifying image variants
US8675933B2 (en) 2010-04-30 2014-03-18 Vucomp, Inc. Breast segmentation in radiographic images
JP5878117B2 (en) 2010-05-11 2016-03-08 株式会社テレシステムズ Radiation imaging apparatus and phantom used in the apparatus
FR2960762B1 (en) * 2010-06-07 2013-04-05 Designers Developers Distributors Associates D3A Medical Systems METHODS AND SYSTEMS FOR IMAGING AND CHARACTERIZING BONE TISSUE
GB201009725D0 (en) 2010-06-11 2010-07-21 Univ Leuven Kath Method of quantifying local bone loss
US9256799B2 (en) 2010-07-07 2016-02-09 Vucomp, Inc. Marking system for computer-aided detection of breast abnormalities
KR101819257B1 (en) 2010-07-13 2018-01-16 다카라 텔레시스템즈 가부시키가이샤 X-ray tomogram imaging device
US8620045B2 (en) * 2010-10-15 2013-12-31 Bruce William Adams System , method and article for measuring and reporting craniomandibular biomechanical functions
US9299001B2 (en) * 2010-10-29 2016-03-29 Analogic Corporation Object identification using sparse spectral components
US20120115107A1 (en) * 2010-11-04 2012-05-10 Adams Bruce W System and method for automated manufacturing of dental orthotics
EP2637570A4 (en) * 2010-11-10 2014-07-02 Echometrix Llc System and method of ultrasound image processing
EP2754419B1 (en) 2011-02-15 2024-02-07 ConforMIS, Inc. Patient-adapted and improved orthopedic implants
US9235892B2 (en) 2011-03-31 2016-01-12 Denise De Andrade Castro Method and device for comparing radiographic images
WO2012165991A1 (en) * 2011-05-31 2012-12-06 Schlumberger Holdings Limited Method for determination of spatial distribution and concentration of contrast components in a porous and/or heterogeneous sample
US9351662B2 (en) 2011-06-17 2016-05-31 Microsoft Technology Licensing, Llc MRI scanner that outputs bone strength indicators
RU2467315C1 (en) 2011-06-23 2012-11-20 Шлюмберже Текнолоджи Б.В. Method to detect spatial distribution and concentration of clay in core sample
RU2467316C1 (en) 2011-06-23 2012-11-20 Шлюмберже Текнолоджи Б.В. Method to detect spatial distribution and concentration of component in pore space of porous material
JP5923889B2 (en) * 2011-07-29 2016-05-25 株式会社島津製作所 Trabecular bone analyzer
KR20140124875A (en) 2011-08-15 2014-10-27 에픽 리서치 앤드 다이어그노스틱스 인코포레이티드 Localized physiologic status from luminosity around fingertip or toe
CN103827874B (en) * 2011-09-26 2017-02-22 皇家飞利浦有限公司 Medical image system and method
US8534115B2 (en) * 2011-10-17 2013-09-17 Schlumberger Technology Corporation Systems and methods of determining parameter values in a downhole environment
US20140376685A1 (en) 2011-10-18 2014-12-25 Schlumberger Technology Corporation Method for 3d mineral mapping of a rock sample
RU2482465C1 (en) 2011-11-29 2013-05-20 Шлюмберже Текнолоджи Б.В. Frozen rock specimen investigation method
RU2486495C1 (en) 2011-12-20 2013-06-27 Шлюмберже Текнолоджи Б.В. Method to examine samples of non-consolidated porous media
JP5829921B2 (en) * 2012-01-06 2015-12-09 学校法人 龍谷大学 Computer operating method for fracture risk assessment
US9245069B2 (en) * 2012-02-03 2016-01-26 The Regents Of The University Of California Methods for calculating bone fracture load
US9361672B2 (en) * 2012-03-26 2016-06-07 Google Technology Holdings LLC Image blur detection
US9044186B2 (en) 2012-06-25 2015-06-02 George W. Ma Portable dual-energy radiographic X-ray perihpheral bone density and imaging systems and methods
WO2014052782A1 (en) 2012-09-28 2014-04-03 Children's Medical Center Corporation Diffusion-weighted mri using multiple b-values and constant echo time
EP2956046B1 (en) 2013-02-18 2018-05-09 OrthoGrid Systems S.a.r.l. Computer system to digitally quantify alignment and placement parameters in a musculoskeletal application
JP6030002B2 (en) * 2013-02-27 2016-11-24 富士フイルムRiファーマ株式会社 Image processing program, image processing apparatus, and image processing method
US9642560B2 (en) * 2013-04-03 2017-05-09 Brainlab Ag Method and device for determining the orientation of a co-ordinate system of an anatomical object in a global co-ordinate system
EP2789308B1 (en) 2013-04-12 2018-01-31 CADFEM GmbH Computer-implemented technique for generating a data set that geometrically defines a bone cut configuration
US9390502B2 (en) 2013-04-22 2016-07-12 Kabushiki Kaisha Toshiba Positioning anatomical landmarks in volume data sets
JP5680703B2 (en) * 2013-05-08 2015-03-04 株式会社日立メディコ Ultrasonic diagnostic equipment
US9466012B2 (en) 2013-07-11 2016-10-11 Radiological Imaging Technology, Inc. Phantom image classification
JP6345178B2 (en) * 2013-07-23 2018-06-20 富士フイルム株式会社 Radiation image processing apparatus and method
US11850061B2 (en) * 2013-08-09 2023-12-26 O.N.Diagnostics, LLC Clinical assessment of fragile bone strength
US9848818B1 (en) * 2013-08-09 2017-12-26 O.N.Diagnostics, LLC Clinical assessment of fragile bone strength
EP3035891B1 (en) 2013-08-21 2020-05-27 Laboratoires Bodycad Inc. Anatomically adapted orthopedic implant
CA2919717C (en) 2013-08-21 2021-06-22 Laboratoires Bodycad Inc. Bone resection guide and method
US10166109B2 (en) 2013-09-18 2019-01-01 Stryker Corporation Patient specific bone preparation for consistent effective fixation feature engagement
DE102013218819B3 (en) * 2013-09-19 2014-10-30 Siemens Aktiengesellschaft Method of reducing artifacts in an image data set and X-ray device
WO2015042416A1 (en) * 2013-09-20 2015-03-26 Children's Medical Center Corporation Methods and apparatus for modeling diffusion-weighted mr data acquired at multiple non-zero b-values
US10006271B2 (en) * 2013-09-26 2018-06-26 Harris Corporation Method for hydrocarbon recovery with a fractal pattern and related apparatus
US9937011B2 (en) * 2013-10-09 2018-04-10 Persimio Ltd Automated patient-specific method for biomechanical analysis of bone
JP6107609B2 (en) * 2013-11-08 2017-04-05 株式会社島津製作所 Trabecular bone analyzer
US9642585B2 (en) 2013-11-25 2017-05-09 Hologic, Inc. Bone densitometer
AT515149B1 (en) * 2013-11-29 2017-01-15 Braincon Handels-Gmbh Method for the early detection of bone joint diseases
JP6107627B2 (en) * 2013-12-03 2017-04-05 株式会社島津製作所 Trabecular bone analyzer
JP6252750B2 (en) * 2013-12-03 2017-12-27 株式会社島津製作所 Trabecular bone analyzer
EP3077990A1 (en) 2013-12-06 2016-10-12 Koninklijke Philips N.V. Bone segmentation from image data
EP3098594B1 (en) 2014-01-23 2021-12-22 Job Corporation X-ray inspection apparatus and x-ray inspection method
US9418415B2 (en) 2014-02-05 2016-08-16 Shimadzu Corporation Trabecular bone analyzer
JP6345468B2 (en) * 2014-04-09 2018-06-20 キヤノンメディカルシステムズ株式会社 Medical diagnostic imaging equipment
JP6133231B2 (en) 2014-04-18 2017-05-24 株式会社日立製作所 X-ray energy spectrum measuring method, X-ray energy spectrum measuring apparatus and X-ray CT apparatus
US10180483B2 (en) 2014-04-24 2019-01-15 David W Holdsworth 3D printed physical phantom utilized in measuring 3D geometric distortion occurring in MRI and CT images
US9510757B2 (en) 2014-05-07 2016-12-06 Align Technology, Inc. Identification of areas of interest during intraoral scans
ES2707100T3 (en) * 2014-05-22 2019-04-02 Nestec Sa Helical movement device
US10039513B2 (en) 2014-07-21 2018-08-07 Zebra Medical Vision Ltd. Systems and methods for emulating DEXA scores based on CT images
US10588589B2 (en) 2014-07-21 2020-03-17 Zebra Medical Vision Ltd. Systems and methods for prediction of osteoporotic fracture risk
