WO2007026598A1

WO2007026598A1 - Medical image processor and image processing method

Info

Publication number: WO2007026598A1
Application number: PCT/JP2006/316595
Authority: WO
Inventors: Hiroshi Fujita; Yoshikazu Uchiyama; Toru Iwama; Hiromichi Ando; Hitoshi Futamura
Original assignee: Gifu University
Priority date: 2005-08-31
Filing date: 2006-08-24
Publication date: 2007-03-08
Also published as: JPWO2007026598A1; JP4139869B2

Abstract

A lesion is accurately detected from a head image. A specific blood vessel portion can be intensively observed. A medical image processor (10) extracts a brain blood vessel region from an MRA image created by imaging a head, and a GC filter bank carries out a primary detection by using the image from which the brain blood vessel region is extracted. An analysis bank of the GC filter bank makes multiresolution analysis to calculate the vector concentration index from a partial image for each resolution level. A weighted image in which voxels having a vector concentration index above a threshold are given a weight of 1 and those having vector concentration index equal to or less than the threshold are given a weight of 0. A reconstruction bank carries out weighting by multiplying each partial image by the weighted image and reconstructs the original image by using the weighted partial images. The averages of the size (volume), spherity, and vector concentration index of the blood vessel region reproduced in the reconstructed image are calculated, and secondary detection is conducted by the rule base method. Tertiary detection of the secondary detection candidate is conducted by the judgment analysis. The result of the detection is outputted and displayed on a display section (13).

Description

Specification

Medical image processing apparatus and image processing method

Technical field

The present invention relates to a medical image processing apparatus and an image processing method for performing image analysis and image processing of a head image obtained by imaging a patient's head.

Background art

In recent years, with the spread and high performance of magnetic resonance imaging devices (hereinafter referred to as MRI; Magnetic Resonance Imaging!), The number of brain dock examinations has increased rapidly. One of the purposes of the brain dock test is to detect unruptured cerebral aneurysms that occur in blood vessels at an early stage and take appropriate measures and treatments to prevent serious diseases such as subarachnoid hemorrhage due to ruptured cerebral aneurysms. It is to prevent disease.

[0003] Detection of an unruptured cerebral aneurysm by a doctor is performed using an MRA image (Magnetic Resonance Angiography) in which the blood flow in the blood vessel is imaged by MRI. Usually, when a doctor interprets a 3D image data from various angles, a 2D image is used by MIP processing (Maximum Intensity Projection). Because the unruptured cerebral aneurysm that occurs in the blood vessel is small Therefore, it is necessary to distinguish from the surrounding blood vessel images that are displayed in an overlapping manner, and the doctor's fatigue is severe. There is also the possibility of oversight due to this fatigue.

[0004] Conventionally, in order to support such a doctor's diagnosis, an apparatus for detecting an image area of a lesion portion by image processing has been developed (for example, see Patent Document 1, Non-Patent Documents 1 and 2). ). This apparatus is generally called CAD (Computer-Aided Diagnosis).

On the other hand, when an image by MIP is generated and displayed in a medical image processing apparatus at the time of interpretation, a plurality of blood vessels may be displayed in an overlapping manner depending on the viewing direction. In this case, in order to detect a small unruptured aneurysm, it is necessary to distinguish between the blood vessel image that the doctor wants to focus on and the surrounding blood vessel image. . Fatigue may overlook an unruptured aneurysm that should be detected.

[0006] It is known that blood vessel sites where aneurysms frequently occur are the middle cerebral artery bifurcation, the anterior communicating artery, the internal carotid artery posterior communicating artery bifurcation, and the like. Therefore, observe these multiple sites in detail. To do this, it is only necessary to generate a target MIP image that focuses only on the blood vessel image of the blood vessel region of interest to the doctor.

[0007] Further, in a region where a plurality of blood vessel parts intersect, using the depth information along the line-of-sight direction, the blood vessel image with the larger depth information, that is, the blood vessel part on the rear side in the line-of-sight direction. A technique for displaying by reducing the luminance of the intersection is also disclosed (for example, see Patent Document 2). According to this method, it is possible to give a sense of perspective to the blood vessel image, and it becomes easy to observe the blood vessel site on the near side.

Patent Document 1: Japanese Patent Laid-Open No. 2002-112986

Patent Document 2: JP-A-5-277091

Non-patent document 1: Norio Hayashi et al., Head MRI image using morphological processing, Automatic extraction of cerebellum and brain affected area, Journal of Medical Image Information Society, vol21.nol.ppl09-115,200 4

Non-Patent Document 2: Ryuro Yokoyama et al., Automatic detection of lacunar infarct area in brain MR images, Journal of Japanese Society of Radiological Technology, 58 (3), 399-405, 2002

Disclosure of the invention

Problems to be solved by the invention

[0008] However, a method for accurately detecting an unruptured cerebral aneurysm by distinguishing it from a vascular tissue has not yet been proposed!

[0009] In addition, since the conventional medical image processing apparatus cannot automatically determine each blood vessel part in the blood vessel image, in order to automatically create the target MIP image, a three-dimensional image data power is required. It is necessary to manually specify only the blood vessel image of the blood vessel site of interest. Such an operation is not easy and requires time.

[0010] In addition, the method described in Patent Document 2 makes it easier to observe a blood vessel portion closer to the observer. Therefore, in the MIP image created according to the line-of-sight direction that the doctor wants to observe, if another blood vessel part intersects the front side of the blood vessel part that the doctor is interested in, it cannot be removed. It is not always possible to observe the desired blood vessel site in detail from the desired observation direction.

[0011] An object of the present invention is to detect a lesion with high accuracy in the head image force. Also specific It is possible to observe while paying attention to the blood vessel site.

Means for solving the problem

[0012] The invention described in claim 1 is a medical image processing apparatus,

Analysis means for performing multi-resolution analysis of the head image and calculating a vector concentration degree for each resolution level using partial images decomposed for each resolution level;

When reconstructing the original head image from the partial image, reconstructing means for reproducing the original image only in a candidate region of a lesion part where the calculated vector concentration degree is a predetermined value; Delete means for deleting a false positive candidate region that is a normal blood vessel from candidate regions of a lesion portion reproduced in the head image using

It is characterized by providing.

[0013] The invention according to claim 8 is the medical image processing apparatus according to claim 1, wherein:

Extracting means for extracting a blood vessel image from the head image;

Image control means for discriminating one or a plurality of blood vessel sites included in the extracted blood vessel image and attaching blood vessel site information relating to the discriminated blood vessel site to the head image.

The invention's effect

[0014] According to the invention described in claims 1 to 4 and 14 to 17, according to the partial image obtained by performing the multi-resolution analysis on the head image, Calculate the concentration of beta. By calculating the vector concentration for each resolution level, it is possible to cope with detection of aneurysms of various sizes, and detection processing with higher accuracy is possible. In addition, a lesion such as an aneurysm exhibits a certain high value of vector concentration, but according to the present invention, reconstruction is performed so that only a region that is highly likely to be a lesion is reproduced. It is possible to obtain a reconstructed image in which only the candidate areas are imaged. By calculating the feature value using the reconstructed image, the calculated feature value force can also exclude image elements other than the candidate region, and the feature value for the candidate region can be calculated accurately. Become. Therefore, it is possible to further improve the accuracy of the detection process itself. [0015] According to the inventions of claims 5 to 7, and 18 to 20, the region in which the vector concentration degree is substantially calculated is the blood vessel region in which the lesion part of the aneurysm exists. This makes it possible to reduce the calculation time.

[0016] According to the inventions of claims 8 and 21, by referring to the blood vessel part information attached to the target image, one or a plurality of blood vessel portions included in the blood vessel image of the target image It is possible to easily determine the position. If the blood vessel part can be identified, for example, each blood vessel part is identified and displayed when the target image is displayed, and each blood vessel part included in the target image is specified when the target image is used, and the information is provided to the doctor. It becomes possible. Therefore, the doctor can observe the target image while paying attention to a specific blood vessel site, and can improve the interpretation efficiency.

[0017] According to the inventions described in claims 9 and 22, by referring to the blood vessel part information attached to the target image, the position and position of each blood vessel part included in the blood vessel image of the target image are referred to. And the name can be easily identified. By determining the position and name of the blood vessel part, the position and name of each blood vessel part included in the target image can be specified when the target image is used, and the information can be provided to the doctor. Therefore, the doctor can easily grasp the position and name of a specific blood vessel site in the target image.

