WO2000062241A1

WO2000062241A1 - Method and apparatus for determining microscope specimen preparation type

Info

Publication number: WO2000062241A1
Application number: PCT/US2000/009987
Authority: WO
Inventors: James J. Boisseranc; Andrew D. Silber; Michael A. Levine; Mark Shuxing Sun; Richard K. Johnson
Original assignee: Tripath, Inc.
Priority date: 1999-04-14
Filing date: 2000-04-14
Publication date: 2000-10-19
Also published as: AU4240200A

Abstract

A system to determine biological specimen slide preparation type (20) and a system to perform automated slide preparation type based classification and analysis. An automated biological specimen analysis system obtains an image of a biological specimen slide (12). The image is processed to compute a number of analysis scores (86) from portions of the image of the biological specimen slide. Geometric features (22) and texture features (28) are derived from the image of the biological specimen (12). Patterns (106A, 106B, 106C,) based on the known biological specimen slide preparation types are compared to the map. A ratio of the number of map elements containing specimen shape to the total number of map elements forms a figure of merit. The system may also confirm a manual determination of slide preparation type (20) and provide error warnings if inconsistencies are found.

Description

METHOD AND APPARATUS FOR DETERMINING MICROSCOPE

SPECIMEN PREPARATION TYPE

FIELD OF THE INVENTION

The invention relates to a method and apparatus for the determination of the type of preparation process used to create a microscope specimen slide, and more particularly to a method and apparatus to determine microscope specimen preparation type from microscope specimen slide image information and an analysis of scores obtained from an automated cytology system.

BACKGROUND OF THE INVENTION

Techniques used to prepare biological specimen slides have evolved and continue to evolve. This evolution presents a number of challenges to a cytology laboratory, such as a Pap screening laboratory, and producers of automated biological specimen systems, such as producers of automated biological specimen systems that automatically provide a quality control or screening function. Different slide preparation types create slides with different specimen characteristics. Different specimen characteristics may affect specimen presentation to human and automated reviewers and may affect processing results. Cells and artifacts present differently depending on biological specimen slide preparation type. Specimen preparation type may also affect the way processing results should be interpreted. Additionally, each type of preparation covers different proportions of the biological specimen slide.

The invention for the first time recognizes that the impact of slide preparation type on the biological specimen indicates that manual and automated procedures must be tailored to slide preparation type to accomplish successful diagnosis. Furthermore, intermingling slides of different slide preparation types should be avoided or handled. Additionally, slide preparation type should be checked prior to processing to assure accurate slide-processing results.

Currently, in the area of cervical cytology, the methods of preparing a conventional Pap smear have been joined by a number of liquid based methods. Conventional techniques of obtaining cytological samples from patients, including cervical samples, involve time tested and well- known procedures. As part of these procedures, a clinician obtains a certain amount of specimen material from the patient, the cellular sample, using a collection device. The clinician then prepares the slide, the conventional cervical smear, by smearing the cellular sample from the collection device onto a glass slide surface. This process creates a smear of cells, blood and mucus that can be examined by cytologists looking for abnormal cells. A smear prepared from a cervical sample that has been treated with the Papanicolaou stain is called a Pap smear.

Recently, modified versions of the conventional Pap smear, utilizing a liquid based emulsion of cells that are placed on a small area of the slide, have been developed. These methods are in use in labs both internationally and domestically. Typically, the specimen is suspended in a liquid preparation, processed to remove blood, mucous and other artifacts and placed in a consistent pattern near the center of the slide. The slides used in this preparation may contain silk-screen markings, called fiducial markings or fiducials, on the slide surface that frame the specimen area. Silk-screen text, for example a company name, may also be included on the slide surface. One such preparation is called a THINPREP from Cytyc Corporation and another type of preparation is called an AUTOCYTE PREP from AUTOCYTE Corporation. Such preparations have the advantages of good cell fixation, lack of air drying, relative cell segmentation, and generally more even distribution of cellular material across the slide surface The process that is used to produce liquid-based slides results in a fairly consistent pattern on the slide space. The pattern is unique for each specific manufacturer's process. Conventional smears are typically a random pattern covering a larger area of the slide.

Therefore, there exist different cervical cytology preparations that may be analyzed by an automatic biological analysis system. These systems must be able to discriminate between different specimen preparation types because the type of specimen preparation has a direct bearing on the analytical steps required to optimally screen the specimen. Therefore, there is a need in the art to automatically determine specimen type in order to direct specimen analysis and confirm slide preparation type determination.

The problem of detecting specimen preparation type is complicated for several reasons:

1. The slides often contain silk screen markings on the surface . The marks must be differentiated from the cellular material in the specimen.

2. The liquid preparation is sometimes annular or contains irregular edges . 3. The conventional preparations may contain circular swirls near the center of the slide making it difficult to distinguish from liquid preparations .

4. The slides may contain bubbles in the adhesive under the coverslip, which partially obscure portions of the specimen.

Therefore, it is one motivation of the invention to provide a system that analyzes slides prepared using different processes to differentiate between slide preparation types because the analysis and setup methods are potentially different .

Additionally, it is another motivation of the invention to provide a preparation detection system that does not add significant time to the processing of a slide

Additionally, it is a motivation of the invention to provide a method and apparatus to determine the type of slide being analyzed automatically without human intervention. Therefore, there is a need in the art to handle the impact of different specimen preparation techniques on automated processing and classification.

SUMMARY OF THE INVENTION The invention provides a method and apparatus for determining a slide preparation type of a biological specimen slide. The invention obtains an image of a biological specimen slide to provide a digital representation of the image. The invention determines the slide preparation type from the digital representation based on the analysis of geometric or texture features or a combination of both.

The invention also provides a method and apparatus for determining the presence of dust on a slide. The invention scores the slide to generate a filtered score array. A processor scans all elements in the filtered score array to determine any elements that have a score within a first predetermined range, scans a neighborhood of any elements that have a score within the first predetermined range to determine a total number of neighbors that have a score within a second predetermined range, counts a total number of neighbors having a score in a predetermined range, and determines whether or not the count is less than a predetermined value and if the count exceeds a first predetermined value, applying a second predetermined value to a mask for all neighbors wherein the mask indicates the presence of dust on the slide.

The invention further provides a method and apparatus for determining the presence of a fiducial mark on a slide. The invention scores the slide to generate a filtered score array. A processor scans all elements in an array to determine any elements that have a score in a first predetermined range, scans a neighborhood of any elements that have a score within the first predetermined range to determine the total number of neighbors that have a score in a second predetermined range, counts the total number of neighbors that have a score in the second predetermined range having a count output, determines whether or not the count output exceeds a first predetermined value and if the count is less than a first predetermined value, applying a second predetermined value to a mask for all neighbors wherein the mask indicates the presence of a fiducial mark on the slide. Other objects, features and advantages of the present invention will become apparent to those skilled in the art through the description of the preferred embodiment, claims and drawings herein wherein like numerals refer to like elements.

BRIEF DESCRIPTION OF THE DRAWINGS To illustrate this invention, a preferred embodiment will be described herein with reference to the accompanying drawings .

Figures 1A, IB and IC show an apparatus for automated slide preparation type determination and slide preparation type based classification.

Figure 2 shows the method of the invention to determine slide preparation type.

Figure 3 shows the method of the invention to determine geometric features . Figure 4 shows the method of the invention to determine texture features .

Figure 5 shows the method of the invention to perform statistical analysis of texture features.

Figure 6 shows the method of the invention to create a mask array.

Figure 7 shows the method of the invention to determine the presence of fiducial marks on a slide.

Figure 8 shows the method of the invention to determine the presence of dust on a slide.

Figure 9 shows one method of the invention to automatically determine slide preparation type using feature threshold comparison.

Figure 10 shows an alternate method of the invention to automatically determine slide preparation type using a slide preparation type feature classifier.

Figures 11A, 11B, 11C, 11D and HE show a scaled schematic of a field of view score array from a conventional Pap smear slide undergoing the various processing steps of the invention. Figures 12A, 12B, 12C, 12D and 12E show a scaled schematic of a field of view score array from a THINPREP slide undergoing the various processing steps of the invention. Figures 13A, 13B, 13C, 13D and 13E show a scaled schematic of a field of view score array from a AUTOCYTE PREP slide undergoing the various processing steps of the invention.

Figures 14A, 14B and 14C show a schematic of a PREP slide, THINPREP slide, and CytoSpin slide, respectively.

Figure 15 shows the method of the invention to perform slide preparation type based classification. Figure 16 shows the method of the invention to perform slide preparation type based classification with automatic slide preparation type determination.

Figures 17A, 17B and 17C show reports that are dependent on slide preparation type.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT

Refer to Figures 1A, IB and IC that show a schematic diagram of one embodiment of the apparatus of the invention for performing slide preparation type based classification and slide preparation type determination. While the method and apparatus of the invention will be discussed in terms of an example herein related to an automated cytology apparatus, it will be understood that the invention is not so limited. The features and principles of the invention may be applied to other biological slide preparation types such as sputum slides, blood slides, other pathology slides, etc.

The techniques used for determining the slide preparation type are based on specimen geometry and/or specimen texture analysis and as such have wide applicability.

The slide preparation type determination and slide preparation type based classification apparatus of the invention comprises an imaging system 502, a motion control system 504, an image processing system 536, a central processor 540, ^'and a workstation 542. The imaging system 502 is comprised of an illuminator 508, imaging optics 510, a CCD camera 512, an illumination sensor 514 and an image capture and focus system 516. The camera may be a high-resolution camera as is well known in the art . The image capture and focus system 516 provides video timing data to the CCD cameras 512, the CCD cameras 512 provide images comprising scan lines to the image capture and focus system 516. Illumination sensor intensity is provided to the image capture and focus system 516 where an illumination sensor 514 receives the sample of the image from the optics 510. In one embodiment of the invention, the optics may further comprise an automated microscope. The illuminator 508 provides illumination of a slide. The image capture and focus system 516 provides data to a VME bus 538. The VME bus 538 distributes the data to an image processing system 536. The image processing system 536 is comprised of field-of-view processors 568. The images are sent along the image bus 564 from the image capture and focus system 516. The central processor 540 controls the operation of the invention through the VME bus 538.

In one embodiment, the central processor 562 comprises a Motorola 68060 CPU. The motion control system 504 is comprised of a tray handler 518, a microscope stage 520, a microscope turret 522, and a calibration slide 524. The motor drivers 526 position the slide under the optics. A bar code reader 528 reads a barcode located on the slide 524. A touch sensor 530 determines whether a slide is under the microscope objectives, and door interlock 532 prevents operation in case the doors are open. Motion controller 534 controls the motor drivers 526 in response to the central processor 540. An ETHERNET communication system 560 communicates to a workstation 542 to provide control of the system. Workstation processor 550 controls a hard disk 544. In one embodiment, workstation 542 may comprise a SUN SPARC ULTRA workstation. A tape drive 546 is connected to the workstation processor 550 as well as a modem 548, a monitor 552, a keyboard 554, and a mouse-pointing device 556. A printer 558 is connected to the ETHERNET network system 560.

