JP7740663B2 - Image processing device, method and program - Google Patents

Description

This disclosure relates to an image processing device, method, and program.

In recent years, advances in medical equipment such as CT (Computed Tomography) and MRI (Magnetic Resonance Imaging) devices have made it possible to perform image diagnosis using higher-quality, higher-resolution medical images. Furthermore, computer-aided diagnosis (CAD) has been put into practical use, analyzing medical images to derive information such as the probability of lesion presence and location, and presenting this information to radiologists and other physicians. For example, Patent Document 1 proposes a method for identifying the area of a target organ and extracting areas suspected of being abnormal based on diagnostic criteria established for that organ.

JP 2009-219610 A

When diagnosing a target organ using CAD, it is important to identify changes in the organ's shape, such as atrophy or swelling. For example, if the target organ is the pancreas, the development of a pancreatic tumor will cause the pancreatic parenchyma surrounding the tumor to swell, and the pancreatic parenchyma outside the tumor will atrophy. For this reason, it is important to determine the presence or absence of pancreatic disease by focusing on the diameter of the pancreas contained in medical images.

However, in cases where atrophy has progressed to an extreme degree, the pancreas may become so thin that parts of the pancreas may disappear from medical images. In such cases, it is impossible to determine whether or not there is an abnormality.

The present invention was developed in consideration of the above circumstances, and aims to make it possible to determine whether or not there is an abnormality in the pancreas even if part of the pancreas has disappeared.

An image processing device according to a first aspect of the present disclosure includes at least one processor,
The processor extracts a region of the pancreas from the medical image;
Extracting a plurality of sub-regions from the extracted pancreas;
determining the presence or absence of each of a plurality of sub-regions;
Based on the result of the determination, a classifier to be used for determining the presence or absence of an abnormality in the pancreas is selected from the plurality of types of classifiers;
The selected classifier is used to determine whether or not there is an abnormality in the pancreas.

An image processing device according to a second aspect of the present disclosure is the image processing device according to the first aspect, wherein the multiple classifiers may include a classifier that determines the presence or absence of an abnormality based on a combination of the presence or absence of multiple sub-regions.

An image processing device according to a third aspect of the present disclosure may be the image processing device according to the second aspect, in which the classifier is trained using images consisting of combinations of multiple sub-regions with and without presence as training data.

An image processing device according to a fourth aspect of the present disclosure is the image processing device according to the first aspect, wherein the plurality of classifiers may include a classifier that determines the presence or absence of an abnormality based on subregions determined to have an abnormality.

An image processing device according to a fifth aspect of the present disclosure is an image processing device according to any one of the second to fourth aspects, wherein the multiple classifiers determine the presence or absence of an abnormality based on the characteristics of the subregions.

An image processing device according to a sixth aspect of the present disclosure is the image processing device according to the fifth aspect, wherein the feature of the subregion may be at least one of the diameter, size, and texture of the subregion.

An image processing device according to a seventh aspect of the present disclosure is an image processing device according to any one of the first to sixth aspects, wherein the plurality of classifiers may include a classifier that determines that there is an abnormality in the pancreas when at least one of the plurality of subregions is absent.

An image processing device according to an eighth aspect of the present disclosure is an image processing device according to any one of the first to seventh aspects, wherein the multiple sub-regions may be the head, body, and tail of the pancreas.

The image processing method according to the present disclosure includes extracting a pancreas region from a medical image;
Extracting a plurality of sub-regions from the extracted pancreas;
determining the presence or absence of each of a plurality of sub-regions;
Based on the result of the determination, a classifier to be used for determining the presence or absence of an abnormality in the pancreas is selected from the plurality of types of classifiers;
The selected classifier is used to determine whether or not there is an abnormality in the pancreas.

The image processing program according to the present disclosure includes a procedure for extracting a pancreas region from a medical image;
extracting a plurality of sub-regions from the extracted pancreas;
determining the presence or absence of each of a plurality of sub-regions;
a step of selecting a classifier to be used for determining the presence or absence of pancreatic abnormality from among a plurality of types of classifiers based on the result of the determination;
The computer is caused to execute a procedure for determining the presence or absence of an abnormality in the pancreas using the selected classifier.