[0018] According to the inventions of claims 10 and 23, the target image has individual differences in the form of the blood vessel image depending on the subject (patient), but the blood vessel image is substantially reduced by affine transformation. By aligning them so that they match, the reference image and the blood vessel image of the target image can be associated with each other with high accuracy. Therefore, it is possible to distinguish a blood vessel site regardless of individual differences of subjects, and the versatility is high.

[0019] According to the inventions described in claims 11 and 24, the doctor can easily identify each of one or a plurality of blood vessel portions included in the blood vessel image of the target image.

[0020] According to the inventions of claims 12 and 25, the doctor can easily grasp the names of one or a plurality of blood vessel parts included in the blood vessel image of the target image.

[0021] According to the inventions described in claims 13 and 26, it is possible to extract and observe only a blood vessel image of a specific blood vessel site that a doctor wants to observe. Since a blood vessel image may be displayed by overlapping multiple blood vessel parts, it is difficult to observe the overlapping part. It becomes easy. Therefore, by displaying only the specific blood vessel part selected by the doctor, it is possible to eliminate duplication of the blood vessel part and provide an environment where the doctor can easily interpret.

Brief Description of Drawings

圆 1] A diagram showing an internal configuration of the medical image processing apparatus according to the embodiment.

圆 2] This is a flowchart for explaining the flow of detection processing executed by the medical image processing apparatus.

FIG. 3A is a diagram showing an example of an MRA image.

FIG. 3B is a diagram showing an example of an extracted image obtained by extracting a blood vessel region.

FIG. 4 is a diagram showing a vector concentration filter.

[5] A diagram showing a cerebral aneurysm model and a blood vessel region model.

FIG. 6 is a diagram showing an output image example of a vector concentration filter.

FIG. 7A is a diagram showing a filtered image before threshold processing.

FIG. 7B is a diagram showing a filtered image after threshold processing.

[8] FIG. 8 is a diagram for explaining a method of calculating sphericity.

FIG. 9A is a diagram for explaining an identification method based on a rule-based method.

FIG. 9B is a diagram illustrating an identification method based on a rule-based method.

FIG. 10 is a diagram showing an output example of a detection result of a cerebral aneurysm candidate.

[11] This is a flow chart for explaining the blood vessel part discrimination processing executed by the medical image processing apparatus.

圆 12A] is a diagram showing an example of a reference image.

FIG. 12B is a diagram showing an original image used to create a reference image.

FIG. 13A is a diagram showing a target image and a histogram of the target image before and after performing normalization processing.

FIG. 13B is a diagram showing a target image of a subject different from that in FIG. 13A and a histogram of the target image before and after performing normal image processing.

FIG. 14A is a diagram showing a target image.

FIG. 14B is a diagram showing a blood vessel extraction image obtained by extracting blood vessels from the target image shown in FIG. 14A. FIG. 15A is a diagram showing a blood vessel extraction image and a reference image.

FIG. 15B is a diagram in which the blood vessel extraction image and the reference image shown in FIG. 15A are superimposed.

FIG. 16A is a diagram showing landmarks in a reference image.

FIG. 16B is a diagram showing corresponding points in a blood vessel extraction image.

FIG. 17A is a diagram showing a blood vessel extraction image and a discrimination result of the blood vessel site.

FIG. 17B is a diagram showing a blood vessel extraction image and a discrimination result of the blood vessel site.

FIG. 18 is a diagram showing an example of identification display of each blood vessel part discriminated in the target image.

FIG. 19 is a flowchart showing a detection process according to the second embodiment.

FIG. 20 is a diagram showing an analysis bank of a GC filter bank.

FIG. 21 is a diagram showing a filter bank A (z ^j ).

FIG. 22 is a diagram showing a reconfiguration bank of a GC filter bank.

FIG. 2 is a diagram showing a filter bank S (z ^j ).

Explanation of symbols

[0023] 10 medical image processing apparatus

11 Control unit

12 Operation unit

13 Display

14 Communications department

15 Memory

16 Lesion candidate detection unit

BEST MODE FOR CARRYING OUT THE INVENTION

First, the configuration will be described.

FIG. 1 shows the configuration of the medical image processing apparatus 10 in the present embodiment.

The medical image processing apparatus 10 detects a candidate region of the medical image force lesion by performing image analysis on a medical image obtained by examination imaging.

Note that this medical image processing device 10 is attached to an image generation device that generates medical images and a medical doctor that stores and manages medical images. It is good also as providing in a medical image system to which various apparatuses, such as an image interpretation terminal which obtains and displays on a display means, are connected via the network. In the present embodiment, an example in which the present invention is realized by a single medical image processing apparatus 10 will be described. However, the functions of the medical image processing apparatus 10 are distributed to each component of the medical image system, and the entire medical image system is used. As a realization of the present invention.

Hereinafter, each unit of the medical image processing apparatus 10 will be described.

As shown in FIG. 1, the medical image processing apparatus 10 includes a control unit 11, an operation unit 12, a display unit 13, a communication unit 14, a storage unit 15, and a lesion candidate detection unit 16.

[0026] The control unit 11 includes a CPU (Central Processing Unit) and a RAM (Random Access Memory).

), Etc., which reads various control programs stored in the storage unit 15 and performs various calculations, and comprehensively controls processing operations in each unit 12-16

The operation unit 12 includes a keyboard, a mouse, and the like. When these are operated by an operator, an operation signal corresponding to the operation is generated and output to the control unit 11. Note that a touch panel configured integrally with the display in the display unit 13 may be provided.

[0028] The display unit 13 includes display means such as an LCD (Liquid Crystal Display), and various operation screens, medical images, and medical image forces are detected on the display means in response to instructions from the control unit 11. Various display information such as the detection result of the lesion candidate and its detection information is displayed.

[0029] The communication unit 14 includes a communication interface, and transmits and receives information to and from an external device on the network. For example, the communication unit 14 performs a communication operation such as receiving a medical image generated from the image generation device and transmitting detection information of a lesion candidate in the medical image processing device 10 to an interpretation terminal.

[0030] The storage unit 15 is a control program used in the control unit 11, various processing programs such as detection processing used in the lesion candidate detection unit 16, and data such as parameters necessary for the execution of each program and processing results thereof. Is remembered.

In addition, the storage unit 15 stores medical images that are candidates for detection of lesion candidates, information on detection results, and the like.

[0031] The lesion candidate detection unit 16 cooperates with the processing program stored in the storage unit 15. The image to be processed is subjected to various image processing (gradation conversion processing, sharpness adjustment processing, dynamic range compression processing, etc.) as necessary. Further, the lesion candidate detection unit 16 executes detection processing and outputs the detection result. The contents of the detection process will be described later.

Next, lesion candidate detection processing by the medical image processing apparatus 10 will be described.

In this embodiment, an example of detecting a lesion candidate for an unruptured cerebral aneurysm from an MRA image (three-dimensional image) obtained by imaging the patient's head using MRI and imaging the blood flow in the brain is explained. The A cerebral aneurysm is a bulge (expansion) that forms in the wall of an artery, and is caused by the blood pressure exerted on the artery wall. If a thrombus occurs inside the cerebral aneurysm or this cerebral aneurysm ruptures, serious diseases such as subarachnoid hemorrhage may develop.

FIG. 2 is a flowchart for explaining the flow of detection processing. As described above, this detection process is a process executed when the lesion candidate detection unit 16 reads a detection processing program stored in the storage unit 15.

As shown in Fig. 2, in the detection process, MRA 3D image data is first input (step Sl). Specifically, the MRA image to be processed stored in the storage unit 15 is read by the lesion candidate detection unit 16.

[0034] Here, an image obtained by MRI will be described.

MRI is a method for obtaining an image using nuclear magnetic resonance (hereinafter referred to as NMR) in a magnetic field.

In NMR, an object is placed in a static magnetic field, and then an RF pulse (radio wave) at the resonance frequency of the atomic nucleus to be detected in the object is irradiated. For medical purposes, the resonance frequency of the hydrogen atom that constitutes the water that is abundant in the human body is usually used. When the object is irradiated with RF noise, an excitation phenomenon occurs, the nuclear spins of the atoms that resonate with the resonance frequency are aligned, and the nuclear spins absorb the energy of the RF pulse. When the irradiation of the RF pulse is stopped in this excited state, a relaxation phenomenon occurs and the phase of the nuclear spin becomes nonuniform, and the nuclear spin releases energy. The phase relaxation time constant is T2, and the energy relaxation time constant is T1.