During image collection integrity checking, the central processor 540, running a real time operating system, controls the automated microscope and the processor to acquire and digitize images from the microscope. The flatness of the slide may be checked, for example, by contacting the four corners of the slide using a computer controlled touch sensor. The central processor 540 also controls the microscope stage to position the specimen under the microscope objective, and from one to 15 field of view (FOV) processors 568 which receive images under control of the central processor 540. Referring now to Figure IC, there is shown placement of a slide 12 into an optical path of an automated microscope 511 having a turret 522 and a CCD camera 512. The slide 12 may be mounted on a stage 520 substantially in a horizontal X,Y plane that intersects the optical path. The stage 520 is movable in the X, Y plane as well as along a Z axis which is perpendicular to the X, Y plane and which is parallel to the optical axis of the optics 510 of the automated microscope 511. The turret 522 may comprise multiple objective lenses as is well known in the art . The microscope turret control 523 provides signals in a well-known manner for positioning a selected objective lens into position for viewing the slide 12. Illuminator 508 illuminates the slide 12. It is to be understood that the various processes described herein may be implemented in software suitable for running on a digital processor. The software may be embedded, for example, in the central processor 540. The slide 12 may be coverslipped. The coverslip may be detected following the methods described in assignee's United States Patents and copending United States Patent applications referred to herein. The automated analysis system may be operated in two modes : quality control mode and screening mode. In the quality control mode, the automated analysis system generates a quality control score.

In the screening mode, the automated analysis system generates an analysis score. These automated methods of performing screening and quality control steps employ software used by the central processor 540 to provide an analysis score or quality control score. The software operates on data that is a machine representation of a slide, such as a digital representation 15 shown in Figure 2 of a slide 12, obtained using the imaging system 502 where the imaging system 502 views the slide. The software employs thresholds that are used to determine the likelihood of normalcy or malignancy of the slide. The thresholds may also be used to determine subpopulations of slides for quality control checking or to select subpopulations of slides that are so clearly normal that they do not need checking. In any case, the different types of Pap smear preparations often require different thresholds and in general different types of biological specimen slide preparations will often require different analysis procedures and/or parameters . Refer now to Figure 2 , which shows the method of the invention to determine the slide preparation type from a scan of a biological specimen slide, such as a Pap smear slide. In step 14, the slide is scanned at low power magnification to generate a digital representation 15 of the slide 12. The digital representation 15 may be a field of view representation using an orthogonal grid where each field of view has an associated field of view score. In step 27, a geometric analysis of the slide generates geometric features 22 from the digital representation 15. These geometric features 22 include the specimen center location, the percent of specimen found in a pattern and the count of zones that are of a predetermined type in a pattern. Each zone may correspond to a field of view or may contain multiple fields of view. Step 27 is described in more detail with reference to Figure 3. In step 23, a texture analysis of the slide generates texture features 28. These texture features 28 include a statistical analysis of an image and statistical analysis of a number of image convolutions. Step 23 is described in more detail with reference to Figure 4. In step 24, automatic slide preparation type determination is performed based on either the geometric features 22 or the texture features 28 or a combination of both sets of features. The slide preparation type 20 is automatically generated using either a threshold method, described in more detail with reference to Figure 9, or a classifier method, described in more detail with reference to Figure 10.

Refer now to Figure 3 , which shows the method of the invention to determine specimen geometry features. During low resolution processing, the slide 12 is scanned with a 4x-microscope objective imaging a Field of View (FOV) of 1.408 Millimeters Square into a 512x512 pixel array. The images cover the entire area under the slide coverslip. Each 4x image is partitioned into a 5 by 5 grid pattern of 20x fields of view, also called 20x zones. Each 20x zone measures 512 by 512 pixels if imaged using a 2Ox objective lens. Examples of such a partitioning is shown in Figures HA, 12A and 13A. The 4x low resolution processing is designed to detect cellular objects. The biological analysis system performs focus and imaging functions that result in a digital representation 15 of the slide 12. The image may be acquired in a conventional fashion using a CCD camera 512 or equivalent. These focus and imaging functions are described in assignee's United States Patents and copending United States Patent applications referred to herein. Those skilled in the art will recognize that the invention is equally applicable to analyses done at a single magnification or magnifications other than 4x or 2Ox. The process may also be expanded to include other biological slides other than Pap smear slides such as histological slides and immunohistological slides. The central processor 540 implements the process described in Figure 3.

Each 20x zone is scored in step 32. These scores are stored in score array 86. One example score, the SIL score, ranks each 2Ox zone according to its likelihood of containing SIL (Squamous Intraepithelial Lesion) cells. A score of 10 indicates the highest likelihood that an FOV contains SIL cells. A score of 0 indicates there is little likelihood of the FOV containing SIL cells. The basis of the information processing steps relating to the interpretation of the SIL score rests on the assumption that specimen material will score differently from empty slide space, dusty slide space and fiducial slide marks.

Zones with specimen material will generally have higher SIL scores than empty space, dust and fiducial marks. Those skilled in the art will recognize that the SIL score could be replaced with other FOV scores that indicate the presence of specimen material, such as cell count or cell group scores . The SIL scores are stored in a two-dimensional array, called the score array 86, dimensioned such that each of the 2Ox zones is represented by an element in the score array 86. The indexes of the elements of the score array correspond with a 2Ox zone on the slide. For example, shown on Figure HA is a 20x zone 215 with an index of (0,21) with a SIL score of 1. As previously stated, each 1.408- millimeter on the slide contains a 5 by 5 grid of 2Ox zones; therefore, each increment of the results array index represents an increase of 0.282 millimeters in the corresponding direction. For a 25 x 50 millimeter coverslip, this results in a score array 86 size of approximately 90 by 180 elements. The score array 86 is used to generate a mask array 94 in mask generation step 92. During the mask generation step 92, information from the score array 86 is processed with image processing functions that treat the score array 86 as if it were a two dimensional image of the specimen. The steps of mask generation are described in more detail with reference to Figure 6.

A representation of a score array 86 is shown for a conventional slide in Figure HA as map 208, for a THINPREP slide in Figure 12A as map 258, and for an AUTOCYTE PREP slide in Figure 13A as map 298. The example of Figure HA demonstrates the somewhat random patterns typical of conventionally prepared slides. Figure 12A shows an example map for a THINPREP specimen including the semicircular fiducial markings on the slide surface. This example demonstrates the relative consistency of the pattern typical of liquid based preparations. The maps are scaled from an actual array size of 80 by 160 to a 20 by 40 array to more clearly demonstrate the appearance of the array in a limited amount of space.

The image processing techniques used to evaluate the score array 86 are the same regardless of the number of array elements and the type of slide preparation. The invention locates a score pattern that is similarly shaped to a known specimen score pattern. The information content of the relative positions of the scores in the score array 86 contains a signal component and a noise component. For example, the signal component comprises analysis scores corresponding to fields of view of biological material. For example, the noise component comprises fields of view of dust and fiducial marks. Other sources of noise include scratches on the slide or coverslip, dark regions, obscuration, and vignetting from non-uniform illumination. A biological specimen score affected by noise is known as a noisy biological specimen score. The impact of this is reduced so that the signal to noise ratio of the information may be improved. After noise reduction, the information is then further processed to locate a pattern that matches an expected pattern, or the lack thereof in the case of a conventional slide, of a slide preparation type. Therefore, the invention is not limited to the pattern matching steps outlined herein. Any other pattern matching technique may be used without deviating from the spirit and scope of the invention. The score array 86 is exclusive ORed with the mask array 94 in step 88. The exclusive OR step 88 creates a filtered score array 96 where scores related to dust, fiducial markings and other noise have largely been removed. This is also known as spatial filtering, where the spatial filter refers to the space created by the relative positions of the elements of the array. At this point, the filtered score array 96 is treated as a binary image. In step 98, the specimen center 150 of specimen material on the slide is found using a geometric center approach. A method of finding a geometric center is described in Calculus and Analytical Geometry by Sherman K. Stein, McGraw Hill 1982, pp. 484-493. The geometric center of the score array, indexed by x and y, is determined by first summing all score array values of y for each x to generate an x histogram and summing all score array values of x for each y to generate a y histogram. A weighted average for x is computed by first, multiplying the histogram value of x by x and summing over all x, then dividing this sum by the sum of all histogram values. A weighted average for y is computed in a similar manner. The geometric center is the weighted average for x and weighted average for y. The geometric center is taken as the specimen center. The specimen center becomes part of the geometric features 22. In step 30, a number of predetermined patterns, such as pattern 106A, 106B and 106C, are applied to the filtered score array 96 in relation to the specimen center 150. Each pattern corresponds to a specimen preparation type. The manufacturer of the processing equipment used to create a liquid based cervical cytology slide specifies the size and shape of the specimen pattern for each liquid-based preparation type. For example, two shapes are checked, a generally circular pattern of a THINPREP slide and a generally circular pattern of an AUTOCYTE PREP slide. Each one of these slide preparation types is an example of a predetermined slide preparation type. Table A shows four preparation methods and the expected geometrical distribution of the specimen.

TABLE A PREPARATION TYPE AND SPECIMEN PATTERN

In step 108, the percentage of the non-zero elements of the filtered score array 96 that fall within each pattern is calculated. Step 108 generates a percentage for each pattern. Each percentage becomes part of the geometric features 22 available for further processing. In step 110, each predetermined pattern is centered on the specimen center 150 and a count of the non-zero elements of the filtered score array 96 that fall within each pattern is calculated. Each count also becomes part of the geometric features 22 available for further processing. The array elements that are on the border of a pattern are considered to be part of the analysis if any part of the element is in the pattern. In an alternative embodiment of the invention, only those elements that are totally in the pattern participate in the analysis . Two methods of determining slide preparation type from slide features, such as geometric features, are described in detail with reference to Figures 9 and 10.

While geometric analysis of field of view score spatial distribution provides a powerful and effective determination of slide type, there are, however, a few instances where this method fails to determine the slide type correctly. To counter this possibility the invention utilizes texture measures derived from an image of the slide to generate texture features .

Refer now to Figure 4 , which shows the method of the invention to compute texture features 28 from the digital representation 15. In step 39, at least one FOV image 350 is selected from the digital representation 15. Data from the FOV image 350 is processed directly by statistical analysis step 40A.

Refer also to Figure 5, which shows the statistical analysis method 40 in more detail. Texture measures are descriptions of the distribution of material across a Field of View (FOV). Texture measures include:

1. the Mean of all image pixel values;

2. the Standard Deviation of all image pixel values; 3. the Skewness, or asymmetry, of all image pixel values; and

4. the Kurtosis, or peakness, of all image pixel values .