JP6299504B2 (en) * 2014-07-23 2018-03-28 株式会社島津製作所 Image analysis device
FR3026211B1 (en) 2014-09-19 2017-12-08 Univ Aix Marseille METHOD FOR IDENTIFYING THE ANISOTROPY OF THE TEXTURE OF A DIGITAL IMAGE
KR101642425B1 (en) * 2014-10-28 2016-07-25 삼성전자주식회사 A radiographic imaging apparatus and a method of controlling the radiographic imaging
US9964499B2 (en) * 2014-11-04 2018-05-08 Toshiba Medical Systems Corporation Method of, and apparatus for, material classification in multi-energy image data
US20180020999A1 (en) * 2015-02-13 2018-01-25 Shimadzu Corporation Bone analyzing device
US10499865B2 (en) 2015-02-26 2019-12-10 Hologic, Inc. Methods for physiological state determination in body scans
US20160331339A1 (en) * 2015-05-15 2016-11-17 The Trustees Of Columbia University In The City Of New York Systems And Methods For Early Detection And Monitoring Of Osteoarthritis
US10706626B1 (en) * 2015-12-09 2020-07-07 Roger Brent Augmented reality procedural system
US9715721B2 (en) 2015-12-18 2017-07-25 Sony Corporation Focus detection
US10201320B2 (en) 2015-12-18 2019-02-12 OrthoGrid Systems, Inc Deformed grid based intra-operative system and method of use
US10052170B2 (en) 2015-12-18 2018-08-21 MediLux Capitol Holdings, S.A.R.L. Mixed reality imaging system, apparatus and surgical suite
US11386556B2 (en) 2015-12-18 2022-07-12 Orthogrid Systems Holdings, Llc Deformed grid based intra-operative system and method of use
KR102471937B1 (en) * 2016-02-05 2022-11-29 (주)바텍이우홀딩스 Method for Determining the Likelihood of Dental Caries Using an X-ray Image
JP6321703B2 (en) * 2016-03-04 2018-05-09 ファナック株式会社 Wire electrical discharge machine inspection system
JP6810395B2 (en) * 2016-03-18 2021-01-06 メディア株式会社 Osteoporosis diagnosis support device
US10136869B2 (en) 2016-03-25 2018-11-27 Perkinelmer Health Sciences, Inc. Systems and methods for characterizing a central axis of a bone from a 3D anatomical image
EP3440660A1 (en) * 2016-04-06 2019-02-13 Koninklijke Philips N.V. Method, device and system for enabling to analyze a property of a vital sign detector
CA3020362A1 (en) 2016-04-07 2017-10-12 Icahn School Of Medicine At Mount Sinai Apparatus, method and system for providing customizable bone implants
US10043088B2 (en) * 2016-06-23 2018-08-07 Siemens Healthcare Gmbh Image quality score using a deep generative machine-learning model
US11013490B2 (en) 2016-11-15 2021-05-25 Musclesound, Inc. Non-invasive determination of muscle tissue size
USD808524S1 (en) 2016-11-29 2018-01-23 Laboratoires Bodycad Inc. Femoral implant
US11064971B2 (en) 2016-11-30 2021-07-20 Musclesound, Inc. Non-Invasive determination of muscle tissue quality and intramuscular fat
US9987502B1 (en) * 2016-12-06 2018-06-05 International Business Machines Corporation Radiation therapy treatment planning using regression
WO2018131611A1 (en) * 2017-01-10 2018-07-19 メディア株式会社 Bone density measurement device, bone density measurement system, and imaging assisting tool
US11096658B2 (en) 2017-02-02 2021-08-24 Musclesound, Inc. Non-invasive determination of pennation angle and/or fascicle length
US11160493B2 (en) 2017-03-03 2021-11-02 Musclesound, Inc. System and method for determining a subject's muscle fuel level, muscle fuel rating, and muscle energy status
KR102021942B1 (en) * 2017-09-18 2019-09-17 한국생산기술연구원 Fracture line calculation method and fracture line calculation system
EP3467771A1 (en) * 2017-10-05 2019-04-10 Koninklijke Philips N.V. Image feature annotation in diagnostic imaging
US10963750B2 (en) 2018-01-04 2021-03-30 IAS Machine, LLC Procedural language and content generation environment for use in augmented reality/mixed reality systems to support laboratory and related operations
US11182920B2 (en) 2018-04-26 2021-11-23 Jerry NAM Automated determination of muscle mass from images
JP6906479B2 (en) * 2018-05-25 2021-07-21 富士フイルム株式会社 Bone mineral information acquisition device, method and program
WO2020006097A1 (en) * 2018-06-26 2020-01-02 Massachusetts Institute Of Technology Systems and methods for imaging cortical bone and soft tissue
WO2020033656A1 (en) * 2018-08-08 2020-02-13 Loyola University Chicago Methods of classifying and/or determining orientations of objects using two-dimensional images
EP3614337A1 (en) * 2018-08-19 2020-02-26 Chang Gung Memorial Hospital, Linkou Method and system of analyzing medical images
WO2020054738A1 (en) * 2018-09-10 2020-03-19 京セラ株式会社 Estimation device, estimation system, and estimation program
EP3852645A4 (en) 2018-09-12 2022-08-24 Orthogrid Systems, SAS An artificial intelligence intra-operative surgical guidance system and method of use
US11540794B2 (en) 2018-09-12 2023-01-03 Orthogrid Systesm Holdings, LLC Artificial intelligence intra-operative surgical guidance system and method of use
US10991091B2 (en) 2018-10-30 2021-04-27 Diagnocat Inc. System and method for an automated parsing pipeline for anatomical localization and condition classification
USD910652S1 (en) 2019-01-31 2021-02-16 OrthoGrid Systems, Inc Display screen or portion thereof with a graphical user interface
US11456063B2 (en) 2019-09-08 2022-09-27 Novadontics Llc Dental patient management system
ES2828728A1 (en) * 2019-11-27 2021-05-27 Fundacion Para La Investigacion Del Hospital Univ La Fe De La Comunidad Valenciana METHOD TO OBTAIN AN IMAGE BIOMARKER THAT QUANTIFIES THE QUALITY OF THE TRABECULAR STRUCTURE OF THE BONES (Machine-translation by Google Translate, not legally binding)
US11132590B2 (en) 2019-12-12 2021-09-28 Lablightar Inc. Augmented camera for improved spatial localization and spatial orientation determination
US11452492B2 (en) 2020-04-21 2022-09-27 Mazor Robotics Ltd. System and method for positioning an imaging device
US11300695B2 (en) 2020-04-24 2022-04-12 Ronald Nutt Time-resolved positron emission tomography encoder system for producing event-by-event, real-time, high resolution, three-dimensional positron emission tomographic image without the necessity of performing image reconstruction
US11054534B1 (en) 2020-04-24 2021-07-06 Ronald Nutt Time-resolved positron emission tomography encoder system for producing real-time, high resolution, three dimensional positron emission tomographic image without the necessity of performing image reconstruction
CN111724357B (en) * 2020-06-09 2023-05-16 四川大学 Arm bone density measurement method based on digital radiological image and support vector regression
KR102378742B1 (en) 2020-07-30 2022-03-28 한국과학기술연구원 System and method for supporting user to read x-ray image
CN111967540B (en) * 2020-09-29 2021-06-08 北京大学口腔医学院 Maxillofacial fracture identification method and device based on CT database and terminal equipment
USD979578S1 (en) 2021-02-08 2023-02-28 Orthogrid Systems Holdings, Llc Display screen or portion thereof with a graphical user interface
US11701075B2 (en) 2021-02-10 2023-07-18 GE Precision Healthcare LLC X-ray image feedback for DXA scan FOV adjustment
IT202100004376A1 (en) * 2021-02-25 2022-08-25 Esaote Spa METHOD OF DETERMINING SCAN PLANS IN THE ACQUISITION OF ULTRASOUND IMAGES AND ULTRASOUND SYSTEM FOR IMPLEMENTING THE SAID METHOD
EP4312729A1 (en) * 2021-04-02 2024-02-07 Spintech, Inc. Systems and methods for template-based automatic detection of anatomical structures