[0035] In MRI, it is possible to obtain various types of images according to the inspection purpose by changing the imaging method. For example, for repetition time TR and echo time TE, TR = T1, TE <<T2 and shooting conditions adjusted are Tl weighted images, TR>> T1, and ΤΕ = Τ2 and shooting conditions adjusted are called Τ2 weighted images. The T1-weighted image is used to detect anatomical structures, and the Τ2-weighted image is used to detect lesions. In addition, images taken by the FLAIR method are 強調 2 weighted images in which signals from water are attenuated, but are particularly called FLAIR images.

[0036] MRA is a blood vessel imaging method in MRI! In MRI, the direction of the subject's foot to head

By applying a gradient magnetic field (this direction is called the body axis), it is possible to absorb energy only in a specific slice (fault). At the time of imaging, blood is saturated with RF pulses in the blood vessels in the slice. Since blood flows constantly in the blood vessels, the signal strength in the slice increases when unsaturated blood flows in over time. To do. MR A is a method of imaging a blood vessel with blood flow by imaging this high signal.

FIG. 3A shows an example of MRA image.

As shown in Fig. 3A, the blood vessel region with blood flow has a high signal, so the blood vessel region appears white in the MRA image.

[0038] When such 3D data of the MRA image is input, the 3D image data is preprocessed as a preparation stage for detecting candidates (step S2). As preprocessing, normalization processing and gradation conversion processing of image data are performed. Normalization is performed by converting by linear interpolation so that all the edges that make up the botacell become 3D image data of the same size. Then, the density gradation conversion process is performed on the three-dimensional image data converted into the equal-sized button cells, and the signal value of each button cell is linearly converted into a density gradation of 0 to 1024. At this time, the higher the signal value, the closer to the density value 1024, and the lower the signal value, the closer the density value to 0. The density gradation range is not limited to 0 to 24 and can be set as appropriate.

[0039] When the preprocessing is completed, a blood vessel image region is extracted from the three-dimensional MRA image (step S3). First, threshold processing is performed, and the MRA image is binarized. In general, in the MRA image, as shown in FIG. 3A, the blood vessel region is whitened and the other regions appear black. In the binary image, the blood vessel region has a different value from the other regions. Therefore, the blood vessel region is extracted by the region expansion method. First, using the binary image, the starting botasel (the whitest Determine the density value (botacel), and examine the 26-botel in the vicinity of the determined botasel in the 3D MRA image before binarization, and determine certain judgment conditions (for example, the density value is 500 or more) Neighboring bocellels satisfying the condition are determined as blood vessel regions. Then, the same processing as described above is repeated for the neighboring buttonacell determined to be the blood vessel region. In this way, the blood vessel region can be extracted by sequentially extracting the botacell satisfying the determination condition while expanding the region. Fig. 3B shows the MRA image force extracted from Fig. 3A. The extracted blood vessel region is white (concentration value 1024) and the other regions are black (concentration value 0).

[0040] Next, the extracted 3D MRA image of the blood vessel region is subjected to filter processing using a vector concentration filter as shown in Fig. 4, and the cerebral aneurysm is output from the processed image output by the filter processing. A primary candidate area is detected (step S4). The vector concentration degree filter calculates the vector concentration degree in each botacell, and images and outputs the calculated vector concentration value as the botacell value of the botacell. The vector concentration degree focuses on the direction of the gradient vector of density change and evaluates how much the gradient vector in the neighboring area is concentrated at a certain point of interest.

FIG. 5 shows a cerebral aneurysm model and a blood vessel model.

As shown in Fig. 5, in the case of a cerebral aneurysm, there is a spherical aneurysm on a linear blood vessel, so the gradient vector (the arrow in the figure indicates the direction of the gradient vector) goes to the center of the aneurysm. There is a tendency to go. On the other hand, since the blood vessel has a linear shape, such a tendency does not occur. Therefore, the area close to the shape of the cerebral aneurysm model has a higher vector concentration value than other blood vessel areas. Therefore, the primary candidate region of the cerebral aneurysm can be detected by outputting only the region having a high vector concentration level by the vector concentration filter.

[0042] Specifically, when an attention vessel cell P as shown in FIG. 4 is scanned on the extraction image of the blood vessel region, the extracted blood vessel region is within the range of a sphere having a radius R centered on the attention button cell P. If it exists, the vector concentration is calculated.

The vector concentration is calculated by the following formula 1.

[Number 1]

^GC ( ^P ). ⁽¹ )

Here, the angle Θ indicates the angle between the direction vector from the target button cell Ρ to the peripheral button cell Qj and the direction of the vector in the peripheral voxel Qj, and M indicates the number of peripheral button cells Qj to be operated. (Refer to Figure 4).

[0043] When the vector concentration degree is obtained for each of the botacells constituting the blood vessel region, a filtered image as shown in Fig. 6 is output with the vector concentration degree as the botacell value. Since the vector concentration is output in the range of 0–1, in Fig. 6, the higher the vector concentration (closer to 1), the more white the image will appear! /

[0044] Next, a threshold processing for binarizing the filtered image using, for example, a threshold value of 0.5 is performed, and a primary candidate region is detected. In other words, a region with a Botacel force having a vector concentration greater than the threshold value 0.5 by the binary key is extracted as a primary candidate region. FIG. 7A is a diagram showing a part of the filtered image shown in FIG. By binarizing this filtered image, a binary key image as shown in FIG. 7B is obtained. As shown in FIG. 7B, the binary image appears in white! /, That is, the vector concentration degree is large! /, And the image area is output as the primary candidate area.

[0045] Note that the threshold for binarization is determined using a teacher image in which the presence of a cerebral aneurysm is known in advance. In other words, by thresholding the teacher image, a threshold value is obtained so that only the image area of the cerebral aneurysm whose existence has already been identified can be extracted. It is good also as obtaining by analyzing statistically. In the p-tile method, a density histogram is obtained, and a density value at a certain area ratio p% in the density histogram is obtained as a threshold value. In the present embodiment, the density histogram of the filtered image is obtained, and the density value at the maximum density value side force area ratio p% is determined as a threshold value.

[0046] Next, in order to perform secondary detection, a feature amount indicating the feature of the primary candidate region is calculated (step S5). Since a cerebral aneurysm has a certain size and a spherical shape, in this embodiment, the size of the candidate region, the sphericity, and the average value of the vector concentration in each botacell in the region are used as feature quantities. It will be calculated. However, if the cerebral aneurysm can be characterized, there is no particular limitation on what kind of feature is used. The maximum value of the vector concentration may be calculated, or the standard deviation of the density value of each botasel may be calculated.

[0047] As the feature quantity of the size, the volume of each of the botacels constituting the candidate area is calculated. Here, in order to shorten the processing time, the number of botasels that are not the actual volume is calculated and used as the index value indicating the volume in the subsequent calculations. In addition, as shown in FIG. 8, the sphericity feature amount is obtained when a sphere having the same volume as the primary candidate area is arranged so that the centroid of the primary candidate area and the centroid of the sphere coincide with each other. Is obtained from the ratio of the volume of the primary candidate region portion that coincides with the total volume of the primary candidate region.

[0048] When the feature amount is calculated, secondary detection is performed based on the feature amount (step S6). In the secondary detection, a cerebral aneurysm is distinguished from a blood vessel that is a normal tissue. Discrimination is a force that shows an example of a discriminator using the rule-based method. The method is not limited to this, and any method can be used as long as discrimination is possible, for example, artificial-eural network, support vector machine, discriminant analysis.

In the rule-based classifier, as shown in FIGS. 9A and 9B, first, a profile indicating the relationship between the sphericity with respect to the size and the average of the vector concentration with respect to the size is created. In the classifier, the range to be detected as the secondary candidate (the range surrounded by the solid line in FIGS. 9A and 9B) is determined in advance, and the above profile file is used as the variable data for each feature quantity of the primary candidate to be identified. When it is created, it is identified as a true positive candidate if the variable data of each feature quantity of the primary candidate is within the detection range, and a false positive candidate if it does not exist. In other words, only the primary candidate in which the variable data of the feature quantity is distributed within the detection range is secondarily detected.

