WO2019216125A1

WO2019216125A1 - Learning device, method, and program for classifier for classifying infarct region, classifier for classifying infarct region, and device, method and program for classifying infarct region

Info

Publication number: WO2019216125A1
Application number: PCT/JP2019/016052
Authority: WO
Inventors: 赤堀　貞登
Original assignee: 富士フイルム株式会社
Priority date: 2018-05-09
Filing date: 2019-04-12
Publication date: 2019-11-14

Abstract

An image acquisition unit (21) acquires a CT image Bc0 and an MRI image Bm0 of the brain of a subject with a brain infarct, while an infarct region extraction unit (22) extracts an infarct region A1 from the MRI image Bm0. An alignment unit (23) aligns the CT image Bc0 with the MRI image Bm0, while an infarct region identification unit (25) identifies an infarct region A2 in the CT image Bc0 on the basis of a result of the alignment. A learning unit (25), using the infarct region identified in the CT image Bc0 as training data, performs learning about a classifier (26) for classifying an infarct region in a CT image Bc1 that has been inputted thereto.

Description

Learning device, method and program for discriminator for discriminating infarct region, discriminator for discriminating infarct region, infarct region discriminating device, method and program

The present disclosure relates to a learning device, method, and program for a discriminator that discriminates an infarct region in a brain image, a discriminator that discriminates an infarct region, and an infarct region discriminating device, method, and program.

In recent years, with the advance of medical equipment such as CT (Computed Tomography) apparatus and MRI (Magnetic Resonance Imaging) apparatus, it has become possible to perform image diagnosis using medical images with higher quality and higher resolution. In particular, in the case where the target site is the brain, a region causing cerebral vascular disorders such as cerebral infarction and cerebral hemorrhage can be identified by image diagnosis using CT images and MRI images. For this reason, various methods for supporting image diagnosis have been proposed.

For example, in JP2013-165765A, a cerebral infarction site included in an MRI diffusion weighted image (DWI: DiffusionＤWeightedifImage) is detected, and from the abnormal site of the diffusion weighted image and the diffusion weighted image of a healthy person, The position conversion data necessary for the anatomical alignment between the two is obtained, and SPECT (Single Photon) is used by using the position conversion data so that each tissue position of the patient's brain matches each tissue position of the healthy person's brain. There has been proposed a technique of converting a SPECT image taken by an Emission (Computed Tomography) apparatus and discriminating a cerebral infarction site on the SPECT image. Japanese Patent Laid-Open No. 2014-518516 proposes a method for performing diagnosis by aligning and displaying a CT image and an MRI image.

By the way, thrombolytic therapy is performed for cerebral infarction patients. However, thrombolytic therapy is applicable within 4.5 hours from the time at which cerebral infarction has not been confirmed, and the risk of bleeding after treatment increases as the infarct range expands over time. Are known. For this reason, in order to determine the suitability of thrombolytic therapy, it is necessary to quickly and appropriately determine the infarct range using medical images. Here, in the diffusion weighted image, the infarct region has a pixel value different from other regions. In particular, in the region of acute cerebral infarction, the difference in pixel value from other regions becomes significant in the diffusion weighted image. For this reason, a diffusion weighted image is often used for confirmation of the infarct region.

On the other hand, in brain diagnosis, the presence or absence of bleeding in the brain is often confirmed before confirming cerebral infarction. Since hemorrhage in the brain can be clearly confirmed in the CT image, a diagnosis using a CT image is first performed for a patient suspected of having a brain disease. However, in the CT image, the difference in pixel values between the cerebral infarction region in the acute phase and other regions is not so large, and it may be difficult to specify the infarction in the acute phase using the CT image. Many. For this reason, after diagnosis using a CT image, an MRI image is acquired to determine whether or not a cerebral infarction has occurred.