In step 354, the mean 42 of the pixel values in the image 350 is calculated. At step 356, the process flows to calculate the standard deviation 44 of the pixel values in the image 350. At step 358, the skewness 46 of the pixel values in the image 350 is calculated. At step 360, the kurtosis 48 of the pixel values in the image 350 is calculated. These statistics are a reflection of the texture of the image 350. For example, the more texture a FOV has, the higher the standard deviation of the pixel values in a FOV will be. A blank field will have a texture score similar to a field so covered in cells that nearly all light is blocked from transmission. A FOV with many well-separated cells will have a high texture measurement. Conventional Pap Smear slides have regions devoid of cells and other regions where the cells are overlapping: these slides have relatively low texture scores . THINPREP and AUTOCYTE Prep slides have regions with many well-separated cells: these slides have high texture scores. The mean 42, skewness 46 and kurtosis 48 may also be interpreted as a measure of texture. It is well known how to calculate these statistics. A good description of the statistics used can be found in Applied Probability and Statistical Methods by George C. Canavos, available from Little Brown & Company, '1984, pp 64-67. The statistical measures of mean 42, standard deviation 44, skewness 46 and kurtosis 48 comprise the texture features 28. Those skilled in the art will see that other measures of texture are possible, including some based on a Fourier transform analysis or wavelet analysis. As with the geometric features, the texture features are further processed to determine the slide preparation type. The determination of a texture measure may also start with some smoothing of the image to remove low frequency power. For example, the step of convolving an image with a kernel may be used to smooth an image . Examples of kernels that may be used include a 3x3, 7x7 and 15x15 kernel. Each kernel comprises an array of ones . The larger the kernel, the more the image is smoothed. In step 370, the image 350 is convolved with a 3x3 array of ones to produced a 3x3 convolved image 33. At step 372, the 3x3 convolved image 33 is subtracted from the original image obtained in step 372 to create a 3x3 convolved subtracted image 34. When the original image is subtracted from the smoothed one the result is to create an image with the high frequency content intact. Once the image 350 has been processed with the kernels, further statistics may be calculated from the smoothed images . The 3x3 convolved subtracted image 34 is processed with the statistical analysis step 40B which has been described with reference to Figure 5 to generate additional texture features. Additional smoothing operations may also be performed. At step 382, the image is convolved with the 7x7 array of ones to generate a 7x7 convolved image 37. The 7x7 convolved image 37 is subtracted from the original image 350, in step 383, to generate a 7x7 convolved subtracted image 36. Statistical analysis 40C, described with reference to Figure 5, is performed on the 7x7 convolved subtracted image 36 to generate additional texture features 28.

In step 392, the image is convolved with a 15x15 array of ones to provide a 15x15 convolved image 35. The 15x15 convolved image 35 is subtracted from the original image 350, in step 393, to generate a 15x15 convolved subtracted image 38. Statistical analysis 40D, described with reference to Figure 5, is performed on the 15x15 convolved subtracted image 38 to generate additional texture features 28. The central processor 540 implements the processes described in Figures 4 and 5 in software .

Using these statistics and kernels a set of features may be created. All of these features are standard statistical measurements of a distribution. For example, the original image with 3 levels of smoothing with 4 statistics results in 16 features. As an example, one feature is calculated by:

1. Convolving the image with a 15x15 kernel of ones ;

2. Subtracting the convolved image from the original image ; and

3. Calculating the skewness of the difference image . In experimental trials, this single feature correctly distinguished THINPREPS from conventional slides 75% of the time based on a single rich FOV.

A rich FOV is an FOV with many cells. Two methods of determining slide preparation type from slide features, such as texture features, are described in detail with reference to Figures 9 and 10. Refer now to Figure 6 which shows the method of the invention to create a mask array 94 from a score array 86. In step 118, setting all array elements to zero creates a blank binary mask array 124. This mask array is a binary homologous image of the score array. The dimension of the array is the same as the score array 86 from the 4X processing where there is a one-to-one correspondence between the elements of the binary mask array 124 and the score array 86. In step 120, a 3x5 processing method is used to determine the location of fiducial markings and in step 130, a 3x3 processing method is used to determine the location of dust. These processing steps fill in elements in the blank mask array 124 to generate the noise indicated mask array 97.

Refer now to Figure 7, which shows the method of the invention to determine the location of fiducial markings on a slide, step 120 in Figure 6. During step 120 the following procedure is used to determine whether or not a value of one is placed in the binary mask in the corresponding position for array elements in the score array 86. In step 152, an element of the score array 86 is checked for a score of one or two. Those skilled in the art will recognize that other scores indicating a fiducial may be used. If the element has a score of one or two, a 3X5 neighborhood scan centered on the element is implemented in step 154. One by one each neighbor of the element in the score array 86 is checked. If, in step 158, a neighbor has a value equal to the value of the center element, then the process increments a counter in step 160.

Otherwise the process checks the next neighbor in step 164 and repeats step 158. If, in step 174, there are no more neighbors to check, the process flows to step 176 to check if the count was greater than an experimentally determined amount; for example, twelve. If, in step 176, the count was greater than or equal to twelve then the elements in the mask array 94 corresponding to the 3x5 grid are set to one in step 178. For each slide preparation type, the number of neighbors required to give a reliable indication of sources of noise such as fiducial marks should be used. In one example embodiment, the experimentally determined amount of 12 is determined by setting different values and noting the resulting maps for each slide preparation type. The process continues by checking the next FOV, in step 156. In step 162, if there are no more FOV's, then the process ends in step 172; otherwise, the process returns to step 152.

Refer now to Figure 8 , which shows the method of the invention to determine the presence of dust on a slide, step 130 in Figure 6. In step 180, an element of the score array 86 is checked for a score of one or two. Those skilled in the art will recognize that other scores indicating dust may be used. If the element has a score of one or two, a 3x3 neighborhood scan centered on the element is implemented in step 182. One by one each neighbor of the element in the score array 86 is checked. If, in step 186, a neighbor has a value of zero, then the process increments a counter in step 188. Otherwise the process checks the next neighbor in step 192 and repeats step 186. If, in step 196, there are no more neighbors to check, the process flows to step 198 to check if the count was greater than an experimentally determined amount; for example, 4. If, in step 198, the count was greater than 4 , then only the element in the mask array 94 corresponding to the central element is set to one in step 200. The process continues by checking the next FOV in step 184. In step 190, if there are no more FOV's, then the process ends in step 194, otherwise the process continues processing at step 180.

Refer again to Figure 6. The noise indicated mask array 97 is then dilated in step 95 to produce the final mask array 94. The primary effect of this dilation is to fill in the fiducial area within the mask. The dilation step 95 may be a 3x3 plus structuring element dilation or a processing step designed to enhance the signal to noise ratio of the array map. Refer now to Figure 9 , which shows the method of the invention to determine the slide preparation type from slide features 50 and feature thresholds 57. Slide features 50 could be either geometric features 22, Figure 3, or texture features 28, Figure 4, or a combination of both types of features . Feature thresholds are obtained by analyzing a set of known slide preparation types and determining what thresholds indicate certain types of preparations. For example, the slide center, percentage of elements within the pattern, and count of elements within the pattern are features that are indicators of the type of slide preparation. These features can be evaluated against predefined thresholds to indicate the slide preparation type. For example, each predetermined pattern in Figure 3 has an associated threshold. The total number of non-zero elements in the filtered score array 96 totaled in the predetermined pattern is a feature 50 as well as the percentage of the non-zero elements to the total number of elements. These features are compared in step 51 to a predetermined threshold 57 corresponding to a chosen pattern. If the percentage is greater than or equal to the threshold, then the slide is determined to be of the corresponding preparation type. For example, if pattern is set to the THINPREP pattern, then the comparison in step 110 will be to the THINPREP threshold. If the result of the comparison indicates that the percentage is greater than or equal to the THINPREP threshold, then the slide preparation type is determined to be THINPREP.

In one embodiment of the invention, the pattern checks occur in a size order from largest to smallest. For example, slide preparation type is initially assumed to be a conventional slide preparation type. If the check for THINPREP indicates that the slide is a THINPREP, then the slide preparation type is set to THINPREP. A check is then made for AUTOCYTE. If the check for AUTOCYTE indicates that the slide is an AUTOCYTE slide, then the slide preparation type is set to AUTOCYTE .

Experimentally, for the THINPREP slide, the threshold has been determined to be 85% of the non- zero zones in the THINPREP pattern, and for the AUTOCYTE slide the threshold is set to 85% of the non-zero zones in the AUTOCYTE pattern.

In one experiment, the application of thresholds to the geometric features of 300 THINPREP slides and 1200 conventional slides successfully identified all of the THINPREP slides and all but two of the conventional slides.

Refer now to Figure 10, which shows an alternative method of the invention to determine slide preparation type. The slide features 50 are provided to a slide preparation type feature classifier 52. The slide preparation type feature classifier 52 may use only geometric features or texture features or the feature classifier 52 may use a combination of features. The feature classifier combines the features to produce a single number using a linear discriminate function. One example discriminate function is the well known Fisher's linear discriminate function. See "Fisher's Linear Discriminant", Pattern Classification and Scene Analysis, by Richard O. Duda and Peter E. Hart, Copyright ^©1973, pp. 114- 121. A good description on the development of classifiers can be found in Applied Multivariate Statistical Analysis by Richard A. Johnson and Dean W. Wichern, available from Prentice Hall, 1988, Chapter 11. The output of the slide preparation type feature classifier 52 is compared 54 with classifier output thresholds 55 determined during a training phase. The comparison 54 provides the slide preparation type 20.

In one experiment, all sixteen texture features referred to in Figure 4 from a single FOV were used to create a classifier that correctly identified all of the THINPREPs and all but one of the Conventionals on a sample of 140 THINPREP slides and 103 Conventional Pap Smears. The inclusion of geometric features adds to the classifier's discrimination ability. For instance, a classifier with five well-chosen texture features plus five well-chosen geometric features would likely provide better discrimination than a classifier with either fifteen geometric features or fifteen texture features.

There are several ways to combine texture features from multiple FOVs . In one embodiment, each FOV is classified as to its likelihood of membership, i.e. whether it is THINPREP or Conventional. The second classifier may be built based on these likelihoods. The classifiers used in the invention may be trained by running a set of training slides of a known slide type through the classifier during a training phase. The training of a classifier is well known and is also described in U.S. Patent No. 5,740,269, issued 04/14/98, to Oh et al., entitled A METHOD AND APPARATUS FOR ROBUST BIOLOGICAL SPECIMEN CLASSIFICATION.

Examples will now be given showing the geometric analysis of three slide preparation types.

Figure HA shows the map of a conventional slide being processed with the 4x field of view low resolution processing scan. Figure HA shows the array of x 202 and y 204 values. Each value is a field of view score and represents a likelihood of having SIL components. The conventional map comprises a number of fields of view where each field of view represents a 20x high-resolution field of view. Each 4x low-resolution field of view, for example low-resolution field of view 210, comprises a 5x5 array of 20x high-resolution fields of view. An example of a field of view is field of view 212, which has a SIL score of 1. Another representative field of view is field of view 214, which has a representative SIL score of 9. The conventional map shown on Figure HA is a scaled down version of a real map for illustrative purposes. The map shown in Figure HA is a 20x40 field of view map where the real slide will have 80x160 fields of view. The invention will work equally well with a system that can capture an image of the entire slide or portions of the slide using a mechanism to move the slide, such as a slide stage. The conventional map 208 comprises a x direction 202 and a y direction 204. Each field of view is addressed in a conventional manner using a x and y coordinate. The conventional map is made in this instance of a Pap smear.