According to the present disclosure, it is possible to determine whether or not there is an abnormality in the pancreas even if part of the pancreas has disappeared.

FIG. 1 is a diagram illustrating a schematic configuration of a diagnosis support system to which an image processing device according to an embodiment of the present disclosure is applied; FIG. 1 is a diagram showing the hardware configuration of an image processing apparatus according to an embodiment of the present invention; Functional configuration diagram of an image processing apparatus according to this embodiment Diagram for explaining extraction of the pancreas region Diagram illustrating division of the pancreas into head, body, and tail A medical image showing a pancreas with a missing tail. FIG. 1 shows a classifier used in this embodiment. FIG. 10 is a diagram showing a display screen for determining whether or not an abnormality has occurred. A flowchart showing the processing performed in this embodiment FIG. 10 is a diagram showing another example of extracting a subregion.

Embodiments of the present disclosure will now be described with reference to the drawings. First, the configuration of a medical information system to which an image processing device according to this embodiment is applied will be described. Figure 1 is a diagram showing the general configuration of a medical information system. In the medical information system shown in Figure 1, a computer 1 incorporating an image processing device according to this embodiment, an imaging device 2, and an image storage server 3 are connected in a communicable state via a network 4.

Computer 1 incorporates the image processing device of this embodiment and has the image processing program of this embodiment installed. Computer 1 may be a workstation or personal computer operated directly by the physician making the diagnosis, or it may be a server computer connected to either of these via a network. The image processing program is stored in an externally accessible state in the storage device of the server computer connected to the network, or in network storage, and is downloaded and installed on computer 1 used by the physician upon request. Alternatively, the program may be recorded and distributed on a recording medium such as a DVD (Digital Versatile Disc) or CD-ROM (Compact Disc Read Only Memory), and then installed on computer 1 from that recording medium.

The imaging device 2 is a device that captures images of the area of the subject to be diagnosed and generates a three-dimensional image of that area. Specifically, it is a CT device, MRI device, PET (Positron Emission Tomography) device, etc. The three-dimensional image composed of multiple tomographic images generated by the imaging device 2 is sent to and stored in the image storage server 3. In this embodiment, the imaging device 2 is a CT device, and generates a three-dimensional CT image of the subject's abdomen. The acquired CT image may be a contrast-enhanced CT image or a non-contrast-enhanced CT image.

The image storage server 3 is a computer that stores and manages various data and is equipped with 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, sending and receiving image data, etc. Specifically, it acquires various data, including image data of CT images generated by the imaging device 2, via the network and stores and manages it on recording media such as a large-capacity external storage device. The storage format of the image data and communication between devices via the network 4 are based on protocols such as DICOM (Digital Imaging and Communication in Medicine).

Next, an image processing device according to this embodiment will be described. Figure 2 is a diagram showing the hardware configuration of the image processing device according to this embodiment. As shown in Figure 2, the image processing device 20 includes a CPU (Central Processing Unit) 11, non-volatile storage 13, and memory 16 as a temporary storage area. The image processing device 20 also includes a display 14 such as an LCD display, input devices 15 such as a keyboard and mouse, and a network I/F (Interface) 17 connected to the network 4. The CPU 11, storage 13, display 14, input devices 15, memory 16, and network I/F 17 are connected to a bus 18. The CPU 11 is an example of a processor in this disclosure.

Storage 13 is realized by a hard disk drive (HDD), a solid state drive (SSD), flash memory, etc. An image processing program 12 is stored in storage 13 as a storage medium. CPU 11 reads image processing program 12 from storage 13, expands it in memory 16, and executes the expanded image processing program 12.

Next, the functional configuration of the image processing device according to this embodiment will be described. Figure 3 is a diagram showing the functional configuration of the image processing device according to this embodiment. As shown in Figure 3, the image processing device 20 includes an image acquisition unit 21, a first extraction unit 22, a second extraction unit 23, a first determination unit 24, a selection unit 25, a second determination unit 26, and a display control unit 27. When the CPU 11 executes the image processing program 12, the CPU 11 functions as the image acquisition unit 21, the first extraction unit 22, the second extraction unit 23, the first determination unit 24, the selection unit 25, the second determination unit 26, and the display control unit 27.