However, if an MRI image is acquired after diagnosis using a CT image to diagnose whether or not a cerebral infarction has occurred, the elapsed time from the onset of the infarction becomes longer. May increase the risk of bleeding after treatment with

The present disclosure has been made in view of the above circumstances, and an object of the present disclosure is to enable quick diagnosis of whether or not a cerebral infarction has occurred by using a CT image.

A learning apparatus for a discriminator according to the present disclosure includes an image acquisition unit that acquires a CT image of a brain of a subject developing cerebral infarction and an MRI image of the brain of the same subject as the subject;
An infarct region extraction unit that extracts an infarct region from an MRI image;
An alignment unit for aligning the CT image and the MRI image;
An infarct region specifying unit that specifies an infarct region in the CT image based on the result of the alignment;
And a learning unit that learns a discriminator for discriminating the infarct region in the input CT image using the infarct region specified in the CT image as teacher data.

In the discriminator learning device according to the present disclosure, the MRI image may be a diffusion weighted image.

Further, in the discriminator learning device according to the present disclosure, the alignment unit may extract a brain region from the CT image and perform alignment between the extracted brain region and the diffusion weighted image.

The discriminator according to the present disclosure has been learned by the discriminator learning device according to the present disclosure.

An infarct region discriminating apparatus according to the present disclosure includes an image acquisition unit that acquires a CT image of an infarct region discrimination target;
A discriminator according to the present disclosure that discriminates an infarct region in a CT image to be discriminated;

Note that the infarct region discriminating apparatus according to the present disclosure further includes a display control unit that displays the discrimination result by the discriminator on the display unit.

The discriminator learning method according to the present disclosure acquires a CT image of a brain of a subject who has developed cerebral infarction and an MRI image of the brain of the same subject as the subject,
Extract infarct area from MRI image,
Align CT image and MRI image,
Based on the alignment result, identify the infarct region in the CT image,
A discriminator for discriminating the infarct region in the input CT image is learned using the infarct region specified in the CT image as teacher data.

The infarct region determination method according to the present disclosure obtains a CT image of an infarct region determination target,
The discriminator of the present disclosure discriminates the infarct region in the CT image to be discriminated.

Note that the discriminator learning method according to the present disclosure and the infarct region determination method according to the present disclosure may be provided as a program for causing a computer to execute the method.

Another discriminator learning device according to the present disclosure includes a memory for storing instructions to be executed by a computer,
A processor configured to execute stored instructions, the processor comprising:
Obtaining a CT image of the brain of the subject developing cerebral infarction and an MRI image of the brain of the same subject as the subject,
Extract infarct area from MRI image,
Align CT image and MRI image,
Based on the alignment result, identify the infarct region in the CT image,
A process of learning a discriminator for discriminating the infarct region in the input CT image is executed using the infarct region specified in the CT image as teacher data.

Another infarct region discriminating device according to the present disclosure includes a memory for storing instructions for causing a computer to execute,
A processor configured to execute stored instructions, the processor comprising:
Obtain a CT image of the infarct area discrimination target,
By the discriminator of the present disclosure, processing for discriminating the infarct region in the CT image to be discriminated is executed.

According to the present disclosure, a CT image of the brain of a subject developing cerebral infarction and an MRI image of the brain of the same subject as the subject are acquired, an infarct region is extracted from the MRI image, and the CT image and the MRI Registration with the image is performed. Then, based on the alignment result, an infarct region in the CT image is specified, and a discriminator for discriminating the infarct region in the input CT image is learned using the specified infarct region as teacher data. Here, in the CT image, since the bleeding region has a pixel value that is greatly different from other regions, it is easy to specify the bleeding region in the CT image. However, in the CT image, the infarct region has a different pixel value from other regions, but the difference is smaller than the difference between the bleeding region and other regions. On the other hand, in the MRI image, the infarct region has a pixel value that is significantly different from other regions. For this reason, if the MRI image and CT image of the brain of the subject who has developed cerebral infarction are aligned, the infarct region in the CT image can be specified based on the infarct region in the MRI image. Therefore, by learning the discriminator using the identified infarct region as teacher data, the infarct region in the input CT image can be discriminated by the learned discriminator. Thereby, it is possible to discriminate not only the bleeding region of the brain but also the infarct region using only the CT image. Therefore, according to the present disclosure, it is possible to quickly diagnose whether cerebral infarction has developed.