[0050] Note that the detection range is determined using a teacher image that is previously known to be a cerebral aneurysm or a normal blood vessel. First, the average of size, sphericity, and vector concentration is obtained from cerebral aneurysm and normal blood vessel teacher images, and the sphericity and size of the vector with respect to size are calculated from these features. A profile of the average value relationship is created. In Figure 9A and Figure 9B, the variable data indicated by the ○ marker is true positive, that is, teacher data that is known as a cerebral aneurysm, and the variable data indicated by the 參 marker is false positive, that is, it is known as a blood vessel. Teacher data. At this time, since the image features differ between the cerebral aneurysm and normal blood vessels, there is a bias in the distribution of the feature values of the average values of size, sphericity, and vector concentration. Jiru. Therefore, two threshold values are determined for the average values of size, sphericity, and vector concentration so that the range in which teacher data for all cerebral aneurysms is included in the profile is minimized. A region range surrounded by the threshold is a detection range.

[0051] In the examples of FIGS. 9A and 9B, the range surrounded by four dotted lines indicating the threshold (only the portion surrounding the range is indicated by a solid line) is the detection range, and “size sphericity”, “ The detection range is determined for each of “average vector density”. In any profile, the primary candidate (the variable data is indicated by a △ marker) located within the detection range is secondarily detected as a cerebral aneurysm candidate.

[0052] When secondary detection is performed in this manner, tertiary detection is further performed on the secondary detection candidates (step S7). In tertiary detection, discriminant analysis is performed using three feature quantities. As a discriminant analysis method, any of Mahalanobis distance, principal component analysis, linear discriminant function, etc. can be applied, but a method different from the method at the time of secondary detection is applied.

[0053] Then, the detection result is output assuming that the candidate region that has been thirdarily detected is the final cerebral aneurysm candidate (step S8).

FIG. 10 shows an example of a detection result displayed on the display unit 13 as a detection result. As shown in FIG. 10, in the display unit 13, marker information (marked by an arrow in FIG. 10) indicating a candidate region of the third detected cerebral aneurysm in the MIP image created from the 3D MRA image Is displayed. By performing such display, the cerebral aneurysm candidate region and other regions can be identified. A MIP image is a 2D image created by applying MIP processing to 3D MRA image data, and the structure in the image can be displayed in 3D. MIP processing is called the maximum brightness projection method, and projection is performed with parallel rays from a certain direction, and the maximum brightness (signal value) in the button cell is reflected on the projection surface to enable 3D observation. This is a process for creating an image.

[0054] Information regarding detection of cerebral aneurysm candidates, such as the average value of the vector concentration calculated for the cerebral aneurysm candidate region, may be output and used as reference information at the time of diagnosis by a doctor. Further, the color of the marker information may be changed according to the degree of vector concentration. For example, change the color of the arrow marker to red if the vector concentration is 0.8 or higher, yellow if it is 0.7 to 0.8, blue if it is 0.7 to 0.5, etc. The vector concentration, The doctor can easily grasp visually that the degree of the spherical shape of the aneurysm is strong.

FIG. 10 shows an example in which the position of a cerebral aneurysm candidate can be identified by marker information. However, if the cerebral aneurysm candidate region is displayed so as to be distinguishable from other regions, such a case is displayed. It is not limited to the display method. For example, the volume rendering method that is not used in MIP processing may be used to display 3D text. The volume rendering method is a method of performing three-dimensional display by giving color information and opacity to each botacell for each partial area. For the attention area, the opacity is set high, and other than that By setting the area low, the area of interest can be raised. Therefore, at the time of display, opacity is set for each area, and color information setting processing corresponding to the opacity is performed.

[0056] Further, the detection result may be shown using a filter processing image (see FIG. 6) obtained by a vector concentration filter that is not used for MIP images or the like. At this time, for example, the vector concentration degree is low, the value area is colored blue, the high value area is colored red, etc., so that the doctor can visually grasp the calculated vector concentration degree. You may do it. Further, a filtered image in which the color of the blood vessel region is changed according to the vector concentration degree may be displayed superimposed on the corresponding position of the MIP image. In this way, it is possible to provide vector concentration information as reference information for detecting a cerebral aneurysm by a doctor.

Next, the blood vessel part determination process will be described with reference to FIG.

This blood vessel part discrimination process is a software process realized in cooperation with the control program 11 and the blood vessel part discrimination process processing program stored in the storage unit 15. In the blood vessel part discrimination process, one or a plurality of blood vessel parts included in the blood vessel image are discriminated from the blood vessel image appearing on the 3D MRA image obtained by photographing the head.

[0058] MRA is a kind of MRI blood vessel imaging method. In MRI, energy can be absorbed only in a specific slice (fault) by applying a gradient magnetic field in the direction from the foot of the subject to the head (this direction is called the body axis). When imaging, the blood vessels in the slice are saturated with RF pulses. Since blood vessels always flow, the signal intensity in the slice increases when non-saturated blood flows over time. . MRA is a method for imaging blood vessels with blood flow by imaging this high signal. [0059] First, a reference image necessary for determining a blood vessel site will be described.

As shown in FIG. 12A, the reference image is a blood vessel image on the three-dimensional MRA image in which the positions and names of one or a plurality of blood vessel parts are preset. Here, the vascular part refers to the anatomical classification of blood vessels, and the position of the vascular part refers to the position of the botacel belonging to the vascular part.

[0060] In FIG. 12A, eight blood vessel sites included in the blood vessel image (anterior cerebral artery, right middle cerebral artery, left middle cerebral artery, right internal carotid artery, left internal carotid artery, right posterior cerebral artery, left posterior cerebral artery) , The position of the basilar artery) and the name of the blood vessel part are shown. In FIG. 12A, of the eight vascular sites, the names of the three vascular sites (right middle cerebral artery, anterior cerebral artery, and basilar artery) are shown, but all eight vascular sites are shown. The name is set.

[0061] The reference image g2 is generated as a three-dimensional data force of the head MRA image gl selected for the reference image as shown in FIG. 12B. First, an axial image (a two-dimensional tomographic image obtained by cutting out a botasel in a plane perpendicular to the body axis) is created at intervals of this three-dimensional data force. Then, based on the doctor's indication in one axial image, the botacell belonging to each vascular site is designated by manual operation, and the name of the vascular site is further designated. By repeating this for each axial image sliced by changing the position in the body axis direction, it is possible to set the position of the vessel cell belonging to each blood vessel part and the name of the blood vessel part among all the botacells constituting the three-dimensional data. it can.

[0062] In addition, landmark botasels are set in the reference image g2 at characteristic points such as the inflection point of the blood vessel, the end point, and the intersection of the blood vessel portions. The landmark is used for alignment between the target image and the reference image, and will be described in detail later. Landmarks are also set according to manual operations based on doctors' indications.

[0063] The reference image g2 created as described above is stored in the storage unit 15.

The reference image g2 may be created by the control unit 11 of the medical image processing apparatus 10, or an externally created image may be stored in the storage unit 15. In addition, in order to show the eight blood vessel parts included in the blood vessel image, each blood vessel part is identified and displayed in FIG. 12A, but the actual reference image g2 has a black background (low signal value) and a white blood vessel image (high signal value). ) And binarized image. Then, the position information, name information, and landmarks of the botasels belonging to each blood vessel site The position information of the button cell that is a key is attached to the reference image or stored in the storage unit 15 as a separate file in association with the reference image.

[0064] Next, a blood vessel part discrimination process using the reference image will be specifically described.

In the medical image processing apparatus 10, a normalization process is first performed by the control unit 11 on the three-dimensional MRA image (hereinafter referred to as the target image and V ヽぅ) to be determined (step S11).

Since the MRA image used as the target image is an image of the blood flow, depending on the subject and the shooting conditions, the botacell may be a rectangular parallelepiped, or the maximum and minimum values of the voxel may vary. It occurs. Therefore, normalization processing is performed to unify the preconditions regarding the target image.

In the normalization process, first, the target image is converted by the linear interpolation method so that all the sides constituting the botacell have the same size. Next, a histogram is created for all the botacell values of the target image, and all the botacel values of the target image are 0, with the top 5% or more of the histograms having a value of 1024 for the top 5% or more and 0 for the minimum. Linear conversion to 1024 tones. At this time, the higher the botacell value is, the closer the density value is to 1024. The lower the signal value is, the closer the density value is to 0. The density gradation range is not limited to 0 to 1024 and can be set as appropriate.

FIG. 13A and FIG. 13B show an example of normalization processing.