Family Cites Families (193)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US2274808A (en) 1941-01-07 1942-03-03 Irwin C Rinn Bite wing for dental film packs and the like
US7366676B2 (en) 2001-05-29 2008-04-29 Mevis Breastcare Gmbh & Co. Kg Method and system for in-service monitoring and training for a radiologic workstation
DE2042009C3 (en) 1970-08-25 1975-02-27 Siemens Ag, 1000 Berlin U. 8000 Muenchen Arrangement for the non-destructive density measurement of substances of living objects by means of penetrating rays
US4012638A (en) 1976-03-09 1977-03-15 Altschuler Bruce R Dental X-ray alignment system
US4126789A (en) 1977-06-06 1978-11-21 Vogl Thomas M X-ray phantom
US4298800A (en) 1978-02-27 1981-11-03 Computome Corporation Tomographic apparatus and method for obtaining three-dimensional information by radiation scanning
GB2023920A (en) 1978-06-19 1980-01-03 Thoro Ray Inc Dental X-ray apparatus
US4686695A (en) 1979-02-05 1987-08-11 Board Of Trustees Of The Leland Stanford Junior University Scanned x-ray selective imaging system
US4233507A (en) 1979-05-07 1980-11-11 General Electric Company Computer tomography table containing calibration and correlation samples
US4251732A (en) 1979-08-20 1981-02-17 Fried Alan J Dental x-ray film holders
US4356400A (en) 1980-08-04 1982-10-26 General Electric Company X-Ray apparatus alignment method and device
US4400827A (en) 1981-11-13 1983-08-23 Spears James R Method and apparatus for calibrating rapid sequence radiography
FR2547495B1 (en) 1983-06-16 1986-10-24 Mouyen Francis APPARATUS FOR OBTAINING A DENTAL RADIOLOGICAL IMAGE
US4649561A (en) 1983-11-28 1987-03-10 Ben Arnold Test phantom and method of use of same
JPS61109557A (en) 1984-11-02 1986-05-28 帝人株式会社 Evaluation of bone
JPH07102210B2 (en) 1986-05-14 1995-11-08 帝人株式会社 Evaluation method of bone atrophy of alveolar bone
US4782502A (en) 1986-10-01 1988-11-01 Schulz Eloy E Flexible calibration phantom for computer tomography system
US4985906A (en) 1987-02-17 1991-01-15 Arnold Ben A Calibration phantom for computer tomography system
CA1288176C (en) 1987-10-29 1991-08-27 David C. Hatcher Method and apparatus for improving the alignment of radiographic images
US4922915A (en) 1987-11-27 1990-05-08 Ben A. Arnold Automated image detail localization method
US5127032A (en) 1987-12-03 1992-06-30 Johns Hopkins University Multi-directional x-ray imager
US4956859A (en) 1989-03-10 1990-09-11 Expert Image Systems, Inc. Source filter for radiographic imaging
US5090040A (en) 1989-03-10 1992-02-18 Expert Image Systems, Inc. Data acquisition system for radiographic imaging
US5001738A (en) 1989-04-07 1991-03-19 Brooks Jack D Dental X-ray film holding tab and alignment method
FR2649883B1 (en) 1989-07-20 1991-10-11 Gen Electric Cgr METHOD FOR CORRECTING THE MEASUREMENT OF BONE DENSITY IN A SCANNER
US5537483A (en) 1989-10-03 1996-07-16 Staplevision, Inc. Automated quality assurance image processing system
US6031892A (en) 1989-12-05 2000-02-29 University Of Massachusetts Medical Center System for quantitative radiographic imaging
US5150394A (en) 1989-12-05 1992-09-22 University Of Massachusetts Medical School Dual-energy system for quantitative radiographic imaging
US5864146A (en) 1996-11-13 1999-01-26 University Of Massachusetts Medical Center System for quantitative radiographic imaging
US5562448A (en) 1990-04-10 1996-10-08 Mushabac; David R. Method for facilitating dental diagnosis and treatment
US5122664A (en) 1990-04-27 1992-06-16 Fuji Photo Film Co., Ltd. Method and apparatus for quantitatively analyzing bone calcium
US5228445A (en) 1990-06-18 1993-07-20 Board Of Regents, The University Of Texas System Demonstration by in vivo measurement of reflection ultrasound analysis of improved bone quality following slow-release fluoride treatment in osteoporosis patients
US5172695A (en) 1990-09-10 1992-12-22 Cann Christopher E Method for improved prediction of bone fracture risk using bone mineral density in structural analysis
DE9016046U1 (en) 1990-11-26 1991-02-14 Kalender, Willi, Dr., 8521 Kleinseebach, De
US5577089A (en) 1991-02-13 1996-11-19 Lunar Corporation Device and method for analysis of bone morphology
US5533084A (en) * 1991-02-13 1996-07-02 Lunar Corporation Bone densitometer with improved vertebral characterization
JP2641078B2 (en) 1991-03-28 1997-08-13 富士写真フイルム株式会社 Bone mineral analysis
US5200993A (en) 1991-05-10 1993-04-06 Bell Atlantic Network Services, Inc. Public telephone network including a distributed imaging system
US5270651A (en) 1991-05-21 1993-12-14 The Trustees Of The University Of Pennsylvania Method and apparatus for diagnosing osteoporosis
US5247934A (en) 1991-08-09 1993-09-28 Trustees Of The University Of Pennsylvania Method and apparatus for diagnosing osteoporosis with MR imaging
JP2973643B2 (en) 1991-10-04 1999-11-08 松下電器産業株式会社 Quantitative measurement of substances
US5271401A (en) 1992-01-15 1993-12-21 Praxair Technology, Inc. Radiological imaging method
EP0570936B1 (en) 1992-05-20 2000-08-09 Aloka Co. Ltd. Bone assessment apparatus
WO1993024055A1 (en) 1992-05-29 1993-12-09 Ge Yokogawa Medical Systems, Ltd. Method of quantitative determination of bone salt with ct equipment
US5321520A (en) 1992-07-20 1994-06-14 Automated Medical Access Corporation Automated high definition/resolution image storage, retrieval and transmission system
US5281232A (en) 1992-10-13 1994-01-25 Board Of Regents Of The University Of Arizona/ University Of Arizona Reference frame for stereotactic radiosurgery using skeletal fixation
US5320102A (en) 1992-11-18 1994-06-14 Ciba-Geigy Corporation Method for diagnosing proteoglycan deficiency in cartilage based on magnetic resonance image (MRI)
US5335260A (en) 1992-11-25 1994-08-02 Arnold Ben A Calibration phantom and improved method of quantifying calcium and bone density using same
US5592943A (en) 1993-04-07 1997-01-14 Osteo Sciences Corporation Apparatus and method for acoustic analysis of bone using optimized functions of spectral and temporal signal components