Hardware configuration diagram showing an outline of a diagnosis support system to which a learning device, a discriminator, and an infarct region discrimination device according to an embodiment of the present disclosure are applied The figure which shows schematic structure of the infarct area | region discrimination apparatus by this embodiment. The figure which shows the example of an MRI image The figure for demonstrating position alignment with CT image and MRI image The figure for demonstrating specification of the infarction area | region in CT image A figure showing an example of display of discrimination results A flowchart showing processing performed at the time of learning in the present embodiment The flowchart which shows the process performed at the time of discrimination | determination of an infarction area | region in this embodiment.

Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. FIG. 1 is a hardware configuration diagram illustrating an outline of a diagnosis support system to which a discriminator learning device, a discriminator, and an infarct region discrimination device according to an embodiment of the present disclosure are applied. As shown in FIG. 1, in the diagnosis support system, the infarct region discriminating device 1, the three-dimensional image capturing device 2, and the image storage server 3 according to the present embodiment are connected in a communicable state via a network 4. Yes. The infarct region discriminating device 1 includes the learning device and the discriminator according to the present embodiment.

The 3D image capturing apparatus 2 is an apparatus that generates a 3D image representing a part as a medical image by capturing the part to be diagnosed of the subject. The medical image generated by the 3D image capturing apparatus 2 is transmitted to the image storage server 3 and stored. In the present embodiment, the diagnosis target part of the patient who is the subject is the brain, and the three-dimensional imaging apparatus 2 is the CT apparatus 2A and the MRI apparatus 2B. Then, the CT apparatus 2A generates a three-dimensional CT image Bc0 including the subject's brain, and the MRI apparatus 2B generates a three-dimensional MRI image Bm0 including the subject's brain. In the present embodiment, the MRI image Bm0 is a diffusion weighted image. In this embodiment, the CT image Bc0 is a non-contrast CT image acquired by performing imaging without using a contrast agent. However, a contrast CT acquired by performing imaging using a contrast agent. An image may be used.

The image storage server 3 is a computer that stores and manages various data, and includes a large-capacity external storage device and database management software. The image storage server 3 communicates with other devices via a wired or wireless network 4 to transmit and receive image data and the like. Specifically, various types of data including image data of CT images and MRI images generated by the three-dimensional image capturing apparatus 2 are acquired via a network, stored in a recording medium such as a large-capacity external storage device, and managed. Note that the storage format of the image data and the communication between the devices via the network 4 are based on a protocol such as DICOM (Digital-Imaging-and-Communication-in-Medicine).

The infarct region discriminating apparatus 1 is obtained by installing the learning program and the infarct region discriminating program of the present disclosure on one computer. The computer may be a workstation or personal computer directly operated by a doctor who performs diagnosis, or may be a server computer connected to them via a network. The learning program and the infarct region discrimination program are recorded and distributed on a recording medium such as a DVD (Digital Versatile Disc) or a CD-ROM (Compact Disc Read Only Memory), and are installed in the computer from the recording medium. Alternatively, it is stored in a storage device of a server computer connected to a network or a network storage in a state where it can be accessed from the outside, and is downloaded and installed on a computer used by a doctor upon request.

FIG. 2 is a diagram showing a schematic configuration of the infarct region discriminating apparatus according to the present embodiment realized by installing a learning program and an infarct region discriminating program in a computer. As shown in FIG. 2, the infarct region discriminating apparatus 1 includes a CPU (Central Processing Unit) 11, a memory 12, and a storage 13 as a standard workstation configuration. The infarct region discriminating apparatus 1 is connected with a display 14 and an input unit 15 such as a keyboard and a mouse. The display 14 corresponds to a display unit.