Now refer to Figure HB, which shows the conventional mask map superimposed on the original conventional map 208 of Figure HA. The mask 218 is shown with solid fields of view from the 3x3- processing step and gray fields of view from the 3x5 processing step. One example of a mask field of view is the 3x3 processing mask 216. Even though certain areas such as the gray field of view mask 220 are scored as a fiducial on the mask, the method and apparatus of the invention is robust enough to properly classify a conventional slide as conventional. Therefore, the dust and fiducial processing of the invention does not adversely affect the determination of slide preparation type. Figure HB shows that the 3x3 processing has masked off a number of fields of view and the 3x5 processing has masked off one area.

Figure 11C shows the conventional map with a dilated mask superimposed on the original map. The dilated mask 228 is shown with the 3x3 plus dilation around the 3x3 processing mask 216. The 3x3 plus dilation is also shown around the fiducial mask elements in the gray field of view area mask 220. The 3x3 plus around the field of view processing mask 216 results in additional mask fields of view surrounding the field of view processing mask 216. Likewise, the 3x3 plus dilation results in additional fields of view being masked around the gray field of view area mask 220. Figure HD shows a filtered conventional map 238 after applying the dilated mask 228 where dust 235 and fiducial marks 237 have been removed. The filtered conventional map 238 is a digital representation of slide scores resulting from the dilated mask 228 being exclusive ORed with the conventional map 208. The result of this operation is a filtered conventional map 238 with dirt and fiducial elements removed. Figure 11D shows that the conventional field of view scores are predominately preserved in this processing step for a conventional slide.

Now refer to Figure HE, which shows an illustration of the filtered conventional map 238 where predefined slide type patterns have been superimposed upon the filtered conventional map 238. The superimposition illustrates that AUTOCYTE and THINPREP slide pattern matching result in a substantial amount of material being outside of both the AUTOCYTE pattern 247 and THINPREP pattern 245. The method and apparatus of the invention would indicate strong geometrical evidence that this is a conventional slide because a substantial number of scores lay outside the patterns.

Figure 12A shows the THINPREP map or the field of view scores for x, and y. Figure 12A shows a THINPREP pattern 260 with the field of view scores in substantially the same array configuration as Figure HA However, there is a circular area in the center, which is the THINPREP preparation and the fiducial marks 262 and 264 with field of view scores of 1 or 2.

Refer now to Figure 12B, which shows the THINPREP map with a THINPREP map mask superimposed on map 258. The THINPREP mask 268 shows 3x3 processing 270 and 5x5 processing 272 and 274 to remove fiducials and dust .

Refer now to Figure 12C, which shows the 3x3 plus dilation in the mask 268 where the fiducial marks 272 and 274 have been dilated and the 3x3 processing fields of view have also been dilated.

Refer now to Figure 12D, which shows the filtered THINPREP map 278 after the exclusive OR of the original map 258 and the dilated mask 268. Figure 12D shows that nearly all the remaining fields of view lie within the THINPREP specimen area demonstrating that the invention is very effective in reducing the effects of dust and fiducials.

Now refer to Figure 12E, which shows an illustration of the filtered THINPREP map 288 where predefined slide type patterns have been superimposed upon the filtered THINPREP map 288. The superimposition 248 illustrates that slide pattern matching may result in a substantial amount of material being outside of the AUTOCYTE pattern 287 but within the THINPREP pattern 285. The method and apparatus of the invention would indicate strong geometrical evidence that this is a THINPREP slide because a high percentage of the remaining scores fall within this pattern.

Refer now to Figure 13A, which shows an AUTOCYTE PREP map 298. The AUTOCYTE PREP map has a preparation in the center with two fiducial markings. The AUTOCYTE PREP slides have plus- shaped fiducial marking in the upper right hand corner 302 and lower left hand corner 304. Figure 13B shows the AUTOCYTE PREP mask 308 superimposed on the AUTOCYTE PREP map 298 showing the 3x3 processing 310 and the 3x5 processing to remove fiducials 312.

Figure 13C shows the AUTOCYTE PREP mask 308 after the 3x3 processing dilation as described herein, showing the final results of the mask dilation.

Figure 13D shows the resulting exclusive OR of the original map with the mask showing that the only material left for the analysis is the preparation itself.

Figure 13E shows the filtered AUTOCYTE PREP map 318 where predefined slide type patterns have been superimposed upon the filtered AUTOCYTE PREP map 318. The superimposition 319 illustrates that slide pattern matching results in a high percentage of material being inside of the AUTOCYTE PREP pattern 321 and the THINPREP pattern 323. Although the method and apparatus of the invention would indicate strong geometrical evidence that this is a THINPREP slide because a high percentage of the remaining scores fall within this pattern, it provides even stronger evidence that this is an AUTOCYTE PREP slide because of the smaller area of the AUTOCYTE PREP pattern 321. Specific examples were given for conventionally prepared Pap smear slides, slides prepared using the THINPREP process developed by Cytyc Corporation, and slides prepared using the PREP process developed by AUTOCYTE Corporation. Those skilled in the art will appreciate that these methods may be extended to the detection of other slide preparation types, for example the CYTOSPIN process developed by Shandon Corporation.

By way of example, a schematic of the PREP slide is shown in Figure 14A where the slide 320 has a small circular specimen pattern 322, and a schematic of the THINPREP slide is shown in Figure 14B where the slide 324 has a larger circular specimen pattern 326. Also, a schematic of the CYTOSPIN slide is shown in Figure 14C where the slide 328 has a rectangular pattern 330. The foregoing discussion has described in considerable detail methods of determining the slide preparation type based on the geometric features and texture features derived from an image of a biological specimen slide. Each geometric analysis evaluates specimen geometry including obtaining an image and mapping the location of the specimen. Also, the filtering methods used to identify slide fiducials and coverslip dust were shown to be the same for all preparation types. The final step in determining the slide preparation type is to threshold or classify the features. It was also shown that the method and apparatus of the invention may be extended to other preparation types with other specimen patterns .

Refer now to Figure 15, which shows the method of the invention to perform slide preparation type based classification. The slide 12 is imaged with an automated biological analysis system as described with reference to Figures 1A, IB and IC and as described in assignee ' s United States Patents and co-pending applications described herein and incorporated by reference hereto. The slide 12 is scanned at low power magnification in step 14. The biological analysis system performs focus and imaging functions that result in a digital representation 15 of the slide 12, such as a digital representation from the CCD camera 512. After a digital representation 15 of the slide 12 is obtained, additional processing is performed to select areas to scan with high power magnification in step 16. Classifiers select regions of interest during the low resolution processing. In step 17, biological features obtained at high power magnification are evaluated. In step 18, evaluation thresholds based on the slide preparation type 25 are applied to classify the slide 12 and a slide classification result is generated as a user report in step 29. During the step of applying evaluation thresholds, feature scores are compared to established thresholds to determine a classification category for the slide. Different report types and report options may be selected based on specimen preparation type.

The slide preparation type 20 may be either manually determined wherein the automated analysis system processes batches of slides that are of the same preparation type or each slide may be manually identified individually prior to processing. In an alternate embodiment the slide preparation type may be automatically determined as described herein. Alternatively, the slide preparation type may be determined manually and confirmed automatically. The central processor 540 implements the processes described in Figure 15 in software. Refer now to Figure 16 which shows the method of the invention to perform slide preparation type based slide classification with automatic slide preparation type determination. The slide 12 is and scanned at a low power magnification in step 14. Focus and imaging functions that result in a digital representation 15 of the slide 12 as described herein. Areas of the slide are selected for high power magnification processing in step 16.

In step 17, biological features obtained at high power magnification are evaluated. In step 21, geometric features 22 of the specimen are generated. In step 23, texture features 28 from the specimen are generated. In step 24, automatic slide preparation type determination is performed using either the geometric features 22, texture features 28, or a combination of both as described herein. The slide preparation type 20 is used to apply evaluation thresholds in step 18. Evaluation thresholds based on slide preparation type 20 are applied to classify the slide. During the evaluation step 18 a slide processing result based on the slide preparation type is determined using thresholds . Example thresholds that may depend on slide preparation type are described herein. The slide scores and classification information is printed on a user report in step 27. Example user reports are described below.

Slide preparation type detection is performed on a computer and image acquisition system, such as the one described in Figures 1A, IB and IC, comprising a host computer running a real time operating system; a microscope with a xenon strobe for illumination; a processor to acquire and digitize images from the microscope; a computer controlled microscope stage to position the specimen under the microscope objective; and from one to fifteen Field of View (FOV) processors which receive images under control of the host computer, from the image acquisition processor, and perform image processing and object classification operations on the images and transmit the results to the host computer. During the scan at low power magnification, step 14, a 4x algorithm analyses the entire slide within the boundaries of the cover slip. The method of determining these boundaries is disclosed in assignee's U.S. Patent No. 5,638,459, issued 6/10/97 to Rosenlof et al . , entitled "Method and Apparatus for Detecting a Microscope Slide Coverslip"; U.S. Patent No. 5,566,249, issued 10/15/96 to Rosenlof et al . , entitled "Apparatus for Detecting Bubbles in Coverslip Adhesive" and U.S. Patent 5,812,692, issued 9/22/98 to Rosenlof et al . , which is a divisional of U.S. Patent 5,638,459, ibid. The 4x scan breaks the slide down into a number of fields of view at 4x magnification. The 4x algorithm scores each field of view by breaking down the field of view into 5x5 subfields of view which will be subsequently processed by a 2Ox-magnification algorithm, the evaluation of biological features at high power magnification, in step 17. Those skilled in the art will recognize that a field of view is an arbitrary designation dictated by a number of factors including the type and resolution of the imaging equipment being used. For example, if a 2Ox microscope objective / imager combination is being used that only can image a portion of a slide, then multiple images of the slide need to be acquired to cover the entire slide. Otherwise, if a 2Ox microscope objective/imager combination is being used that can image an entire slide or significant portion of a slide, only a single image need be acquired. This single image or these multiple images may be logically divided or combined into any number of subimages and analyzed to provide a set of scores . Multiple images may be combined to create a composite image or combined images that are subsequently analyzed to provide a set of scores. The invention is equally applicable to other types of imaging equipment and to any method that can obtain an equivalent set of scores that may be further processed. Therefore, the field of view score is a term not limited to a single microscope object field but could correspond to any portion of a microscope object field or any composite or combination of a number of microscope object fields. Neither is the invention limited to orthogonal grids; other topologies, such as a circular grid or other regular pattern or non- regular pattern, may be used that incorporates position information related to the set of analysis scores such that a suitable specimen pattern may be detected as described herein.