The target image g3 shown in FIG. 13A and the target image g4 shown in FIG. 13B are obtained by using different patients as subjects. For this reason, the histogram hi (see Fig. 13A) obtained from the target image g3 and the histogram h3 (see Fig. 13B) obtained from the target image g4 share the common feature that there are two local maxima. It can be seen that there is a considerable difference in the range of values, and the histogram characteristics as a whole are different. If the target images g3 and g4 having such a histogram characteristic are subjected to the above-mentioned regularity processing and then created again, the histogram h2 shown in FIG. 13A and the histogram h4 shown in FIG. 13B are obtained. . Histogram h2 and h4 force As shown by the component force, the histogram characteristics of the target images g3 and g4 are almost the same by the normalization process.

When the normalization process is completed, the control unit 11 extracts a blood vessel image from the normalized target image (step S12). First, threshold processing is performed on the target image, and binarization is performed. In general, in the MRA image, as shown in Fig. 14A, the blood vessel image appears white and other tissue parts appear black. Therefore, in the binary image, the blood vessel image has a different value from the other regions. Therefore, the region having the same signal value as the blood vessel image is extracted by the region expansion method.

[0068] In the area expansion method, a binary cell is used to determine the starting botasel (the whitest and highest density voxel), and is determined to be the starting point in the target image before the binary image processing. In the vicinity of the Botacel, 26 Botacels are examined, and a neighboring Botacell that satisfies a certain determination condition (for example, a density value of 500 or more) is determined as a blood vessel image. Then, the same processing as described above is repeated for the neighboring buttonacell determined to be the blood vessel image. In this way, the image region of the blood vessel image can be extracted by sequentially extracting the botacell satisfying the determination condition while expanding the region.

[0069] FIG. 14B shows a blood vessel extraction image g6 obtained by extracting blood vessel images.

The blood vessel extraction image g6 is obtained by extracting a blood vessel image from the target image g5 after normal input shown in FIG. 14A, and the region of the blood vessel image is white (density value 1024) and the other regions are black (density value 0). It is a valuation.

Next, in the control unit 11, in order to make the position of the blood vessel image of the blood vessel extraction image substantially coincide with the position of the blood vessel image of the reference image, alignment is performed based on the position of the center of gravity of each image (step S 13). The position of the center of gravity is the position of the button cell that is the center of gravity of all the button cells belonging to the blood vessel image.

Specific description will be given with reference to FIGS. 15A and 15B.

FIG. 15A is a diagram in which the blood vessel extraction image and the reference image before alignment are superimposed. From FIG. 15A, it can be seen that the positions of the respective blood vessel images coincide with each other simply by combining the blood vessel extraction image and the reference image.

Therefore, the control unit 11 obtains the positions of the centroid P (xl, yl, zl) of the blood vessel extraction image and the centroid Q (x2, y2, y3) of the reference image as shown in FIG. 15A. Next, the blood vessel extraction image or the reference image is translated so that the barycentric positions P and Q match. FIG. 15B is a diagram showing the resultant force obtained by matching the center of gravity positions P and Q by translation. Fig. 15B Force The fact that the blood vessel image in the blood vessel extraction image and the blood vessel image in the reference image are roughly coincident with each other is divided. [0072] Further, in order to perform alignment with high accuracy, the control unit 11 performs rigid deformation on the blood vessel extraction image (step S14).

First, a corresponding point search using a cross-correlation coefficient is performed as preprocessing for rigid body deformation. This is because a plurality of corresponding points are set for each of the two images to be aligned, and one of the images is rigidly deformed so that the corresponding points set in the two images match each other. Here, a land mark button that is determined in advance in the reference image and a blood vessel extracted image that has locally similar image characteristics are set as corresponding points. The similarity of image characteristics is determined based on the cross-correlation coefficient obtained for the blood vessel extraction image and the reference image.

Specifically, as shown in FIG. 16A, corresponding points corresponding to 12 landmarks set in advance in the blood vessel image of the reference image g7 are searched from the blood vessel extraction image. When searching for the corresponding points, as shown in FIG. 16B, in the blood vessel extraction image g8, the start point is the button cell at the position corresponding to each landmark of the reference image g7, and the start point in the blood vessel extraction image g8 and the reference image g7. In addition, a search is made for a button cell in the range of 10 to +10 (in a cubic region of 21 X 21 X 21) in the X-axis, Y-axis, and Z-axis directions and the landmarks. Correlation coefficient C (hereinafter referred to as correlation value C) is calculated.

[0074] [Equation 2]

In the above equation 2, A (i, j, k) represents the botacell position of the reference image g7, and B (i, j, k) represents the botacell position of the blood vessel extraction image g8. UK indicates the size of the search area, UK = 21 X 21 X 21.

Further, | 8 is the average value of the voxel values in the search region in the reference image g7 and the blood vessel extraction image g8, and is represented by the following expressions 3 and 4. σ and σ are reference images, respectively.

A B

This is the standard deviation of the botacel value in the search area in image g7 and blood vessel extraction image g8, and is shown by the following equations (5) and (6). [Equation 3]

«= ^ ∑∑∑ (iJ, k) (3)

丄 k = l j = l i = l

∑∑∑B (, k) (4)

k = l j = l i = l

K J I

∑∑ {Α (^)-ο;

k = l j = l i = l

σ (5)

UK

K J I

∑∑∑ {B (i, j, k)-^} ²

k = l j = l i = l

σ (6)

UK

[0076] The correlation value C has a value range of 1.0 to 1.0, and the closer to the maximum value 1.0, the more similar the image characteristics of the reference image g7 and the blood vessel extraction image g8! / Show.

Therefore, the position force of the botel cell having the largest correlation value C is set as the corresponding point of the blood vessel extraction image g8 corresponding to the landmark of the reference image g7.

[0077] When the corresponding point is set, the control unit 11 performs a rigid body deformation on the blood vessel extraction image g8 based on the corresponding point, and thereby the blood vessel image of the blood vessel extraction image g8 and the blood vessel image of the reference image g7. And are aligned. Rigid body deformation is one of the affine transformations, in which coordinate transformation is performed by rotation and translation. The alignment is performed so that the corresponding point of the blood vessel extraction image g8 matches the landmark of the reference image g7 by an ICP (Iterative Closest Point) algorithm that repeats rigid body deformation using the least squares method multiple times. According to this algorithm, every time a rigid body is deformed, the least square error of the distance at the corresponding point between the landmark of the reference image and the blood vessel extraction image is calculated, and the end condition such as the least square error exceeding a certain threshold is set. The rigid body deformation is repeated until it is satisfied. Note that rigid body deformation may be replaced with affine transformation (coordinate transformation by scaling, rotation, or parallel movement).

[0078] Next, based on the reference image, blood in the blood vessel extraction image aligned by rigid body deformation Each blood vessel site included in the tube image is discriminated by the control unit 11 (step S15).

First, the square of the Euclidean distance from a certain target vessel cell in the blood vessel extraction image is obtained for all the vessel cells belonging to each blood vessel part of the reference image (this is called the target button cell). Then, it is determined that the blood vessel part to which the target botacell having the shortest Euclidean distance belongs is the blood vessel part to which the target botacell belongs. At this time, the name of the blood vessel part of the target button cell is determined from the name of the blood vessel part set in the target button cell.

[0079] When the above processing is performed for all the botasels constituting the blood vessel image of the blood vessel extraction image and the above-described processing is performed, and the corresponding blood vessel portions are determined for all the botasels, Blood vessel part information indicating the position of the botacel to which it belongs and the name of the blood vessel part are generated by the control unit 11 and attached to the target image (step S16). For example, if it is determined that the button cell at the position (x3, y3, z3) is a blood vessel part of the anterior cerebral artery, the button cell at the position “(x3, y3, z3)” has the blood vessel name “anterior cerebral artery”. The blood vessel part information indicating that this is a blood vessel part is appended to the header area of the target image.

FIG. 17A and FIG. 17B show the results of determining the blood vessel site.

The blood vessel extraction image g9 shown in FIG. 17A and the blood vessel extraction image gl l shown in FIG. 17B are images obtained from different subjects, respectively, and the image glO shown in FIG. 17A and the image gl 2 shown in FIG. Each blood vessel part is discriminated from the blood vessel extraction images g9 and gl l, and is an image that is identified and displayed by changing the color for each blood vessel part. By aligning the images glO and gl2 by rigid body deformation, the blood vessel images of the images glO and gl2 are the same regardless of their different forms (blood vessel position, size, extension direction, etc.). It can be seen that the blood vessel site can be identified.

The above is a flow from determining a blood vessel part in the target image to attaching the blood vessel part information.