The storage 13 is composed of a hard disk drive or the like, and stores various information including medical images of the subject acquired from the image storage server 3 via the network 4 and information necessary for processing.

Further, the memory 12 stores a learning program and an infarct region discrimination program. The learning program executes an image acquisition process for acquiring a CT image Bc0 and an MRI image Bm0 of the brain of a subject who has developed a cerebral infarction, and an infarct area extraction process for extracting an infarct area from the MRI image Bm0. , An alignment process for aligning the CT image Bc0 and the MRI image Bm0, an infarct area specifying process for specifying an infarct area in the CT image Bc0 based on the result of the alignment, and an infarct area specified in the CT image Bc0 A learning process for learning a discriminator for discriminating an infarct region in the input CT image Bc1 is defined as teacher data. In addition, the infarct region determination program executes, as processing to be executed by the CPU 11, an image acquisition process for acquiring a CT image Bc1 that is a determination target of the infarct region, a determination process that determines an infarct region in the CT image Bc1 of the determination target, and a determination result. A display control process to be displayed on the display 14 is defined.

When the CPU 11 executes these processes according to the program, the computer acquires the image acquisition unit 21, the infarct region extraction unit 22, the alignment unit 23, the infarct region specifying unit 24, the learning unit 25, the discriminator 26, and display control. It functions as the unit 27. Here, the image acquisition unit 21, the infarct region extraction unit 22, the alignment unit 23, the infarct region specifying unit 24, and the learning unit 25 constitute a discriminator learning device of the present embodiment. Further, the discriminator 26 and the display control unit 27 constitute an infarct region discriminating apparatus of the present embodiment.

In the present embodiment, the CPU 11 executes the function of each unit by the learning program and the infarct region determination program. However, as a general-purpose processor that functions as various processing units by executing software, the CPU 11 In addition, a programmable logic device (PLD) which is a processor whose circuit configuration can be changed after manufacturing FPGA (Field Programmable Gate Array) or the like can be used. Further, the processing of each unit may be executed by a dedicated electric circuit or the like that is a processor having a circuit configuration designed exclusively for executing specific processing such as ASIC (Application Specific Specific Integrated Circuit).

One processing unit may be composed of one of these various processors, or a combination of two or more processors of the same type or different types (for example, a plurality of FPGAs or a combination of a CPU and an FPGA). It may be configured. Further, the plurality of processing units may be configured by one processor. As an example of configuring a plurality of processing units with one processor, first, as represented by a computer such as a client or a server, one processor is configured with a combination of one or more CPUs and software. There is a form in which the processor functions as a plurality of processing units. Second, as represented by a system-on-chip (SoC), a form of using a processor that realizes the functions of the entire system including a plurality of processing units with a single IC (integrated circuit) chip. is there. As described above, various processing units are configured using one or more of the various processors as a hardware structure.

Further, the hardware structure of these various processors is more specifically an electric circuit (circuitry) in which circuit elements such as semiconductor elements are combined.

The image acquisition unit 21 acquires the CT image Bc0 and the MRI image Bm0 of the brain of the subject who has developed cerebral infarction from the image storage server 3 for learning by the discriminator 26. In addition, in order to determine the infarct region, a CT image Bc1 that is a determination target of the infarct region is acquired from the image storage server 3. If the CT image Bc0, CT image Bc1, and MRI image Bm0 are already stored in the storage 13, the image acquisition unit 21 acquires the CT image Bc0, CT image Bc1, and MRI image Bm0 from the storage 13. It may be. The image acquisition unit 21 acquires CT images Bc0 and MRI images Bm0 for a large number of subjects for learning by the discriminator 26 described later.