US5513240A (en) 1993-05-18 1996-04-30 The Research Foundation Of Suny Intraoral radiograph alignment device
FR2705785B1 (en) 1993-05-28 1995-08-25 Schlumberger Ind Sa Method for determining the attenuation function of an object with respect to the transmission of a reference thickness of a reference material and device for implementing the method.
US5657369A (en) 1993-11-22 1997-08-12 Hologic, Inc. X-ray bone densitometry system having forearm positioning assembly
US5432834A (en) * 1993-11-22 1995-07-11 Hologic, Inc. Whole-body dual-energy bone densitometry using a narrow angle fan beam to cover the entire body in successive scans
US5931780A (en) 1993-11-29 1999-08-03 Arch Development Corporation Method and system for the computerized radiographic analysis of bone
EP0660599B2 (en) 1993-12-24 2002-08-14 Agfa-Gevaert Partially-transparent-shield-method for scattered radiation compensation in x-ray imaging
US5948692A (en) 1994-02-19 1999-09-07 Seikagaku Corporation Method and measurement kit for assay of normal agrecan, and method for evaluation of informations on the joint
GB9506050D0 (en) 1995-03-24 1995-05-10 Osteometer A S Assaying collagen fragments in body fluids
US5600574A (en) 1994-05-13 1997-02-04 Minnesota Mining And Manufacturing Company Automated image quality control
US5476865A (en) 1994-07-06 1995-12-19 Eli Lilly And Company Methods of inhibiting bone loss
EP0905638A1 (en) 1994-08-29 1999-03-31 Torsana A/S A method of estimation
US5493593A (en) 1994-09-27 1996-02-20 University Of Delaware Tilted detector microscopy in computerized tomography
WO1996012187A1 (en) 1994-10-13 1996-04-25 Horus Therapeutics, Inc. Computer assisted methods for diagnosing diseases
JPH08186762A (en) 1994-12-27 1996-07-16 Toshiba Medical Eng Co Ltd Mammography device
SE9601065L (en) 1996-03-20 1997-03-03 Siemens Elema Ab Anesthesia System
US5594775A (en) * 1995-04-19 1997-01-14 Wright State University Method and apparatus for the evaluation of cortical bone by computer tomography
US5886353A (en) 1995-04-21 1999-03-23 Thermotrex Corporation Imaging device
US5565678A (en) 1995-06-06 1996-10-15 Lumisys, Inc. Radiographic image quality assessment utilizing a stepped calibration target
US6038287A (en) 1995-10-10 2000-03-14 Miles; Dale A. Portable X-ray device
JPH11501190A (en) 1995-12-22 1999-01-26 フィリップス エレクトロニクス エヌ ベー X-ray inspection device including subtraction unit
US5772592A (en) 1996-01-08 1998-06-30 Cheng; Shu Lin Method for diagnosing and monitoring osteoporosis
US6215846B1 (en) 1996-02-21 2001-04-10 Lunar Corporation Densitometry adapter for compact x-ray fluoroscopy machine
US5785041A (en) 1996-03-26 1998-07-28 Hologic Inc. System for assessing bone characteristics
AU2763597A (en) 1996-05-06 1997-11-26 Torsana A/S A method of estimating skeletal status
US6108635A (en) 1996-05-22 2000-08-22 Interleukin Genetics, Inc. Integrated disease information system
DE19625835A1 (en) 1996-06-27 1998-01-02 Siemens Ag Medical system architecture
US5837674A (en) 1996-07-03 1998-11-17 Big Bear Bio, Inc. Phosphopeptides and methods of treating bone diseases
US5919808A (en) 1996-10-23 1999-07-06 Zymogenetics, Inc. Compositions and methods for treating bone deficit conditions
US5945412A (en) 1996-12-09 1999-08-31 Merck & Co., Inc. Methods and compositions for preventing and treating bone loss
US8545569B2 (en) 2001-05-25 2013-10-01 Conformis, Inc. Patient selectable knee arthroplasty devices
GB9702202D0 (en) 1997-02-04 1997-03-26 Osteometer Meditech As Diagnosis of arthritic conditions
JP3863963B2 (en) 1997-03-27 2006-12-27 大日本印刷株式会社 Digital data correction and storage method and apparatus for X-ray image
US6226393B1 (en) 1997-07-04 2001-05-01 Torsana Osteoporosis Diagnostics A/S Method for estimating the bone quality or skeletal status of a vertebrate
EP1014857A4 (en) 1997-08-19 2006-10-25 John D Mendlein Multi-site ultrasound methods and devices, particularly for measurement of fluid regulation
JP3706719B2 (en) 1997-08-19 2005-10-19 キヤノン株式会社 Image processing apparatus and method, and storage medium
US6064716A (en) 1997-09-05 2000-05-16 Cyberlogic, Inc. Plain x-ray bone densitometry apparatus and method
US5917877A (en) 1997-09-05 1999-06-29 Cyberlogic, Inc. Plain x-ray bone densitometry apparatus and method
AU739275B2 (en) 1997-09-09 2001-10-11 Procter & Gamble Company, The Method of increasing bone volume using non-naturally-occurring FP selective agonists
US5852647A (en) 1997-09-24 1998-12-22 Schick Technologies Method and apparatus for measuring bone density
JP3656695B2 (en) 1997-09-30 2005-06-08 富士写真フイルム株式会社 Bone measuring method and apparatus
US6252928B1 (en) 1998-01-23 2001-06-26 Guard Inc. Method and device for estimating bone mineral content of the calcaneus
JPH11239165A (en) 1998-02-20 1999-08-31 Fuji Photo Film Co Ltd Medical network system
WO1999045371A1 (en) 1998-03-02 1999-09-10 Image Anaylsis, Inc. Automated x-ray bone densitometer
US6013031A (en) 1998-03-09 2000-01-11 Mendlein; John D. Methods and devices for improving ultrasonic measurements using anatomic landmarks and soft tissue correction
US6077224A (en) 1998-03-23 2000-06-20 Lang; Philipp Methods and device for improving broadband ultrasonic attenuation and speed of sound measurements using anatomical landmarks
CA2322420A1 (en) 1998-03-09 1999-09-16 Philipp Lang Methods and devices for improving broadband ultrasonic attenuation and speed of sound measurements
EP0952726B1 (en) 1998-04-24 2003-06-25 Eastman Kodak Company Method and system for associating exposed radiographic films with proper patient information
US6835377B2 (en) 1998-05-13 2004-12-28 Osiris Therapeutics, Inc. Osteoarthritis cartilage regeneration
US6442287B1 (en) 1998-08-28 2002-08-27 Arch Development Corporation Method and system for the computerized analysis of bone mass and structure
US6714623B2 (en) 1998-08-31 2004-03-30 Canon Kabushiki Kaisha Image collecting system
JP3639750B2 (en) 1998-08-31 2005-04-20 キヤノン株式会社 Image acquisition device