When a display instruction operation is performed on such a target image via the operation unit 12, the control unit 11 performs MIP processing on the target image to generate a MIP image and displays it on the display unit 13. Hereinafter, the display of MIP images is referred to as MIP display.

[0082] In the MIP method, a certain directional force is projected by parallel rays and is on this projection line. This is a method of creating a two-dimensional image by reflecting the maximum luminance (botacel value) in the botacel on the projection plane. This projection direction is the line-of-sight direction that the doctor desires to observe. Here, the doctor can freely operate the observation direction, and the control unit 11 creates a MIP image from the target image according to the observation direction instructed through the operation unit 12, and displays the MIP image. It shall be displayed in part 13.

[0083] In the state where the target image is displayed in MIP, when the instruction operation force S for identifying and displaying the vascular part is further performed, the control unit 11 refers to the vascular part information attached to the target image, and Based on this, display control is performed so that each blood vessel part can be identified in the blood vessel image in the MIP image (step S17).

[0084] For example, when the position and name of the vessel cell of each blood vessel part are determined based on the blood vessel part information !, the control unit 11 displays blue for the voxels belonging to the blood vessel part of the anterior cerebral artery in the MIP image. In addition, the color of each blood vessel part is set to the button cell located at the position determined for each blood vessel part, such as green for the botacel belonging to the blood vessel part of the basilar artery. Then, the set color is reflected in the MIP image of the target image. Furthermore, an annotation image indicating the name of the blood vessel part is created and synthesized with the corresponding blood vessel part of the MIP image.

FIG. 18 shows an example of identification display.

In Fig. 18, the MIP image gl3 is the target image displayed with the head upward force MIP. When the identification display of each blood vessel region is instructed with the MIP image gl 3 displayed, the identification display image g 14 is displayed. The identification display image gl4 is obtained by identifying and displaying each blood vessel part by, for example, assigning different colors to the eight kinds of blood vessel parts in the blood vessel extraction image.

In the identification display image gl4, when a doctor selects a blood vessel part, an annotation indicating the name of the blood vessel part such as “basal artery” is displayed in association with the selected blood vessel part by the display control of the control unit 11. Is displayed.

[0086] When MIP display from the side direction is instructed to the MIP image gl3 in the head upward direction, the MIP image gl5 corresponding to the side direction is created by the control unit 11 and displayed. In the MIP image gl5, the blood vessel image on the front side overlaps the blood vessel image on the rear side, making it difficult to observe. Even in such MIP image gl5, it is possible to identify and display blood vessel sites. Noh. By referring to the blood vessel part information, it is possible to determine which botacell corresponds to which blood vessel part, and therefore, it is possible to identify the botacell to be identified and displayed regardless of the change in the observation direction of the MIP display. is there.

[0087] The identification display image corresponding to the MIP image g15 is the image g16. In this identification display image g 16, it is possible to extract and display only one of the blood vessel sites.

When any one blood vessel part is selected and operated via the operation part 12 in the control unit 11, the MIP image in which only the luminance of the selected vessel vessel is projected, that is, the MIP image of the target image. A blood vessel selection image gl7 in which only the blood vessel site selected from gl5 is extracted is displayed. In the blood vessel selection image gl7, only the selected blood vessel part is displayed in MIP and the other blood vessel parts are not displayed, so that the doctor can observe only the blood vessel part of interest.

[0088] In addition, these display images _g 13～Gl7 also by stone as be displayed side by side on the same screen, it is also possible to switch the display as one screen first image. In the former case, the MIP display image gl3, the identification display image gl4, and the blood vessel selection image gl7 can be compared and observed. In the latter case, each image gl3 to gl7 can be observed in full screen display. This makes it easier to observe details.

Example

Hereinafter, examples of the detection process will be described.

Three-dimensional MRA image data for 20 patients were obtained. These image data have a matrix size force S256 X 256, a spatial resolution of 0.625 to 0.78 mm, and a slice thickness of 0.5 to 1.2 mm. These image data are divided into seven unruptured cerebral aneurysms. An unruptured cerebral aneurysm has been determined by an experienced neurosurgeon.

[0090] First, these three-dimensional MRA image data were converted to image data having an equal botacell size using a linear interpolation method, and normalization was performed. As a result of this normalization, all image data became equal-botacel image data with a matrix size power of S400 X 400 X 200 and a spatial resolution of 0.5 mm. Next, after performing gradation conversion processing and linearly converting the density gradation from 0 to 1024, the size (volume), sphericity, and velocity of the cerebral aneurysm region (true positive) determined by the doctor are determined. Each feature amount of the degree of concentration of tuttle is calculated, and the same feature amount is calculated for a normal blood vessel region (false positive). These feature quantities were used to determine the detection range of the rule-based method, which is a discriminator for secondary detection, and as teacher data for discriminant analysis. Similarly, the discriminator was trained as a teacher image in the discriminator at the third detection.

[0091] Using the discriminator constructed as described above, the detection process of Fig. 7 was performed on 3D MRA image data of 20 cases, and 1.85 false positive candidates were included per case. The correct answer rate for detecting cerebral aneurysms was 100%, confirming that it could be detected with high accuracy.

[0092] As described above, according to the first embodiment, a cerebral aneurysm candidate having a characteristic that a gradient vector concentrates in the central portion is accurately detected using a vector concentration filter, and the detected information is obtained by a doctor. Can be provided. Therefore, it is possible to prevent fatigue and oversight during doctor's interpretation work, and it is expected to improve diagnosis accuracy.

[0093] In addition, since the cerebral aneurysm candidate region is displayed so as to be distinguishable from other regions, the doctor can easily grasp the detection result on the image and facilitate the interpretation work. Can

[0094] Further, since the vector concentration filter is applied not to the MRA image itself but to the extracted image obtained by extracting the blood vessel region, it is possible to shorten the processing time required for the filter processing.

Note that the above embodiment is a preferred example to which the present invention is applied, and the present invention is not limited to this. For example, in the above description, a cerebral aneurysm candidate is detected using a 3D MRA image. Detection may be performed using an MRA image. In this case, the size of the cerebral aneurysm candidate is the number of pixels, and the sphericity is a circularity, and a two-dimensional feature value is calculated.

[0096] Although an example of detecting an aneurysm candidate using an MRA image has been described, an MRI image obtained by other imaging methods such as a contrast MRA image obtained by imaging a blood vessel region using a contrast agent may be used. However, it is also possible to use an image in which a blood vessel region is imaged by another imaging device such as CTA (Computed Tomography Angiography) or D¾A (Digital Subtraction Angiography). In addition, the detection target can be detected by applying the present invention as long as it is not only an aneurysm but also a lesion having a spherical aneurysm. [0097] Further, according to the above-described embodiment, by aligning the reference image and the target image with predetermined positions and names of the eight blood vessel portions included in the blood vessel image on the image with respect to the blood vessel image, The position and name of each blood vessel part included in the blood vessel image on the target image is determined. Since the position and name information is attached to the target image as blood vessel part information, when performing MIP display of the target image, each blood vessel part can be easily determined based on the blood vessel part information. The identification display of each blood vessel site is possible. Therefore, the doctor can observe the target image while paying attention to a specific blood vessel site in the target image, and can improve the interpretation efficiency.

[0098] When performing MIP display, depending on the direction in which the MIP display is performed, a plurality of blood vessel sites may overlap, making it difficult to observe the blood vessel site to be noticed.

However, according to the present embodiment, since each blood vessel part can be identified and displayed, the doctor can easily identify the position and name of each blood vessel part, and can perform the interpretation work. Efficiency can be improved.

[0099] In addition, a target MIP image in which only the selected blood vessel part is extracted and displayed in response to a selection operation of an arbitrary blood vessel part among the identified blood vessel parts is generated and displayed. The doctor can observe only the blood vessel site to be noticed. Therefore, duplication of a plurality of blood vessel sites can be eliminated, and observation can be performed by identifying a specific blood vessel site such as a site where multiple aneurysms occur.

[0100] Also, when aligning the blood vessel images of the target image and the reference image, first, rough alignment is performed based on the position of the center of gravity, and then the target image is subjected to rigid body deformation using the cross-correlation coefficient. Then, fine alignment is performed so that the feature point of the blood vessel image of the target image matches the corresponding feature point of the blood vessel image of the reference image.

[0101] The morphology of the major blood vessel sites (the length of the blood vessels, the direction of extension, the thickness, etc.) is generally the same for different subjects (patients), but the shape of the minor blood vessels that are not major varies depending on the individual. Therefore, the shape of the blood vessel part varies depending on the subject.