The infarct region extraction unit 22 extracts the infarct region of the brain from the MRI image Bm0. FIG. 3 is a diagram showing an example of the MRI image Bm0. Although the MRI image Bm0 is a three-dimensional image, for the sake of explanation, the MRI image Bm0 will be described using a two-dimensional tomographic image on one tomographic plane of the MRI image Bm0. As shown in FIG. 3, the MRI image Bm0, which is a diffusion weighted image, is an image including only the brain parenchyma from which the skull has been erased. As shown in FIG. 3, in the MRI image Bm0 that is a diffusion weighted image, the infarct region has a larger pixel value (lower density) than other regions. The infarct region extraction unit 22 extracts, for example, a region having a pixel value larger than a predetermined threshold value as an infarct region in the MRI image Bm0. Note that an infarct region may be extracted from the MRI image Bm0 using a discriminator that has been trained to extract an infarct region. Thereby, the infarct region A1 shown in FIG. 3 is extracted.

The alignment unit 23 performs alignment between the CT image Bc0 and the MRI image Bm0. FIG. 4 is a diagram for explaining alignment between the CT image Bc0 and the MRI image Bm0. The CT image Bc0 and the MRI image Bm0 are both three-dimensional images. However, for the sake of explanation, the CT image Bc0 and the MRI image Bm0 will be described using a two-dimensional tomographic image on one corresponding tomographic plane of the CT image Bc0 and the MRI image Bm0. As shown in FIG. 4, the shape of the brain is substantially the same in the same subject. On the other hand, in the MRI image Bm0, the infarct region has a larger pixel value (lower density) than other regions. In contrast, in the CT image Bc0, the difference in pixel values between the infarct region and other regions is not as great as that of the MRI image Bm0. Note that, unlike the MRI image Bm0 that is a diffusion weighted image, the CT image Bc0 includes the skull and the brain parenchyma. Therefore, the alignment unit 23 extracts a brain parenchymal region from the CT image Bc0 as a brain region, and performs alignment between the extracted brain region and the MRI image Bm0.

In the present embodiment, the alignment unit 23 aligns one of the CT image Bc0 and the MRI image Bm0 with respect to the other by non-rigid alignment. In the present embodiment, the CT image Bc0 is aligned with the MRI image Bm0. However, the MRI image Bm0 may be aligned with the CT image Bc0.

Note that the non-rigid registration is a method in which feature points in the CT image Bc0 are nonlinearly converted into corresponding points corresponding to the feature points in the MRI image Bm0 using functions such as B spline and thin plate spline, for example. Although it can be used, it is not limited to this.

The infarct region specifying unit 24 specifies the infarct region in the CT image Bc0 based on the alignment result by the alignment unit 23. FIG. 5 is a diagram for explaining the specification of the infarct region in the CT image Bc0. Note that the CT image Bc0 shown in FIG. 5 is aligned with the MRI image Bm0. Since the CT image Bc0 is aligned with the MRI image Bm0, the infarct region specifying unit 24 specifies the voxel region of the CT image Bc0 corresponding to the infarct region A1 extracted from the MRI image Bm0 as the infarct region A2.

The learning unit 25 learns the discriminator 26 for discriminating the infarct region in the input CT image using the infarct region specified in the CT image Bc0 as teacher data. In the present embodiment, the discriminator 26 discriminates the infarct region in the CT image Bc1 when the CT image Bc1 to be discriminated is input. Specifically, the discriminator 26 classifies each voxel position of the CT image Bc1 to be discriminated into two classes other than the infarct region and the infarct region, and discriminates the infarct region. For this reason, the learning unit 25 acquires a feature amount in a region having a predetermined size (for example, 3 × 3) from the infarct region A2 specified in the CT images Bc0 of a plurality of subjects, and the acquired feature amount Is input to the discriminator 26, and learning of the discriminator 26, that is, machine learning, is performed so that a discrimination result indicating that the region is an infarct region is output.