US7239908B1 (en) 1998-09-14 2007-07-03 The Board Of Trustees Of The Leland Stanford Junior University Assessing the condition of a joint and devising treatment
JP2002532126A (en) 1998-09-14 2002-10-02 スタンフォード ユニバーシティ Joint condition evaluation and damage prevention device
US6368326B1 (en) 1998-09-28 2002-04-09 Daos Limited Internal cord fixation device
US6501827B1 (en) 1998-09-29 2002-12-31 Canon Kabushiki Kaisha Examination system, image processing apparatus and method, medium, and x-ray photographic system
JP3499761B2 (en) 1998-10-22 2004-02-23 帝人株式会社 Bone image processing method and bone strength evaluation method
DE19853965A1 (en) 1998-11-23 2000-05-31 Siemens Ag Bone contour and bone structure determination
US7283857B1 (en) 1998-11-30 2007-10-16 Hologic, Inc. DICOM compliant file communication including quantitative and image data
US6302582B1 (en) 1998-12-22 2001-10-16 Bio-Imaging Technologies, Inc. Spine phantom simulating cortical and trabecular bone for calibration of dual energy x-ray bone densitometers
US6430427B1 (en) 1999-02-25 2002-08-06 Electronics And Telecommunications Research Institute Method for obtaining trabecular index using trabecular pattern and method for estimating bone mineral density using trabecular indices
JP4067220B2 (en) 1999-03-25 2008-03-26 富士フイルム株式会社 Quality control system for medical diagnostic equipment
WO2000072216A1 (en) 1999-05-20 2000-11-30 Torsana Osteoporosis Diagnostics A/S Method and apparatus for selection and evaluation of substances in treatment of bone disorders
US6178225B1 (en) 1999-06-04 2001-01-23 Edge Medical Devices Ltd. System and method for management of X-ray imaging facilities
US6356621B1 (en) 1999-07-14 2002-03-12 Nitto Denko Corporation Pressure-sensitive adhesive sheet for radiography
US6694047B1 (en) 1999-07-15 2004-02-17 General Electric Company Method and apparatus for automated image quality evaluation of X-ray systems using any of multiple phantoms
US6285901B1 (en) 1999-08-25 2001-09-04 Echo Medical Systems, L.L.C. Quantitative magnetic resonance method and apparatus for bone analysis
US6490476B1 (en) 1999-10-14 2002-12-03 Cti Pet Systems, Inc. Combined PET and X-ray CT tomograph and method for using same
US6246745B1 (en) 1999-10-29 2001-06-12 Compumed, Inc. Method and apparatus for determining bone mineral density
US6605591B1 (en) 1999-11-12 2003-08-12 Genelabs Technologies, Inc. Treatment of subnormal bone mineral density
US6336903B1 (en) 1999-11-16 2002-01-08 Cardiac Intelligence Corp. Automated collection and analysis patient care system and method for diagnosing and monitoring congestive heart failure and outcomes thereof
US6219674B1 (en) 1999-11-24 2001-04-17 Classen Immunotherapies, Inc. System for creating and managing proprietary product data
US6385283B1 (en) 1999-11-24 2002-05-07 Hologic, Inc. Device and method for determining future fracture risk
US6315553B1 (en) 1999-11-30 2001-11-13 Orametrix, Inc. Method and apparatus for site treatment of an orthodontic patient
FR2801776B1 (en) * 1999-12-03 2002-04-26 Commissariat Energie Atomique METHOD OF USING AN OSTEODENSITOMETRY SYSTEM, BY BI-ENERGY X-RADIATION, WITH A CONICAL BEAM
KR100343777B1 (en) 1999-12-10 2002-07-20 한국전자통신연구원 Method for calibrating trabecular index using sawtooth-shaped rack
US6463344B1 (en) 2000-02-17 2002-10-08 Align Technology, Inc. Efficient data representation of teeth model
WO2001063488A2 (en) 2000-02-25 2001-08-30 Healthscreen International, Inc. Method for centralized health data management
AU2001238726A1 (en) 2000-03-01 2001-09-12 Medeview.Com, Inc. A medical diagnosis and prescription communications delivery system, method and apparatus
US6775401B2 (en) 2000-03-29 2004-08-10 The Trustees Of The University Of Pennsylvania Subvoxel processing: a method for reducing partial volume blurring
US7088847B2 (en) 2000-07-19 2006-08-08 Craig Monique F Method and system for analyzing animal digit conformation
JP2002045722A (en) 2000-08-03 2002-02-12 Inax Corp Garbage crusher
US6249692B1 (en) 2000-08-17 2001-06-19 The Research Foundation Of City University Of New York Method for diagnosis and management of osteoporosis
WO2002017210A2 (en) 2000-08-18 2002-02-28 Cygnus, Inc. Formulation and manipulation of databases of analyte and associated values
US6904123B2 (en) 2000-08-29 2005-06-07 Imaging Therapeutics, Inc. Methods and devices for quantitative analysis of x-ray images
US7050534B2 (en) 2000-08-29 2006-05-23 Imaging Therapeutics, Inc. Methods and devices for quantitative analysis of x-ray images
US7467892B2 (en) 2000-08-29 2008-12-23 Imaging Therapeutics, Inc. Calibration devices and methods of use thereof
WO2002017789A2 (en) 2000-08-29 2002-03-07 Imaging Therapeutics Methods and devices for quantitative analysis of x-ray images
US20020186818A1 (en) 2000-08-29 2002-12-12 Osteonet, Inc. System and method for building and manipulating a centralized measurement value database
EP1322224B1 (en) 2000-09-14 2008-11-05 The Board Of Trustees Of The Leland Stanford Junior University Assessing condition of a joint and cartilage loss
WO2002022014A1 (en) 2000-09-14 2002-03-21 The Board Of Trustees Of The Leland Stanford Junior University Assessing the condition of a joint and devising treatment
US6799066B2 (en) 2000-09-14 2004-09-28 The Board Of Trustees Of The Leland Stanford Junior University Technique for manipulating medical images
US7660453B2 (en) 2000-10-11 2010-02-09 Imaging Therapeutics, Inc. Methods and devices for analysis of x-ray images
US8639009B2 (en) 2000-10-11 2014-01-28 Imatx, Inc. Methods and devices for evaluating and treating a bone condition based on x-ray image analysis
US20070047794A1 (en) 2000-10-11 2007-03-01 Philipp Lang Methods and devices for analysis of x-ray images
EP1324695B1 (en) 2000-10-11 2011-07-13 ImaTx, Inc. Methods and devices for analysis of x-ray images
AU2002213351A1 (en) 2000-10-17 2002-04-29 Maria-Grazia Ascenzi System and method for modeling bone structure
EP1346325A2 (en) 2000-10-31 2003-09-24 Ecole de Technologie Superieure High precision modeling of a body part using a 3d imaging system