However, as shown in this embodiment, by aligning the positions of the feature points such as the main inflection points of the blood vessel image by the rigid body deformation finally, what form each blood vessel part in the target image takes on Even with this, it is possible to accurately correspond to the blood vessel portion of the reference image. Therefore, the covered The blood vessel part can be identified uniformly regardless of the subject, and is highly versatile.

[0102] Furthermore, since the alignment is performed in two stages (based on the position of the center of gravity and based on rigid body deformation), it is possible to improve the determination accuracy for determining the blood vessel site. In addition, by performing alignment based on the position of the center of gravity before the deformation of the rigid body, the processing time for the rigid body deformation can be shortened, and the processing efficiency is good.

[0103] The above-described embodiment is a preferred example to which the present invention is applied, and the present invention is not limited to this.

For example, although an example using an MRA image has been described, an MRI image obtained by another imaging method such as a contrast MRA image obtained by imaging a blood vessel using a contrast agent may be used. In addition, an image obtained by imaging a blood vessel with another imaging apparatus such as CTA (Computed Tomograpny Angiography) or DSA (Digital subtraction Angiography) may be used.

In the second embodiment, an example of detection processing by the GC filter bank will be described. The medical image processing apparatus according to the second embodiment has the same configuration as the medical image processing apparatus 10 according to the first embodiment, and only the operation is different. Therefore, the same components as those in the medical image processing apparatus 10 (see FIG. 1) according to the first embodiment are denoted by the same reference numerals, and the operation of the medical image processing apparatus 10 in the second embodiment will be described below.

FIG. 19 is a flowchart showing the detection process according to the second embodiment.

In the detection process shown in FIG. 19, the MRA 3D image data is first input (step S101), and the 3D image data is preprocessed (step S102). When the preprocessing is completed, the image region of the 3D image data force blood vessel is extracted (step S 103). Since steps S101 to S103 are the same processing as steps Sl to 3 described with reference to FIG. 2 in the first embodiment, detailed description thereof is omitted here.

Next, the primary candidate region of the cerebral aneurysm is detected by the GC filter bank using the extracted three-dimensional MRA image of the blood vessel region (step S104).

The detection method using the GC filter bank is described below.

The GC filter bank is a combination of various filter processes and is divided into an analysis bank and a reconstruction bank. The analysis bank is the original image (3D MRA image of the blood vessel region) Multi-resolution analysis is performed to create images with different resolution levels (hereinafter referred to as partial images!) And weight images are created from these partial images. On the other hand, the reconstruction bank weights each partial image with a weighted image, and then reconstructs the weighted partial image force original image.

FIG. 20 shows an analysis bank.

As shown in Figure 20, the analysis bank uses a 3D MRA image of the blood vessel region as the original image S

0 Filter processing is performed in filter bank A (z ^j ), and partial images of each resolution level j are sequentially created. Here, an example of creating partial images with resolution levels 1 to 3 will be described.

Filter bank A (z ^j ) decomposes image S into partial images S, Wz, Wy, and Wx through filter processing with filters H (z ^j ) and G (z ^j ) as shown in FIG. is there. Where S is

]-1]]] j] A smoothing filter H (z ^j ) is applied to each of x, y, and z! Smoothing

]-1

The filter H (z ^j ) is expressed by the following formula 7.

Drawing z- ² (7)

Note that z used in Equation 7 above indicates z conversion (hereinafter, the same applies to Equations 8 to 10 indicating filters!).

[0108] Partial images Wz, Wy, and Wx are respectively filtered in the x, y, and z directions of image S!

j]]]-ι

This is the first-order difference image obtained by applying G (z ^j ). Filter G (z ^j ) is expressed by Equation 8 below.

[Equation 5]

G (z) = z _ z 1 ¹ ■ ■ ■ (8) Partial images Wz, Wy, and Wx are more sensitive to fine fluctuations as the resolution level is smaller! / ヽ (j = 1). On the other hand, the higher the resolution level (j = 3), the higher the sensitivity to coarse fluctuations. Therefore, partial images Wz, Wy, Wx with small resolution levels are suitable for detecting small aneurysms, and partial images Wz, Wy, Wx with large resolution levels are large. Suitable for detecting aneurysms.

[0109] For S, the filter processing by the filter bank A (z ^{j + 1} ) at the resolution level j + 1 j

Will be performed. In this way, by performing repeated filtering, partial images S, Wz, Wy, and Wx are created for each resolution level j (j = 1 to 3).

[0110] Next, as shown in FIG. 20, the partial image Wz, Wy, Wx at each resolution level is filtered by the vector concentration filter GC, and the vector concentration at each resolution level is calculated. Is done. Since the method for calculating the vector concentration has been described above, a description thereof is omitted here.

[0111] The calculated vector concentration is input to the -Eural network NN.

The neural network NN outputs the output value in the range of 0 to 1 so that the higher the possibility of a cerebral aneurysm, the higher the possibility of a cerebral aneurysm, the lower the value is 0. Designed to be.

[0112] When the output value is obtained from the neural network NN, the determination unit NM generates and outputs a weighted image V based on the output value. In the weighted image V, the botacell value is set to 1 for a botacell having an output value larger than a certain threshold value (here, 0.8), and the botacell value is set to 0 for a botacell having an output value less than or equal to a threshold value 0.8. Is created. In other words, it is highly possible that a botacel for which a vector concentration degree exceeding the threshold value of 0.8 is calculated is a botacell constituting a region of a cerebral aneurysm. On the other hand, it is highly probable that a vessel cell having a vector concentration of less than or equal to the threshold value 0.8 is a vessel cell constituting a normal blood vessel region. Therefore, the weighted image is created by binarizing the botacell values with the threshold as the boundary. The weight image V is input to the reconstruction bank.

FIG. 22 is a diagram showing a reconfiguration bank.

The reconstruction bank weights each partial image s, Wz, Wy, Wx,

The original image S is reconstructed through the ruta bank S (z ^j ).

0

In the weighting process, each partial image S, Wz, Wy, Wx is multiplied by the weight image V. In other words, the value of 1 or 0 set for each botacell in the weighted image V is used as a weighting coefficient in the weighting process. Each partial image multiplied by the weight image V is input to the filter bank S (z ^j ). [0114] Filter bank S (z ^j ) is filtered by filters L (z ^j ), K (z ^j ), H (z ^j ) L (z ^j ) as shown in Fig. 23, and S, Wz , Wy, Wx and others reconstruct S.

]]] j] -ι

Here, the filters L (z ^j ) and K (z ^j ) are expressed by the following equations 9 and 10.

[Equation 6]

^{Kfz = - (-Z 3 - 5z} + 5z _1 + z "3) · · · (9)

16

L (z) = -z ² +-+ -z " ² ■ ■ ■ (10)

4 2 4

That is, S is subjected to filtering by the filter L (z ^j ) in the x, y, and z directions, respectively. For Wz, filter K (z ^j ) is applied in the X direction, and filter H (z ^j ) L (z ^j ) is applied in the y and z directions. For Wy, the filter K (z ^j ) and z direction are applied in the y direction, and the filter H (z ^j ) L (z ^j ) is applied, and for Wx, it is in the z direction! And filter) apply. And S, Wz, Wy, Wx after each filtering

By adding j j j, S is obtained.

] j-i

[0116] S is further filtered by filter bank S (z ^j_1 ).

j-l j-2 will be reconstructed. The original image S is re-created by repeating such filtering.

0 can be configured.

[0117] The reconstruction of the original image S consists of the filters H (z ^j ) and G (z ^j o) in the filter banks A (z ^j ) and S (z ^j ).

), L (z ^j ), and K (z ^j ) can reproduce the original image completely only when the combination of filters shown in FIGS. 21 and 23 is used. Therefore, as a result of reconstruction, the original image S is completely reproduced for the votacel whose voxel value is set to 1 in the weighted image V.

0

On the other hand, in the weighted image V, the original image S is not reproduced for the botacell whose botacell value is set to 0, and the botacell value becomes 0. Therefore, the GC filter bank

0

The output image S that is output is only the image area that is likely to be a cerebral aneurysm.

0

The resulting image. The image that appears in this output image S is the primary candidate for cerebral aneurysm

[0118] When the output image S is obtained, the feature amount is calculated using the output image S (steps). S105). As in the first embodiment, the feature amount is calculated by calculating the average value of the size of the candidate region, the sphericity, and the vector concentration degree in each of the botasels in the region. After secondary detection is performed using the calculated feature quantity (step S106), further tertiary detection is performed (step S107), and the final detection result is output (step S108). Since the processing of steps S105 to S108 is the same as steps S5 to S8 described with reference to FIG. 2, detailed description thereof is omitted.