When the CT image Bc1 is input by performing learning in this way, the voxel values of the CT image Bc1 are classified into an infarct region and a region outside the infarct region, and the infarct region in the CT image Bc1 to be determined Is generated.

The display control unit 27 displays the discrimination result by the discriminator 26 of the CT image Bc1 to be discriminated on the display 14. FIG. 6 is a diagram illustrating an example of display of the discrimination result. FIG. 6 shows a tomographic image in one cross section of the CT image Bc1 to be discriminated. As shown in FIG. 6, in the discrimination result, the infarct region A3 is specified in the CT image Bc1 to be discriminated.

As a machine learning method, a support vector machine (SVM (SupportＳMachine)), a deep neural network (DNN (Deep Neural Network)), a convolutional neural network (CNN (Convolutional Neural Network)), and a recurrent neural network ( RNN (Recurrent Neural Network)) or the like can be used.

Next, processing performed in the present embodiment will be described. FIG. 7 is a flowchart showing processing performed during learning in the present embodiment. First, the image acquisition unit 21 acquires the CT image Bc0 and MRI image Bm0 of the brain of the subject who has developed cerebral infarction (step ST1), and the infarct region extraction unit 22 extracts the infarct region A1 from the MRI image Bm0. Extract (step ST2). Next, the alignment unit 23 performs alignment between the CT image Bc0 and the MRI image Bm0 (step ST3), and the infarct region specifying unit 24 specifies the infarct region A2 in the CT image Bc0 based on the alignment result. (Step ST4). Then, the learning unit 25 learns the discriminator 26 that discriminates the infarct region in the input CT image Bc1 using the infarct region specified in the CT image Bc0 as teacher data (step ST6), and ends the process.

FIG. 8 is a flowchart showing a process performed when determining an infarct region in the present embodiment. First, the image acquisition unit 21 acquires a CT image Bc1 to be determined (step ST11), and the determiner 26 determines an infarct region in the CT image to be determined (step ST12). And the display control part 27 displays a discrimination | determination result on the display 14 (step ST13), and complete | finishes a process.

As described above, in the present embodiment, the CT image Bc0 and MRI image Bm0 of the brain of the subject developing cerebral infarction are acquired, the infarct region is extracted from the MRI image Bm0, and the CT image Bc0 and the MRI image Bm0 are extracted. Align with. Then, based on the alignment result, the infarct region A1 in the CT image Bc0 is identified, and the discriminator 26 for discriminating the infarct region in the input CT image Bc1 is learned using the identified infarct region A1 as teacher data. I did it.

Here, in the CT image, since the bleeding region has a pixel value that is significantly different from other regions, it is easy to specify the bleeding region in the CT image. However, in the CT image, the infarct region has a different pixel value from other regions, but the difference is smaller than the difference between the bleeding region and other regions. On the other hand, in the MRI image, the infarct region has a pixel value that is significantly different from other regions. In particular, in the diffusion weighted image, the infarct region in the acute phase has a pixel value that is significantly different from other regions. For this reason, if the MRI image Bm0 and the CT image Bc0 of the brain of the subject developing cerebral infarction are aligned, the infarct region A2 in the CT image Bc0 is specified based on the infarct region in the MRI image Bm0. Can do. Therefore, by learning the discriminator 26 using the identified infarct region A2 as teacher data, the learned discriminator 26 can discriminate the infarct region in the CT image Bc1 to be discriminated. Thereby, it is possible to discriminate not only the bleeding region of the brain but also the infarct region using only the CT image. Therefore, according to the present embodiment, it is possible to quickly diagnose whether or not cerebral infarction has developed.

In the above embodiment, a diffusion weighted image is used as the MRI image Bm0. However, an MRI image other than the diffusion weighted image may be used. For example, a T1-weighted image, a T2-weighted image, or a FLAIR (FLuid-Attenuated Inversion Recovery) image may be used. A plurality of MRI images may be used in combination.