KR100419573B1 (en) * 2000-12-14 2004-02-19 한국전자통신연구원 Method for evaluating trabecular bone using X-ray image
DE20100641U1 (en) 2001-01-27 2001-07-26 Steer Sebastian Universally adjustable holder system for easy positioning of a recording medium for X-rays
FR2820966B1 (en) 2001-02-16 2003-04-04 Commissariat Energie Atomique DOUBLE ENERGY RADIOGRAPHY METHOD, AND CALIBRATION DEVICE THEREFOR
US6975894B2 (en) 2001-04-12 2005-12-13 Trustees Of The University Of Pennsylvania Digital topological analysis of trabecular bone MR images and prediction of osteoporosis fractures
US20050037515A1 (en) 2001-04-23 2005-02-17 Nicholson Jeremy Kirk Methods for analysis of spectral data and their applications osteoporosis
US6829378B2 (en) 2001-05-04 2004-12-07 Biomec, Inc. Remote medical image analysis
US8000766B2 (en) 2001-05-25 2011-08-16 Imatx, Inc. Methods to diagnose treat and prevent bone loss
GB0124947D0 (en) 2001-10-17 2001-12-05 Mccue Plc Bone simulation analysis
US6895077B2 (en) 2001-11-21 2005-05-17 University Of Massachusetts Medical Center System and method for x-ray fluoroscopic imaging
US20030133601A1 (en) * 2001-11-23 2003-07-17 University Of Chicago Automated method and system for the differentiation of bone disease on radiographic images
ATE365566T1 (en) 2001-12-20 2007-07-15 Bone Support Ab NEW BONE MINERAL SUBSTITUTE
AU2003206822C1 (en) 2002-02-08 2009-01-22 F. Hoffmann-La Roche Ag Use of 15-lipoxygenase inhibitors for treating and preventing bone loss
JP3799603B2 (en) 2002-02-13 2006-07-19 勇 鹿島 Trabecular structure analysis method and trabecular structure improvement effect judgment support method
EP1349098B1 (en) 2002-03-27 2008-05-28 Agfa HealthCare NV Method of performing geometric measurements on digital radiological images using graphical templates
US20050168460A1 (en) 2002-04-04 2005-08-04 Anshuman Razdan Three-dimensional digital library system
EP1357480A1 (en) 2002-04-17 2003-10-29 Agfa-Gevaert Osteoporosis screening method
US20030198316A1 (en) 2002-04-17 2003-10-23 Piet Dewaele Osteoporosis screening method
AU2003232063A1 (en) 2002-05-06 2003-11-11 Institute For Infocomm Research Simulation system for medical procedures
AU2003229328A1 (en) 2002-05-17 2003-12-02 Jerome L. Ackerman Method and apparatus for quantitative bone matrix imaging by magnetic resonance imaging
KR100442503B1 (en) 2002-05-18 2004-07-30 엘지.필립스 엘시디 주식회사 Image quality analysis method and system for display device by using the fractal dimension
EP1550024A2 (en) 2002-06-21 2005-07-06 Cedara Software Corp. Computer assisted system and method for minimal invasive hip, uni knee and total knee replacement
US6904118B2 (en) 2002-07-23 2005-06-07 General Electric Company Method and apparatus for generating a density map using dual-energy CT
US8965075B2 (en) 2002-09-16 2015-02-24 Imatx, Inc. System and method for predicting future fractures
WO2004025541A1 (en) 2002-09-16 2004-03-25 Imaging Therapeutics, Inc. Imaging markers in musculoskeletal disease
US7840247B2 (en) 2002-09-16 2010-11-23 Imatx, Inc. Methods of predicting musculoskeletal disease
US6836557B2 (en) 2002-10-02 2004-12-28 VirtualS{tilde over (c)}opics, LLC Method and system for assessment of biomarkers by measurement of response to stimulus
US20040101186A1 (en) 2002-11-27 2004-05-27 Xin Tong Initializing model-based interpretations of digital radiographs
US7769214B2 (en) 2002-12-05 2010-08-03 The Trustees Of The University Of Pennsylvania Method for measuring structural thickness from low-resolution digital images
ATE484231T1 (en) 2003-01-07 2010-10-15 Imaging Therapeutics Inc DEVICE FOR PREDICTING MUSCLE/SKELETAL DISEASES
US7848558B2 (en) 2003-02-14 2010-12-07 The University Of Chicago Method and system for fractal-based analysis of medical image texture
US7664298B2 (en) 2003-03-25 2010-02-16 Imaging Therapeutics, Inc. Methods for the compensation of imaging technique in the processing of radiographic images
AU2003218427A1 (en) 2003-03-27 2004-11-23 Wright State University Osteoporosis screening using radiographic absorptiometry of the mandible
US7092749B2 (en) 2003-06-11 2006-08-15 Siemens Medical Solutions Usa, Inc. System and method for adapting the behavior of a diagnostic medical ultrasound system based on anatomic features present in ultrasound images
US20050015002A1 (en) 2003-07-18 2005-01-20 Dixon Gary S. Integrated protocol for diagnosis, treatment, and prevention of bone mass degradation
US20050059887A1 (en) 2003-09-16 2005-03-17 Hassan Mostafavi Localization of a target using in vivo markers
US8290564B2 (en) 2003-09-19 2012-10-16 Imatx, Inc. Method for bone structure prognosis and simulated bone remodeling
US8073521B2 (en) 2003-09-19 2011-12-06 Imatx, Inc. Method for bone structure prognosis and simulated bone remodeling
GB0325523D0 (en) 2003-10-31 2003-12-03 Univ Aberdeen Apparatus for predicting bone fracture risk
DE602004015739D1 (en) 2004-05-18 2008-09-25 Agfa Healthcare Nv Method for the automatic positioning of geometric objects in medical images
EP1789924A2 (en) 2004-09-16 2007-05-30 Imaging Therapeutics, Inc. System and method of predicting future fractures
JP5116947B2 (en) 2005-03-02 2013-01-09 株式会社沖データ Transfer device and image forming apparatus
US20070156066A1 (en) 2006-01-03 2007-07-05 Zimmer Technology, Inc. Device for determining the shape of an anatomic surface
WO2008034101A2 (en) 2006-09-15 2008-03-20 Imaging Therapeutics, Inc. Method and system for providing fracture/no fracture classification
US8377016B2 (en) 2007-01-10 2013-02-19 Wake Forest University Health Sciences Apparatus and method for wound treatment employing periodic sub-atmospheric pressure
US8617175B2 (en) 2008-12-16 2013-12-31 Otismed Corporation Unicompartmental customized arthroplasty cutting jigs and methods of making the same
US8939917B2 (en) 2009-02-13 2015-01-27 Imatx, Inc. Methods and devices for quantitative analysis of bone and cartilage
US9330490B2 (en) 2011-04-29 2016-05-03 University Health Network Methods and systems for visualization of 3D parametric data during 2D imaging