After the 4x algorithm has assigned a specific score to each of the subfields of view, geometric features 22 and texture features 28 are generated. The geometric features 22 and texture features 28 are provided to an automatic slide preparation type determination step 24 which generates a slide preparation type which is stored along with other slide data. The image is further processed with 2Ox field of view algorithms and classified using a thresholding step that utilizes slide data to generate an analysis score, adequacy or quality control score.

The slide data includes thresholds and a classification rate for applying evaluation thresholds in step 18. The slide is processed using higher resolution processing steps that output a number of scores, such as a slide score and an adequacy score. The adequacy score is related to the probability that endocervical cells have been found on the specimen for a Pap smear slide for example. Methods of computing a slide score or an analysis score or a quality control score are well known in the art . Such methods are described in assignee's United States Patents and United States Patent applications referred to herein. Other scores are described in more detail herein. During step 18 the process applies evaluation thresholds based on the slide preparation type 20. The slide preparation type 20 is extracted from a slide database along with a slide preparation type classification rate. The evaluation step 18 then determines the action to be taken and generates a slide results report.

One example of the slide classification decision logic used to determine what action to be taken on a slide, based on slide scores and slide preparation type, is shown in Table B. For example, if in step 1, the Eval slide score is greater than or equal to a quality control threshold, a QC threshold, then the system classifies the slide into a quality control review category. The QC threshold, and other thresholds, may change depending on slide preparation type.

TABLE B SLIDE ACTION CLASSIFICATION

Those skilled in the art will recognize that the No Review population cannot exceed the classification rate. If the application of thresholds results in a higher percentage of No Review slides than the classification rate, the Review population is supplemented with the highest scoring slides, Eval score, from the No Review population until the No Review population is less than or equal to the classification rate. The other parameters will be discussed in order to show how determining slide preparation type will effect slide processing.

The Eval parameter represents the evaluation score. The Eval parameter is a slide score that describes the likelihood that the slide will contain abnormality. The generation of this score and other scores in Table B is disclosed in more detail in assignee's United States Patents and co- pending United States Patent applications. The QC parameter is the quality control threshold for the Eval score. This parameter is set to achieve a predetermined quality control review classification rate based on a calibration slide set for a particular slide preparation type. If the Eval score equals or exceeds the quality control threshold, QC threshold, the slide is classified into the quality control review category also known as the QC Review category. During the quality control review, a cytotech or pathologist manually reviews the slide in a well known manner.

In step 2, the Adj_Eval score is evaluated. The Adj_Eval score is an adjunctive evaluation parameter. The Adj_Eval parameter is a slide score that describes the likelihood that the slide will contain cells indicating glandular abnormality. If the Adj_Eval score equals or exceeds the adjunctive quality control threshold, Adj_QC threshold, the slide is classified into the quality control review category.

In step 3, the Eval score is evaluated. If the Eval score equals or exceeds the screening threshold, Screen threshold, the slide is classified into the Review category. The Screen threshold is the screening threshold for the Eval score. The Screen parameter is set to achieve a predetermined screening review classification rate based on a calibration slide set for a particular slide preparation type. During the review, a cytotech or pathologist manually reviews the slide in a well-known manner.

In step 4, the Adj_Eval score is compared against the adjunctive screening threshold, Adj_Screen threshold. If the Adj_Eval score is greater than or equal to the Adj_Screen threshold, then the slide is classified in the Review category. The Adj_Screen threshold may be changed based on slide preparation type.

In step 5, the Adeq score is compared against the Adeq_Hi threshold. If Adeq score is greater than the Adeq_Hi threshold, then the slide is classified into the Review category. The Adeq_Hi threshold may change based on the slide preparation type.

In step 6, the Squamous score, Squam, is compared against the Squamous threshold. If the squamous score is less than or equal to the Squamous threshold, then the slide is classified into the Review category. The Squamous threshold may change based on the slide preparation type. In step 7, the Adeq score is compared against the Adeq_Lo threshold. If Adeq score is greater than or equal to the Adeq_Lo threshold, then the slide is classified into the No Review category. The Adeq_Lo threshold may change based on the slide preparation type.

In step 8, the Endocervical score, Endocx, compared against the Endocervical threshold, Endocx. If Endocx score is less than or equal to the Endocx threshold, then the slide is classified into the No Review category. The Endocx threshold may change based on the slide preparation type.

In step 9, the Adjunctive Endocervical score, Adj_Endocx, is compared against the Endocervical threshold, Adj_Endocx. If the Adj_Endocx score is less than or equal to the Endocx threshold, then the slide is classified into the No Review category. The Adj_Endocx threshold may change based on the slide preparation type.

Otherwise the slide is classified into the No Review category.

Table C INFLAMMATION/OBSCURATION CLASSIFICATION

Table C shows the Inflammation/Obscuration Classification logic. If the adequacy score, Adeq, is greater than the adequacy high threshold, adeq_hi, then the slide is identified as having inflammation/obscuration on more than 75% of the specimen area. Otherwise, if the adequacy score, Adeq, is greater than or equal to the adequacy low threshold, adeq_lo, the slide is identified as having 50-75% of the specimen area obscured. Otherwise, the slide is identified as having less than 50% inflammation/obscuration. The adeq_hi and adeq_lo thresholds may vary based on the slide preparation type.

TABLE D: SQUAMOUS CLASSIFICATION

Table D shows the Squamous Classification logic. If the Squamous slide score is greater than the squamous threshold, then squamous cells are detected. Otherwise, squamous cells are not detected. The squamous threshold varies based on the slide preparation type.

TABLE E ENDOCERVICAL CLASSIFICATION

Table E shows the Endocervical Classification logic. If the Endocervical slide score is less than or equal to the endocervical threshold, then endocervical cells have not detected. If the adjunctive endocervical score is greater than or equal to the adjunctive endocervical threshold, then endocervical cells have not detected.

Otherwise, endocervical cells have detected. The endocervical threshold and adjunctive threshold vary based on the slide preparation type.

TABLE F ADEQUACY CLASSIFICATION

Table F shows the Adequacy Classification logic. Depending on the state of the squamous, endocervical and inflammation/obscuration score, the adequacy score is determined.

The central processor 540 implements the processes described in Figure 16 in software.

During a slide processing run, a cytology laboratory, such as a laboratory running Pap smear slides, follows a setup procedure. These procedures are disclosed in assignee's co-pending United States Patent applications and issued United States Patents. As part of the setup procedure, the determination of slide preparation type may be performed. Thus labs may run different types of slides either in batches on different days or on batches on the same day. Labs may also elect to review differing slide preparation types from run to run as outlined herein. During the setup procedure, a representative number of normal slides are run, for example, 240 slides. These normal slides are then used to set the normal threshold to achieve a desired sort rate, based additionally on the slide preparation type. The invention may also be used to verify that a slide is indeed a designated slide preparation type. This verification step may be performed during a machine setup phase for example. This test may also be performed to test whether a foreign slide preparation type has been intermingled in a batch. Error warnings may be provided to indicate that a slide's preparation type has been automatically determined to be inconsistent with other slides in a batch or is inconsistent with an expected slide preparation type.

Table G shows examples of thresholds from one embodiment of the invention. This table demonstrates the differences in threshold by preparation type.

TABLE G THRESHOLD VALUES

conventional : adeq_hi 0.9601 conventional : adeq_lo 0.7614 conventional : adj_endocx 0.1775 conventional :adj_qc 0.742096 conventional :adj_screen 0.334177 conventional : endocx 26.66 conventional :qc 0.867010 conventional : screen 0.109055 conventional : sort_rate 0.25 conventional : squam 4252.8

THINPREP : adeq_hi 0.5922

THINPREP : adeq_lo 0.0367

THINPREP : adj_endocx 0.1009

THINPREP: adj_qc 0.972840

THINPREP :adj_screen 0.474872

THINPREP: endocx 23.53

THINPREP :qc 0.834857

THINPREP: screen 0.249146

THINPREP : sort_rate 0.25

THINPREP : squam 1787.7

Figures 17A-17C show reports that are dependent on slide preparation type. An example of each report is attached. One difference in the report appearance between preparation types is that the slide preparation type is printed on the top of each report . The automated biological specimen analysis system of the invention provides on-line, electronic and printed reports to aid subsequent review by laboratory technicians, cytologists and pathologists . The reports will differ based on slide preparation type and the features on the reports that are dependent on specimen preparation type. Three report types are produced, the slide results report, the Pap Map report, and the process review report. The slide results report shows the review action and indications for each slide in the slide run. The report contains slides from only one specimen preparation type. The specimen preparation type appears in the report header. All actions and indications are based on thresholds specific to the specimen preparation type such as the action to be taken, Review, QC Review or No Further Review, the Squamous adequacy, the Endocervical adequacy, and the Inflammation and Obscuration adequacy. The Pap Map report shows the review action and indications in a pictorial representation of each slide with circles indicating potential areas of interest within the specimen. The report contains slides from only one specimen preparation type. The specimen preparation type appears in the report header. The number of circles is dependent on specimen preparation type. All actions and indications are based on thresholds specific to the specimen preparation type such as the action to be taken: Review, QC Review or No Further Review, the Squamous adequacy, the Endocervical adequacy, and the Inflammation and Obscuration adequacy. The process review report shows slides that have a processing problem. The report contains slides from only one specimen preparation type where the specimen preparation type appears in the report header. A digest of each report is given below.

The following United States Patents and Patent Applications are incorporated by reference hereto:

U.S. Patent No. 5,315,700, issued 05/24/94 to Johnston et al . , entitled METHOD AND APPARATUS FOR RAPIDLY PROCESSING DATA SEQUENCES;

U.S. Patent No. 5,361,140, issued 11/01/94, to Hayenga et al . , entitled METHOD AND APPARATUS FOR DYNAMIC CORRECTION OF MICROSCOPIC IMAGE SIGNALS;

U.S. Patent No. 5,699,794, issued 12/23/97, to Fleck, entitled APPARATUS FOR AUTOMATED URINE SEDIMENT SAMPLE HANDLING;

U.S. Patent No. 5,528,703, issued 06/18/96, to Lee, entitled FWC: METHOD FOR IDENTIFYING OBJECTS USING DATA PROCESSING TECHNIQUES, which is a file wrapper continuation of abandoned U.S. Patent Application Serial No. 07/838,395, filed 02/18/92; pending U.S. Patent Application Serial No. 08/485,182, filed 06/07/95, to Lee et al . , entitled INTERACTIVE METHOD AND APPARATUS FOR SORTING BIOLOGICAL SPECIMENS;

U.S. Patent No. 5,647,025, issued 07/08/97, to Frost et al . , entitled AUTOMATIC FOCUSING OF BIOMEDICAL SPECIMENS APPARATUS; allowed U.S. Patent Application Serial No.