[0119] In addition, the blood vessel part discrimination process is also performed in the second embodiment, but since the process contents are the same as those in the first embodiment, description of the process contents and effects is omitted.

As described above, according to the second embodiment, the vector concentration degree is calculated for each resolution level from the partial image obtained by performing the multi-resolution analysis on the head image by the GC filter bank. By calculating the vector concentration for each resolution level, it is possible to cope with detection of aneurysms of various sizes, and more accurate detection processing is possible.

[0121] In addition, a weighted image in which the botacell value is set to 1 so that the original image is reproduced only for the botacell whose vector concentration is a predetermined value equal to or greater than the threshold value, and the botacell value is set to 0 for the other botacels. Each partial image is generated and multiplied to be weighted, and the original image is reconstructed from the weighted partial images. Therefore, only the region having a high vector concentration degree, that is, the region that is highly likely to be an aneurysm is reconstructed, and a reconstructed image in which only the candidate region of the aneurysm is displayed can be obtained. By calculating the feature value using such a reconstructed image, it is possible to eliminate image elements other than the feature value force candidate area to be calculated, and it is possible to accurately calculate the feature value for the candidate area. It becomes. Therefore, the accuracy of the detection process itself can be further improved.

Industrial applicability

[0122] The present invention can be used in the field of image processing, and can be applied to a medical image processing apparatus that performs image analysis and image processing of a head image obtained by a medical imaging apparatus.

Claims

The scope of the claims

[1] Analyzing means for performing multi-resolution analysis of the head image and calculating a vector concentration for each resolution level using partial images decomposed for each resolution level;

A medical image processing apparatus comprising:

[2] The reconstruction means weights the partial image so that only an image of an area where the vector concentration level is a predetermined value is reproduced, and uses the weighted partial image to restore the original head image. The medical image processing apparatus according to claim 1, wherein the medical image processing apparatus is reconfigured

[3] The deletion unit calculates a feature amount for a candidate region of a lesion portion reproduced in the reconstructed head image, and deletes a false positive candidate region from the candidate region based on the feature amount. 3. The medical image processing apparatus according to claim 1 or 2, wherein the medical image processing apparatus is characterized in that:

[4] The feature amount is the size of the candidate area of the lesion, the sphericity, the average of the concentration of the outer region in the candidate region, or the degree of the outer region concentration, which is reproduced in the reconstructed head image. 4. The medical image processing apparatus according to claim 3, wherein the medical image processing apparatus is at least one or more of the maximum values.

[5] The lesion is an aneurysm that occurs in a blood vessel,

Comprising extraction means for extracting a blood vessel image from the head image,

5. The image forming apparatus according to claim 1, wherein the image construction unit performs multi-resolution analysis and reconstruction of the original image using the image from which the blood vessel image has been extracted. The medical image processing apparatus described in 1.

[6] The medical image processing apparatus according to any one of [1] to [5], wherein the head image is an MRI image taken by MRI.

[7] The medical image processing apparatus according to any one of [1] to [6], wherein the MRI image is an MRA image obtained by an MRA imaging method in MRI. .

[8] extraction means for extracting a blood vessel image from the head image;

An image control means for discriminating one or a plurality of blood vessel parts included in the extracted blood vessel image and attaching blood vessel part information relating to the discriminated blood vessel part to the head image. The medical image processing apparatus according to claim 1, wherein

[9] The image control means determines the position and name of the blood vessel part in the head image using a reference image in which the position and name of one or more blood vessel parts included in the blood vessel image are predetermined, 9. The medical image processing apparatus according to claim 8, wherein information on the position and name of each determined blood vessel part is attached to the head image as the blood vessel part information.

[10] The image control means performs affine transformation on the head image to substantially match the position of the blood vessel image of the head image subjected to the affine transformation with the position of the blood vessel image in the reference image. 10. The blood vessel image portion of the head image corresponding to a predetermined blood vessel portion in the matched reference image is determined as the predetermined blood vessel portion. Medical image processing apparatus.

[11] display means for displaying the head image;

Based on the blood vessel part information attached to the head image, one or more blood vessel parts included in the blood vessel image of the displayed head image are determined, and each of the blood vessel parts is displayed on the displayed head image. Display control means for identifying and displaying in

The medical image processing apparatus according to claim 9 or 10, characterized by comprising:

[12] The display control means determines the name of each identified and displayed blood vessel part based on the blood vessel part information and displays the determined name in association with each corresponding blood vessel part. The medical image processing apparatus according to claim 11.

[13] An operation means for performing a display operation is provided.

The display control means is adapted to display the operation target blood vessel portion on the head image. When a selection operation is performed through the means, only a blood vessel image corresponding to the selected blood vessel part is extracted from the head image, and only the extracted blood vessel image is displayed on the display means. The medical image processing apparatus according to claim 12.

[14] An analysis step of performing multi-resolution analysis of the head image and calculating a vector concentration for each resolution level using partial images decomposed for each resolution level;

In reconstructing the original head image from the partial image, a reconstruction step of reproducing the original image only in a candidate region of a lesion part where the calculated vector concentration degree is a predetermined value; and A deletion step of deleting a false positive candidate region that is a normal blood vessel from candidate regions of a lesion portion reproduced in the head image using

An image processing method comprising:

[15] In the reconstruction step, weighting of the partial image is performed so that only the image of the candidate region of the lesion part having a predetermined vector concentration degree is reproduced, and the weighted partial image is used to reproduce the original image. 15. The image processing method according to claim 14, wherein the head image is reconstructed.

[16] In the deletion step, a feature amount is calculated for a candidate region of a lesion portion reproduced in the reconstructed head image, and a false positive candidate region is deleted from the candidate region based on the feature amount 16. The image processing method according to claim 14 or 15, characterized in that:

[17] The feature amount is the size of the candidate area of the lesion portion reproduced in the reconstructed head image, the sphericity, the average of the outer concentration of the candidate area, or the outer concentration of the candidate area. 17. The image processing method according to claim 16, wherein the image processing method is at least one of the maximum values.

[18] The lesion is an aneurysm that occurs in a blood vessel,

An extraction step of extracting a blood vessel image from the head image,

18. The reconstruction process according to any one of claims 14 to 17, wherein in the reconstruction step, multi-resolution analysis and reconstruction of an original image are performed using an image from which the blood vessel image has been extracted. An image processing method described in 1.

[19] The image processing method according to any one of [14] to [18], wherein the head image is an MRI image taken by MRI.

[20] The image processing method according to any one of claims 14 to 19, wherein the MRI image is an MRA image captured by an MRA imaging method in MRI.

[21] An extraction step of extracting the head image force blood vessel image;

An image control step of discriminating one or a plurality of blood vessel parts included in the extracted blood vessel image and attaching blood vessel part information relating to the discriminated blood vessel part to the head image. Item 15. The image processing method according to Item 14.

[22] In the image control step, the position and name of the blood vessel part in the head image are determined using a reference image in which the position and name of one or more blood vessel parts included in the blood vessel image are determined in advance. The image processing method according to claim 21, wherein information on the position and name of each determined blood vessel part is attached to the head image as the blood vessel part information.

[23] In the image control step, the head image is affine transformed to substantially match the position of the blood vessel image of the head image subjected to the affine transformation and the position of the blood vessel image in the reference image. 23. The range according to claim 22, wherein a blood vessel image portion of the head image corresponding to a predetermined blood vessel portion corresponding to the matched reference image is determined to be the predetermined blood vessel portion. Image processing method.

[24] a display step of displaying the head image on a display means;

Based on the blood vessel part information attached to the head image, one or more blood vessel parts included in the blood vessel image of the displayed head image are determined, and each of the blood vessel parts is displayed on the displayed head image. A display control process for identifying and displaying in

24. The image processing method according to claim 22 or 23, comprising:

[25] In the display control step, the name of each identified and displayed blood vessel part is determined based on the blood vessel part information, and the determined name is displayed in association with each corresponding blood vessel part. The image processing method according to claim 24.

[26] In the display control step, the blood vessel part to be displayed is displayed on the head image as an operating means. When the selection operation is performed via the button, only the blood vessel image corresponding to the selected blood vessel part is extracted from the head image, and only the extracted blood vessel image is displayed on the display means. The image processing method according to claim 25.