In the above embodiment, a non-contrast CT image or a contrast CT image is used as the CT image Bc0 used for learning by the discriminator 26. However, both the contrast CT image and the non-contrast CT image are learned by the discriminator 26. You may make it use for. By using the discriminator 26 learned in this way, the infarct region can be discriminated regardless of whether the CT image to be discriminated is a contrast CT image or a non-contrast CT image.

In the above embodiment, the infarct region discriminating apparatus 1 includes the learning apparatus, but the present invention is not limited to this. That is, the diagnosis support system includes an image acquisition unit 21, an infarct region extraction unit 22, an alignment unit 23, an infarct region specifying unit 24, and a learning unit 25 separately from the infarct region determination device 1. You may make it provide the learning apparatus which performs learning. In this case, the infarct region discriminating apparatus 1 includes only the image acquisition unit 21, the discriminator 26, and the display control unit 27.

DESCRIPTION OF SYMBOLS 1 Infarction area | region discrimination device 2 3D imaging device 3 Image storage server 4 Network 11 CPU
DESCRIPTION OF SYMBOLS 12 Memory 13 Storage 14 Display 15 Input part 21 Image acquisition part 22 Infarct area extraction part 23 Positioning part 24 Infarct area specific part 25 Learning part 26 Discriminator 27 Display control part A1, A2, A3 Infarct area Bc0 CT image Bc1 Discrimination object CT image of Bm0 MRI image

Claims

An image acquisition unit for acquiring a CT image of the brain of a subject developing cerebral infarction and an MRI image of the brain of the same subject as the subject;
An infarct region extraction unit for extracting an infarct region from the MRI image;
An alignment unit for aligning the CT image and the MRI image;
Based on the result of the alignment, an infarct region specifying unit that specifies the infarct region in the CT image,
A learning device for a discriminator, comprising: a learning unit that learns a discriminator that discriminates an infarct region in an input CT image using the infarct region specified in the CT image as teacher data.
The discriminator learning device according to claim 1, wherein the MRI image is a diffusion weighted image.
The discriminator learning device according to claim 2, wherein the alignment unit extracts a brain region from the CT image and aligns the extracted brain region with the diffusion weighted image.
A discriminator learned by the discriminator learning device according to any one of claims 1 to 3.
An image acquisition unit for acquiring a CT image of a discrimination target of the infarct region;
An infarct region discriminating apparatus comprising the discriminator according to claim 4, which discriminates an infarct region in the CT image to be discriminated.
The infarct region discriminating apparatus according to claim 5, further comprising a display control unit that displays a discrimination result by the discriminator on a display unit.
Obtaining a CT image of the brain of a subject developing cerebral infarction and an MRI image of the brain of the same subject as the subject;
Extracting an infarct region from the MRI image;
Align the CT image and the MRI image,
Identifying the infarct region in the CT image based on the alignment result;
A learning method for a discriminator that learns a discriminator for discriminating an infarct region in an input CT image using the infarct region specified in the CT image as teacher data.
Obtain a CT image of the infarct area discrimination target,
An infarct region discriminating method for discriminating an infarct region in the CT image to be discriminated by the discriminator according to claim 4.
A procedure for obtaining a CT image of the brain of a subject developing cerebral infarction and an MRI image of the brain of the same subject as the subject;
A procedure for extracting an infarct region from the MRI image;
A procedure for aligning the CT image and the MRI image;
A procedure for identifying the infarct region in the CT image based on the alignment result;
A discriminator learning program for causing a computer to execute a procedure for learning a discriminator for discriminating an infarct region in an input CT image using the infarct region specified in the CT image as teacher data.
A procedure for obtaining a CT image of an infarct area discrimination target;
An infarct region discrimination program for causing a computer to execute a procedure for discriminating an infarct region in the CT image to be discriminated by the discriminator according to claim 4.