08/302,355, filed 09/07/94, for which the issue fee has been paid, to Hayenga et al . , entitled CIP: METHOD AND APPARATUS FOR RAPID CAPTURE OF

FOCUSED MICROSCOPIC IMAGES, which is a continuation-in-part of abandoned U.S. Patent

Application Serial No. 07/838,063, filed 02/18/92;

U.S. Patent No. 5,715,326, issued 02/03/98, to Ortyn et al . , entitled CYTOLOGICAL SYSTEM

ILLUMINATION INTEGRITY CHECKING APPARATUS AND METHOD ;

U.S. Patent No. 5,581,631, issued 12/03/96, to Ortyn et al . , entitled CYTOLOGICAL SYSTEM IMAGE COLLECTION INTEGRITY CHECKING APPARATUS; U.S. Patent No. 5,557,097, issued 09/17/96, to Ortyn et al . , entitled CYTOLOGICAL SYSTEM AUTOFOCUS INTEGRITY CHECKING APPARATUS;

U.S. Patent No. 5,499,097, issued 03/12/96, to Ortyn et al . , entitled METHOD AND APPARATUS FOR CHECKING AUTOMATED OPTICAL SYSTEM PERFORMANCE REPEATABILITY;

U.S. Patent No. 5,757,954, issued 05/26/98, to Kuan et al., entitled FIELD PRIORITIZATION APPARATUS AND METHOD; U.S. Patent No. 5,627,908, issued 05/06/97, to Lee et al . , entitled METHOD FOR CYTOLOGICAL SYSTEM DYNAMIC NORMALIZATION;

U.S. Patent No. 5,638,459, issued 06/10/97, to Rosenlof et al . , entitled METHOD AND APPARATUS FOR DETECTING A MICROSCOPE SLIDE COVERSLIP;

U.S. Patent No. 5,566,249, issued 10/15/96, to Rosenlof et al . , entitled APPARATUS FOR DETECTING BUBBLES IN COVERSLIP ADHESIVE; allowed U.S. Patent Application Serial No. 08/309,250, filed 09/20/94, for which the issue fee has been paid, to Lee et al., entitled

APPARATUS FOR THE IDENTIFICATION OF FREE-LYING

CELLS ;

U.S. Patent No. 5,740,269, issued 04/14/98, to Oh et al., entitled A METHOD AND APPARATUS FOR ROBUST BIOLOGICAL SPECIMEN CLASSIFICATION;

U.S. Patent No. 5,715,327, issued 02/03/98, to Wilhelm et al . , entitled METHOD AND APPARATUS FOR DETECTION OF UNSUITABLE CONDITIONS FOR AUTOMATED CYTOLOGY SCORING;

U.S. Patent No. 5,692,066, issued 11/25/97, to Lee et al . , entitled METHOD AND APPARATUS FOR IMAGE PLANE MODULATION PATTERN RECOGNITION;

U.S. Patent No. 5,671,288, issued 09/23/97, to Wilhelm et al . , entitled METHOD AND APPARATUS FOR ASSESSING SLIDE AND SPECIMEN PREPARATION QUALITY;

U.S. Patent No. 5,619,428, issued 04/08/97, to Lee et al . , entitled METHOD AND APPARATUS FOR INTEGRATING AN AUTOMATED SYSTEM TO A LABORATORY; U.S. Patent No. 5,621,519, issued 04/15/97, to Frost et al., entitled IMAGING SYSTEM TRANSFER FUNCTION CONTROL METHOD AND APPARATUS;

U.S. Patent No. 5,642,441, issued 06/24/97, to Riley et al . , entitled APPARATUS AND METHOD FOR MEASURING FOCAL PLANE SEPARATION;

U.S. Patent No. 5,787,208, issued 07/28/98, to Oh et al., entitled IMAGE ENHANCEMENT METHOD AND APPARATUS; pending U.S. Patent Application Serial No. 08/924,351, filed 09/05/97, to Kuan et al . , entitled DYNAMIC CONTROL AND DECISION MAKING METHOD AND APPARATUS;

U.S. Patent No. 5,625,706, issued 04/29/97, to Lee et al . , entitled METHOD AND APPARATUS FOR CONTINUOUSLY MONITORING AND FORECASTING SLIDE AND SPECIMEN PREPARATION FOR A BIOLOGICAL SPECIMEN POPULATION;

U.S. Patent No. 5,745,601, issued 04/28/98, to Lee et al . , entitled ROBUSTNESS OF CLASSIFICATION MEASUREMENT APPARATUS AND METHOD;

U.S. Patent No. 5,781,667, issued 07/14/98, to Schmidt et al . , entitled APPARATUS FOR HIGH SPEED MORPHOLOGICAL PROCESSING;

U.S. Patent No. 5,642,433, issued 06/24/97, to Lee et al . , entitled METHOD AND APPARATUS FOR IMAGE CONTRAST QUALITY EVALUATION; U.S. Patent No. 5,797,130, issued 08/18/98, to Nelson et al . , entitled FWC : METHOD FOR TESTING PROFICIENCY IN SCREENING IMAGES OF BIOLOGICAL SLIDES, which is a file wrapper continuation of abandoned U.S. Patent Application Serial No. 08/153,293, filed 11/16/93;

U.S. Patent No. 5,787,188, issued 07/28/98, to Nelson et al., entitled FWC: METHOD FOR IDENTIFYING NORMAL BIOMEDICAL SPECIMENS, which is a file wrapper continuation of abandoned U.S. Patent Application Serial No. 07/838,064, filed 02/18/92, to Nelson et al . ;

U.S. Patent No. 5,710,842, issued 01/20/98, to Lee entitled DIV: METHOD FOR IDENTIFYING OBJECTS USING DATA PROCESSING TECHNIQUES, which is a divisional of U.S. Patent No. 5,528,703, ibid., which is a file wrapper continuation of abandoned U.S. Patent Application Serial No. 07/838,395, filed 02/18/92; allowed U.S. Patent Application Serial No. 08/788,239, for which the issue fee has been paid, filed 01/25/97, to Oh et al . , entitled METHOD AND APPARATUS FOR ALIAS FREE MEASUREMENT OF OPTICAL TRANSFER FUNCTION; U.S. Patent No. 5,677,762, issued 10/14/97, to Ortyn et al . , entitled FWC: APPARATUS FOR ILLUMINATION STABILIZATION AND HOMOGENIZATION, which is a file wrapper continuation of abandoned U.S. Patent Application Serial No. 08/309,064, filed 09/20/94;

U.S. Patent No. 5,787,189, issued 07/28/98, to Lee et al . , entitled FWC: BIOLOGICAL ANALYSIS SYSTEM SELF CALIBRATION APPARATUS, which is a file wrapper continuation of abandoned U.S. Patent Application Serial No. 08/309,115, filed 09/20/94; U.S. Patent No. 5,654,535, issued 08/05/97, to Ortyn et al . , entitled DIV: CYTOLOGICAL SYSTEM AUTOFOCUS INTEGRITY CHECKING APPARATUS, which is a divisional of U.S. Patent No. 5,557,097, ibid.;

U.S. Patent No. 5,828,776, issued 10/27/98, to Lee et al . , entitled FWC: APPARATUS FOR IDENTIFICATION AND INTEGRATION OF MULTIPLE CELL PATTERNS, which is a file wrapper continuation of abandoned U.S. Patent Application Serial No. 08/308,992, filed 09/20/94, to Lee et al . ; pending U.S. Patent Application Serial No. 09/121,012, filed 07/22/98, to Lee et al . , entitled DIV: APPARATUS FOR THE IDENTIFICATION OF FREE-LYING CELLS, which is a divisional of allowed U.S. Patent Application Serial No. 08/309,250, ibid.; pending U.S. Patent Application Serial No. 09/120,860, filed 07/22/98, to Lee et al . , entitled DIV: APPARATUS FOR THE IDENTIFICATION OF FREE-LYING CELLS, which is a divisional of allowed U.S. Patent Application Serial No. 08/309,250, ibid. ; pending U.S. Patent Application Serial No. 09/120,612, filed 07/22/98, to Lee et al . , entitled DIV: APPARATUS FOR THE IDENTIFICATION OF FREE-LYING CELLS, which is a divisional of allowed U.S. Patent Application Serial No. 08/309,250, ibid. ;

U.S. Patent No. 5,875,258, issued 2/23/99, to Ortyn et al . , entitled FWC: BIOLOGICAL SPECIMEN ANALYSIS SYSTEM PROCESSING INTEGRITY CHECKING APPARATUS, which is a file wrapper continuation of abandoned U.S. Patent Application Serial No. 08/309,249, filed 09/20/94;

U.S. Patent No. 5,812,692, issued 09/22/98, to Rosenlof et al . , entitled DIV: METHOD AND APPARATUS FOR DETECTING A MICROSCOPE SLIDE COVERSLIP, which is a divisional of U.S. Patent No. 5,638,459, ibid.; pending U.S. Patent Application Serial No. 08/767,457, filed 12/16/96, to Lee et al . , entitled METHOD AND APPARATUS FOR EFFICACY IMPROVEMENT IN MANAGEMENT OF CASES WITH EQUIVOCAL SCREENING RESULTS; pending U.S. Patent Application Serial No. 08/877,368, filed 06/17/97, to Lee et al . , entitled DIV: METHOD AND APPARATUS FOR IMAGE PLANE MODULATION PATTERN RECOGNITION, which is a divisional of U.S. Patent No. 5,692,066, ibid.;

U.S. Patent No. 5,892,218, issued 4/6/99, to Ortyn et al . , entitled DIV: CYTOLOGICAL SYSTEM AUTOFOCUS INTEGRITY CHECKING APPARATUS, which is a divisional of U.S. Patent No. 5,654,535, ibid., which is a divisional of U.S. Patent No. 5,557,097, ibid.;

U.S. Patent No. 5,760,387, issued 06/02/98, to Ortyn et al . , entitled DIV: CYTOLOGICAL SYSTEM AUTOFOCUS INTEGRITY CHECKING APPARATUS, which is a divisional of U.S. Patent No. 5,654,535, ibid., which is a divisional of U.S. Patent No. 5,557,097, ibid.; U.S. Patent No. 5,841,124, issued 11/24/98, to Ortyn et al . , entitled DIV: CYTOLOGICAL SYSTEM AUTOFOCUS INTEGRITY CHECKING APPARATUS, which is a divisional of U.S. Patent No. 5,654,535, ibid., which is a divisional of U.S. Patent No. 5,557,097, ibid.;

U.S. Patent No. 5,763,871, issued 06/09/98, to Ortyn et al . , entitled DIV: CYTOLOGICAL SYSTEM AUTOFOCUS INTEGRITY CHECKING APPARATUS, which is a divisional of U.S. Patent No. 5,654,535, ibid., which is a divisional of U.S. Patent No. 5,557,097, ibid.; U.S. Patent No. 5,877,489, issued 3/2/99, to Ortyn et al . , entitled DIV: CYTOLOGICAL SYSTEM AUTOFOCUS INTEGRITY CHECKING APPARATUS, which is a divisional of U.S. Patent No. 5,654,535, ibid., which is a divisional of U.S. Patent No. 5,557,097, ibid.;

U.S. Patent No. 5,883,982, issued 3/16/99, to

Riley et al . , entitled DIV: ASTIGMATISM

MEASUREMENT APPARATUS AND METHOD BASED ON A FOCAL PLANE SEPARATION, which is a divisional of U.S.