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8913818B2 (en) 2000-10-11 2014-12-16 Imatx, Inc. Methods and devices for evaluating and treating a bone condition based on X-ray image analysis
US9275469B2 (en) 2000-10-11 2016-03-01 Imatx, Inc. Methods and devices for evaluating and treating a bone condition on x-ray image analysis
US9267955B2 (en) 2001-05-25 2016-02-23 Imatx, Inc. Methods to diagnose treat and prevent bone loss
US8965075B2 (en) 2002-09-16 2015-02-24 Imatx, Inc. System and method for predicting future fractures
US9460506B2 (en) 2002-09-16 2016-10-04 Imatx, Inc. System and method for predicting future fractures
US9155501B2 (en) 2003-03-25 2015-10-13 Imatx, Inc. Methods for the compensation of imaging technique in the processing of radiographic images
US8965087B2 (en) 2004-09-16 2015-02-24 Imatx, Inc. System and method of predicting future fractures
US8939917B2 (en) 2009-02-13 2015-01-27 Imatx, Inc. Methods and devices for quantitative analysis of bone and cartilage
JP7187714B2 (en) 2019-04-12 2022-12-12 フラウンホーファー-ゲゼルシャフト・ツール・フェルデルング・デル・アンゲヴァンテン・フォルシュング・アインゲトラーゲネル・フェライン A method for calculating the deformation of an object in a time-resolved form and a computer program for the calculation