Patent No. 5,642,441, ibid.; pending U.S. Patent Application Serial No. 08/900,341, filed 07/25/97, to Riley et al . , entitled MODULATION TRANSFER FUNCTION TEST COMPENSATION FOR TEST PATTERN DUTY CYCLE; pending U.S. Patent Application Serial No.

08/888,115, filed 07/03/97, to Lee et al . , entitled METHOD AND APPARATUS FOR MASKLESS

SEMICONDUCTOR AND LIQUID CRYSTAL DISPLAY INSPECTION; pending U.S. Patent Application Serial No.

08/888,120, filed 07/03/97, to Lee et al . , entitled METHOD AND APPARATUS FOR A REDUCED

INSTRUCTION SET ARCHITECTURE FOR MULTIDIMENSIONAL IMAGE PROCESSING; pending U.S. Patent Application Serial No. 08/888,119, filed 07/03/97, to Lee et al . , entitled METHOD AND APPARATUS FOR INCREMENTAL CONCURRENT LEARNING IN AUTOMATIC SEMICONDUCTOR WAFER AND LIQUID CRYSTAL DISPLAY DEFECT CLASSIFICATION; pending U.S. Patent Application Serial No. 08/888,116, filed 07/03/97, to Lee et al . , entitled METHOD AND APPARATUS FOR SEMICONDUCTOR WAFER AND LCD INSPECTION USING MULTIDIMENSIONAL IMAGE DECOMPOSITION AND SYNTHESIS; pending U.S. Patent Application Serial No.

09/006,457, filed 01/13/98, to Kuan et al . , entitled A METHOD AND APPARATUS FOR OPTIMIZING

BIOLOGICAL AND CYTOLOGICAL SPECIMEN SCREENING AND DIAGNOSIS; allowed U.S. Patent Application Serial No. 08/867,017, filed 06/03/97, for which the issue fee has been paid, to Lee et al . , entitled FWC: CYTOLOGICAL SLIDE SCORING APPARATUS, which is a file wrapper continuation of abandoned U.S. Patent Application Serial No. 08/309,931, filed 09/20/94; U.S. Patent No. 5,587,833, issued 12/24/96, to Kamentsky entitled COMPUTERIZED MICROSCOPE SPECIMEN ENCODER; U.S. Patent No. 5,602,674, issued 02/11/97, to Weissman et al . , entitled COMPUTERIZED SPECIMEN ENCODER;

U.S. Patent No. 5,561,556, issued 10/01/96, to Weissman entitled SLIDE ANALYSIS SYSTEM WITH SLIDE HAVING SELF CONTAINED MICROSCOPE ANALYSIS INFORMATION;

U.S. Patent No. 5,793,969, issued 08/11/98, to Kamentsky et al . , entitled NETWORK REVIEW AND ANALYSIS OF COMPUTER ENCODER SLIDES; U.S. Patent No. 5,790,308, issued 08/04/98, to Kamentsky entitled COMPUTERIZED MICROSCOPE SPECIMEN ENCODER;

U.S. Patent No. 5,694,212, issued 12/02/97, to Weissman, entitled METHOD FOR CALIBRATING SPECIMEN WITH SPECIMEN HOLDER OF A MICROSCOPE;

U.S. Patent No. 5,581,487, issued 12/03/96, to Kelly entitled METHOD AND APPARATUS FOR MICROSCOPIC SCREENING OF CYTOLOGICAL SAMPLES;

U.S. Patent No. 5,867,610, issued 2/2/99, to Lee entitled DIV: METHOD FOR IDENTIFYING OBJECTS

USING DATA PROCESSING TECHNIQUES, which is a divisional of U.S. Patent No. 5,710,842, ibid.;

U.S. Patent No. 5,799,101, issued 08/25/98, to Lee et al . , entitled FWC: METHOD AND APPARATUS FOR HIGHLY EFFICIENT COMPUTER AIDED SCREENING, which is a file wrapper continuation of abandoned U.S. Patent Application Serial No. 08/315,719, filed 09/30/94; pending U.S. Patent Application Serial No.

08/912,061, filed 08/15/97, to Ortyn et al . , entitled DIV: CYTOLOGICAL SYSTEM ILLUMINATION

INTEGRITY CHECKING APPARATUS AND METHOD, which is a divisional of U.S. Patent No. 5,715,326, ibid.; pending U.S. Patent Application Serial No. 08/911,807, filed 08/15/97, to Ortyn et al . , entitled DIV: CYTOLOGICAL SYSTEM ILLUMINATION INTEGRITY CHECKING APPARATUS AND METHOD, which is a divisional of U.S. Patent No. 5,715,326, ibid.; pending U.S. Patent Application Serial No. 08/911,612, filed 08/15/97, to Ortyn et al . , entitled DIV: CYTOLOGICAL SYSTEM ILLUMINATION INTEGRITY CHECKING APPARATUS AND METHOD, which is a divisional of U.S. Patent No. 5,715,326, ibid.; pending U.S. Patent Application Serial No.

08/912,115, filed 08/15/97, to Ortyn et al . , entitled DIV: CYTOLOGICAL SYSTEM ILLUMINATION

INTEGRITY CHECKING APPARATUS AND METHOD, which is a divisional of U.S. Patent No. 5,715,326, ibid.; pending U.S. Patent Application Serial No. 08/911,644, filed 08/15/97, to Ortyn et al . , entitled DIV: CYTOLOGICAL SYSTEM ILLUMINATION INTEGRITY CHECKING APPARATUS AND METHOD, which is a divisional of U.S. Patent No. 5,715,326, ibid.; pending U.S. Patent Application Serial No.

08/911,611, filed 08/15/97, to Ortyn et al . , entitled DIV: CYTOLOGICAL SYSTEM ILLUMINATION

INTEGRITY CHECKING APPARATUS AND METHOD, which is a divisional of U.S. Patent No. 5,715,326, ibid.; pending U.S. Patent Application Serial No. 08/969,970, filed 11/13/97, to Meyer et al . , entitled FWC: APPARATUS FOR AUTOMATED IDENTIFICATION OF THICK CELL GROUPINGS ON A BIOLOGICAL SPECIMEN, which is a file wrapper continuation of abandoned U.S. Patent Application Serial No. 08/309,116, filed 09/20/94;

A Continued Prosecution Application entitled CPA: METHOD AND APPARATUS FOR DETECTION OF UNSUITABLE CONDITIONS FOR AUTOMATED CYTOLOGY SCORING of prior U.S. Patent Application Serial No. 08/914,292, filed 08/18/97, to Wilhelm et al . , which is a divisional of U.S. Patent No. 5,715,327, ibid.; pending U.S. Patent Application Serial No. 08/927,379, filed 09/12/97, to Wilhelm et al . , entitled FWC: APPARATUS FOR AUTOMATED IDENTIFICATION OF CELL GROUPINGS ON A BIOLOGICAL SPECIMEN, which is a file wrapper continuation of abandoned U.S. Patent Application Serial No. 08/309,061, filed 09/20/94; pending U.S. Patent Application Serial No.

08/970,904, filed 11/14/97, to Weissman et al . , entitled FWC: FOLDING SLIDE HOLDER, which is a file wrapper continuation of abandoned U.S. Patent

Application Serial No. 08/582,495, filed 01/03/96; pending U.S. Patent Application Serial No. 09/014,984, filed 01/28/98, to Ellison et al . , entitled METHOD AND APPARATUS FOR RANKED REVIEW OF BIOLOGICAL SPECIMENS;

U.S. Patent No. 5,862,265, issued 1/19/99, to

Riley et al . , entitled DIV: ASTIGMATISM

MEASUREMENT APPARATUS AND METHOD, which is a divisional of U.S. Patent No. 5,883,982, ibid., which is a divisional of U.S. Patent No. 5 , 642 , 441 , ibid . ; pending U.S. Patent Application Serial No.

09/082,580, filed 05/21/98, to Kuan et al . , entitled DIV: FIELD PRIORITIZATION APPARATUS AND METHOD, which is a divisional of U.S. Patent No.

5,757,954, ibid.; and pending U.S. Patent Application Attorney Docket No. 74B/2248, filed on even date herewith, to Boisseranc et al . , entitled METHOD AND APPARATUS FOR DETERMINING MICROSCOPE SPECIMEN PREPARATION TYPE.

The invention has been described herein in considerable detail in order to comply with the Patent Statutes and to provide those skilled in the art with the information needed to apply the novel principles and to construct and use such specialized components as are required. However, it is to be understood that the invention can be carried out by specifically different equipment and devices, and that various modifications, both as to the equipment details and operating procedures, can be accomplished without departing from the scope of the invention itself. What is claimed is:

Claims

1. A method for determining a slide preparation type of a biological specimen slide comprising the steps of : a) obtaining at least one image (14) of the biological specimen slide to provide a digital representation (15) of the at least one image; and b) determining the slide preparation type from the digital representation (24) .

2. The method of claim 1 wherein the digital representation further includes texture information, the method further comprising the steps of : a) processing the digital representation

(23) to obtain at least one texture feature (28) ; and b) determining the slide preparation type from the at least one texture feature

(24) .

3. The method of claim 2 wherein the at least one texture feature (28) is mathematically derived from the digital representation (15) .

4. The method of claim 3 wherein the digital representation (15) further comprises image pixel values, and wherein the at least one texture feature (28) is the mean of image pixel values (42) .

5. The method of claim 3 wherein the digital representation (15) further comprises image pixel values, and wherein the at least one texture feature (28) is the standard deviation of image pixel values (44) .

6. The method of claim 3 wherein the digital representation (15) further comprises image pixel values, and wherein the at least one texture feature (28) is the skewness of image pixel values (46) .

7. The method of claim 3 wherein the digital representation (15) further comprises image pixel values, and wherein the at least one texture feature (28) is the kurtosis of image pixel values (48) .

8. The method of claim 2 further comprising the step of convolving the digital representation (15) with a kernel to generate a convolved image representation, and wherein the at least one texture feature (28) is mathematically derived from the convolved image representation.

9. The method of claim 8 wherein the kernel is selected from the group consisting of a 3x3 kernel (370), a 5x5 kernel, or a 15x15 kernel (392) .

10. The method of claim 2 wherein the step of determining the slide preparation type (24) from the at least one texture feature (28) further comprises comparing the at least one texture feature (28) against a threshold (57) .

11. The method of claim 2 wherein the step of determining the slide preparation type (20) from the at least one texture feature (28) further comprises classifying the at least one texture feature (28) with a feature classifier.

12. The method of claim 11 wherein the feature classifier (52) is a classifier.

13. The method of claim 11 wherein the feature classifier (52) is a Fisher's linear discriminant classifier.

14. The method of claim 1 wherein the digital representation (15) is further processed to obtain a plurality of biological specimen scores (86) , wherein each one of the plurality of biological specimen scores (86) has a geometric position relative to the biological specimen slide (12) , the method further comprising the steps of: a) processing the plurality of biological specimen scores and geometric positions (27) to obtain at least one geometric feature (22) ; and b) determining the slide preparation type from the at least one geometric feature (24) .

15. The method of claim 14 wherein the at least one image further includes a plurality of fields of view wherein each one of the plurality of fields of view has a corresponding field of view position, the method further comprising the ^"steps of: a) processing the plurality of fields of view to obtain field of view scores (86) ; and b) determining the slide preparation type (20) from the field of view scores (86) and the plurality of corresponding field of view positions.

16. The method of claim 15 wherein the step of determining the slide preparation type (24) from the field of view scores (86) further comprises matching a predefined pattern associated with a slide preparation type to a pattern formed from the field of view scores.

17. The method of claim 16 further comprising processing the field of view scores with a spatial filter (88) .

18. The method of claim 14 wherein the step of determining the slide preparation type (20) from the at least one geometric feature (22) further comprises the step of removing at least one noisy biological specimen score from consideration (88) .

19. The method of claim 18 wherein the at least one noisy biological specimen score correspond to dust, fiducial marks, scratches, or dark regions on the biological specimen slide (12) .

20. The method of claim 18 wherein the step of removing at least one noisy biological specimen score (88) comprises the steps of arranging the biological specimen scores into an array of scores (86) and creating a mask (92) corresponding to noisy scores, and combining the mask and the array to generate a filtered score array (96) .

21. The method of claim 14 wherein the step of determining the slide preparation type (20) from the at least one geometric feature (27) further comprises the step of locating a position on the biological specimen slide that substantially represents the center of material on the biological specimen slide (98) .

22. The method of claim 21 wherein the biological specimen scores are arranged as an array (86) corresponding to positions on the biological specimen slide (12) , wherein the step of locating a position on the biological specimen slide that substantially represents the center of material on the biological specimen slide (98) comprises the steps of: a) summing all values of y for each x to generate an x histogram; b) summing all values of x for each y to generate a y histogram; c) computing the weighted average for x from the x histogram; d) computing the weighted average for y from the y histogram; and e) calculating the position on the biological specimen slide that substantially represents the center of material (150) on the biological specimen slide from the weighted average for x and the weighted average for y.

23. The method of claim 22 wherein the step of determining the slide preparation type (20) from the biological specimen scores (86) further comprises the step of totaling a number of non- zero fields of view scores in a predetermined pattern around a geometric center (30) , wherein the predetermined pattern corresponds to a slide preparation type (20) .

24. The method of claim 23 wherein the predetermined pattern is a generally circular pattern (322, 326) .

25. The method of claim 24 further comprising the step of computing a percentage of non-zero fields of view to a total number of fields of view (108) .

26. The method of claim 25 further comprising the step of comparing the percentage to a predetermined threshold associated with a predetermined slide preparation type (110) .

27. The method of claim 26 further comprising the step of determining the slide preparation type (20) of the biological specimen slide (12) to be the predetermined slide preparation type if the percentage exceeds the predetermined threshold.

28. The method of claim 14 further comprising the step of generating a binary homologous image (124) .

29. The method of claim 14 wherein the step of determining the slide preparation type (20) from the at least one geometric feature further comprises the steps of: a) creating a score array of field of view scores (86) ; b) creating a mask array corresponding to the score array (118) ; c) neighborhood processing the mask array for corresponding elements of the score array to count (160) a first number of neighbors with a predetermined field of view score in the score array (158); d) assigning a mask value to elements in the mask array that correspond to neighbors in the score array if the first number of neighbors exceeds a first predetermined amount (176) ; e) neighborhood processing the mask array for elements of the score array to count a second number of neighbors with a predetermined field of view score in the score array; and f) assigning a mask value to a corresponding center element of the mask array if the second number of neighbors exceeds a second predetermined amount .

30. The method of claim 29 further comprising the step of dilating (95) an element of the mask array that is set to the mask value.

31. The method of claim 29 wherein the step of dilating an element further comprises a 3x3 plus dilation (Fig.HC).

32. The method of claim 29 wherein the step of neighborhood processing further comprising using a 3x5 neighborhood (154) .

33. The method of claim 29 wherein the step of neighborhood processing further comprises using a 3x3 neighborhood (182) .

34. The method of claim 1 wherein the at least one image (14) comprises a single image from a high-resolution camera.

35. The method of claim 1 wherein the at least one image (14) comprises a plurality of subimages .

36. The method of claim 1 further comprising the step of setting an alarm if the slide preparation type is not an expected slide preparation type.

37. An apparatus for determining slide preparation type (20) of a biological specimen slide (12) comprising: a) means (502) for obtaining at least one image of the biological specimen slide (12) to obtain a digital representation

(15) of the at least one image; and b) means (540) for determining the slide preparation type (20) from the digital representation (15) .

38. The apparatus of claim 37 wherein the digital representation (15) further includes texture information, the apparatus further comprising: a) means for processing (540) the digital representation (15) to obtain at least one texture feature; and b) means for determining the slide preparation type (20) from the at least one texture feature (28) .

39. The apparatus of claim 38 wherein the at least one texture feature (28) is mathematically derived from the digital representation (15) .

40. The apparatus of claim 39 wherein the digital representation (15) further comprises image pixel values, and wherein the at least one texture feature (28) is the mean (42) of image pixel values .

41. The apparatus of claim 39 wherein the digital representation (15) further comprises image pixel values, and wherein the at least one texture feature (28) is the standard deviation (44) of image pixel values.

42. The apparatus of claim 39 wherein the digital representation (15) further comprises image pixel values, and wherein the at least one texture feature (28) is the skewness (46) of image pixel values.

43. The apparatus of claim 39 wherein the digital representation (15) further comprises image pixel values, and wherein the at least one texture feature (28) is the kurtosis (48) of image pixel values.

44. The apparatus of claim 38 further comprising a means for convolving the digital representation (15) with a kernel to generate a convolved image representation, and wherein the at least one texture feature (28) is statistically derived from the convolved image representation.

45. The apparatus of claim 44 wherein the kernel is selected from the group consisting of a 3x3 kernel (370) , a 5x5 kernel, or a 15x15 kernel (392) .

46. The apparatus of claim 38 further comprising a means for comparing the at least one texture feature (28) against a threshold.

47. The apparatus of claim 38 further comprising a means for classifying the at least one texture feature (28) with a feature classifier.

48. The apparatus of claim 47 wherein the feature classifier (52) is a statistical classifier.

49. The apparatus of claim 47 wherein the feature classifier (52) is a Fisher's linear discriminate classifier.

50. The apparatus of claim 37 wherein the digital representation (15) is further processed to obtain a plurality of biological specimen scores (86) , wherein each one of the plurality of biological specimen scores (86) has a geometric position relative to the biological specimen slide (12) , wherein the apparatus further comprises : a) means for processing (540) the plurality of biological specimen scores (86) and geometric positions to obtain at least one geometric feature (22); and b) means for determining the slide preparation type (20) from the at least one geometric feature.

51. The apparatus of claim 50 wherein the plurality of biological specimen scores (86) further comprises field of view scores.

52. The apparatus of claim 50 further comprising a means for determining a center (98) of the plurality of biological specimen scores (86) , responsive to the digital representation (15) to determine a geometric center, wherein the means for determining a center of the plurality of biological specimen scores (86) has a center output (150) .

53. The apparatus of claim 52 further comprising a means for determining a percentage of material that falls within an expected pattern (108) having a percentage output.

54. The apparatus of claim 53 further comprising a means for comparing (30) the percentage output against a predetermined percentage threshold, wherein the predetermined percentage threshold has an associated slide preparation type, to automatically determine a slide preparation type, responsive to the percentage output, wherein the means for comparing the percentage output has a slide preparation type output wherein, if the percentage output exceeds the predetermined percentage threshold, the slide preparation type is determined to be the associated slide preparation type.

55. The apparatus of claim 50 wherein the digital representation further comprises a plurality of fields of view wherein a field of view score is generated for each field of view and the field of view scores are stored in a score array of field of view scores corresponding to a location of the field of view on the biological specimen slide, the apparatus further comprising: a) a means for generating (92) a mask array (94) from the score array (86) ; and b) means for combining (88) the mask array

(94) with the score array (86) to generate a filtered score array (96) .

56. The apparatus of claim 55 wherein the means for combining further comprises an exclusive OR function (88) .

57. The apparatus of claim 55 further comprising a means for removing field of view scores corresponding to dust and fiducial marks .

58. The apparatus of claim 55 wherein the means for generating the mask array further comprises a means for 3x5 neighborhood processing (120) connected to the score array

(86) and a binary blank mask array (124) having a dust indicated mask array output and a means for 3x3 processing connected to the dust indicated mask array output having a filtered mask array.

59. The apparatus of claim 55 further comprising: a) means for totaling a number of non- zero zones in a pattern in the filtered score array having a total non-zero filtered score array output; b) means for computing a percentage of nonzero elements in the filtered score array; and c) means for comparing the percentage of non-zero elements against a predetermined threshold to determine the slide preparation type if the percentage of non-zero elements exceeds the predetermined threshold.

60. A method for determining the presence of dust on a slide wherein the slide has been scored to generate a filtered score array further comprising the steps of: a) scanning all elements in the filtered score array to determine any elements that have a score within a first predetermined range (180) ; b) scanning a neighborhood of any elements that have a score within the first predetermined range to determine a total number of neighbors that have a score within a second predetermined range (186); c) counting a total number of neighbors having a score in a predetermined range to generate a neighbor count (188) ; and d) determining whether or not the neighbor count is less than a predetermined value

(198) and if the neighbor count exceeds a first predetermined value, applying a second predetermined value to a mask (200) for all neighbors wherein the mask indicates the presence of dust on the slide.

61. A method for determining the presence of a fiducial mark on a slide wherein the slide has been scored to generate a filtered score array further comprising the steps of : a) scanning all elements in an array to determine any elements that have a score in a first predetermined range (152) ; b) scanning a neighborhood of any elements that have a score within the first predetermined range to determine the total number of neighbors that have a score in a second predetermined range (158) ; c) counting the total number of neighbors that have a score in the second predetermined range having a count output (160) ; and d) determining whether or not the count output exceeds a first predetermined value (176) and if the count is less than a first predetermined value applying a second predetermined value to a mask (178) for all neighbors wherein the mask indicates the presence of a fiducial mark on the slide.

62. A method for determining a center of a filtered score array wherein the filtered score array is computed from a field of view score from an image of a biological specimen slide, wherein the method comprises the steps of: a) summing all values of y for each x element in the filtered score array, having a sum of y over x output ; b) summing all values of x for each y, having a sum of x over y output ; c) performing an x histogram on the sum of y over x output ; d) computing a y histogram for the sum of x over y output ; e) calculating a weighted average for x from the x histogram output; f) calculating a weighted average for y from the y histogram output; and g) calculating a center from the weighted average for x and the weighted average for y.