Also Published As

Publication number Publication date
US20050010106A1 (en) 2005-01-13
US20120027283A1 (en) 2012-02-02
US9155501B2 (en) 2015-10-13
US20130039592A1 (en) 2013-02-14
US20150003712A1 (en) 2015-01-01
US20100130832A1 (en) 2010-05-27
WO2004086972A2 (en) 2004-10-14
WO2004086972A3 (en) 2005-04-28
US20160253797A1 (en) 2016-09-01
EP1605824A2 (en) 2005-12-21
US7995822B2 (en) 2011-08-09
US8260018B2 (en) 2012-09-04
US8781191B2 (en) 2014-07-15
CA2519187A1 (en) 2004-10-14
JP2007524438A (en) 2007-08-30
US7664298B2 (en) 2010-02-16

Similar Documents

Publication Publication Date Title
US9155501B2 (en) Methods for the compensation of imaging technique in the processing of radiographic images
US9275469B2 (en) Methods and devices for evaluating and treating a bone condition on x-ray image analysis
US9767551B2 (en) Methods and devices for analysis of x-ray images
US6811310B2 (en) Methods and devices for analysis of X-ray images
US20070047794A1 (en) Methods and devices for analysis of x-ray images
US8965075B2 (en) System and method for predicting future fractures
AU2002213193B2 (en) Methods and devices for analysis of X-ray images
AU2007201613A1 (en) Methods and devices for analysis of X-ray images
AU2002213193A1 (en) Methods and devices for analysis of X-ray images

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A2

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BW BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE EG ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NA NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL SY TJ TM TN TR TT TZ UA UG US UZ VC VN YU ZA ZM ZW

AL Designated countries for regional patents

Kind code of ref document: A2

Designated state(s): BW GH GM KE LS MW MZ SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IT LU MC NL PL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
COP Corrected version of pamphlet

Free format text: PAGE 9/17, DRAWINGS, REPLACED BY A NEW PAGE 9/17

WWE Wipo information: entry into national phase

Ref document number: 2519187

Country of ref document: CA

WWE Wipo information: entry into national phase

Ref document number: 2004758337

Country of ref document: EP

Ref document number: 2006509289

Country of ref document: JP

WWE Wipo information: entry into national phase

Ref document number: 20048127115

Country of ref document: CN

WWP Wipo information: published in national office

Ref document number: 2004758337

Country of ref document: EP