The image acquisition unit 21 acquires the medical image G0 to be processed from the image storage server 3 in response to instructions from the operator via the input device 15. In this embodiment, the medical image G0 is a CT image consisting of multiple tomographic images including the abdomen of a human body.

The first extraction unit 22 extracts the pancreas region from the medical image G0. To this end, the first extraction unit 22 has a semantic segmentation model (hereinafter referred to as an SS (Semantic Segmentation) model) that has been machine-learned to extract the pancreas from the medical image G0. As is well known, an SS model is a machine learning model that outputs an output image in which each pixel of an input image is assigned a label representing an extracted object (class). In this embodiment, the input image is a tomographic image that constitutes the medical image G0, the extracted object is the pancreas, and the output image is an image in which the pancreas region is labeled. The SS model is constructed using a convolutional neural network (CNN) such as Residual Networks (ResNet) or U-shaped Networks (U-Net).

As a result, the first extraction unit 22 extracts the region of the pancreas 30 included in the medical image G0 as shown in Figure 4. Note that Figure 4 shows one tomographic image D0 included in the medical image G0.

Extraction of the pancreas is not limited to using the SS model. Any method for extracting the pancreas from the medical image G0 can be applied, such as template matching or threshold processing of CT values.

The second extraction unit 23 extracts multiple sub-regions from the pancreas extracted by the first extraction unit 22. In this embodiment, the second extraction unit 23 divides the pancreas region into a head, body, and tail, and extracts each of the head, body, and tail of the pancreas region as a sub-region.

Figure 5 is a diagram illustrating the division of the pancreas into the head, body, and tail. Figure 5 shows the pancreas as viewed from the front of the human body. In the following description, up, down, left, and right refer to a front view of a human body in an upright position. As shown in Figure 5 , when viewed from the front, veins 31 and arteries 32 run parallel to each other vertically at intervals behind the pancreas 30. Anatomically, the pancreas 30 is divided into the head (left side of vein 31), the body (between vein 31 and artery 32), and the tail (right side of artery 32). Therefore, in this embodiment, the second extraction unit 23 divides the pancreas 30 into three small regions: the head 33, the body 34, and the tail 35, based on the vein 31 and the artery 32. The boundaries of the head 33, the body 34, and the tail 35 are based on the boundary definitions set forth in "Pancreatic Cancer Treatment Guidelines, 7th Edition, Supplementary Edition, Edited by the Japan Pancreas Society, p. 12, September 2020." Specifically, the left edge of the vein 31 (the right edge of the vein 31 when viewed from the front) is the boundary between the head 33 and the body 34, and the left edge of the artery 32 (the right edge of the artery 32 when viewed from the front) is the boundary between the body 34 and the tail 35.

For segmentation, the second extraction unit 23 extracts veins 31 and arteries 32 near the pancreas 30 in the medical image G0. The second extraction unit 23 extracts vascular regions and their core lines (i.e., central axes) from the region near the pancreas 30 in the medical image G0, for example, using techniques described in Japanese Patent Application Laid-Open Nos. 2010-200925 and 2010-220732. This technique first calculates the positions and major axis directions of multiple candidate points constituting the vascular core lines based on the values of the voxel data constituting the medical image G0. Alternatively, a Hessian matrix is calculated for the medical image G0, and the eigenvalues of the calculated Hessian matrix are analyzed to calculate the position information and major axis directions of multiple candidate points constituting the vascular core lines. Then, feature values representing the likelihood of blood vessels are calculated for the voxel data around the candidate points, and the unit determines whether the voxel data represents a blood vessel based on the calculated feature values. This extracts blood vessel regions and their core lines from the medical image G0. The second extraction unit 23 divides the pancreas 30 into a head 33, body 34, and tail 35 using the left edge (the right edge when the body is viewed from the front) of the extracted vein 31 and artery 32 as a reference, and extracts each of the head 33, body 34, and tail 35 as a sub-region.

When a tumor develops in the pancreas, the pancreatic parenchyma around the tumor swells, and the pancreatic parenchyma outside the tumor shrinks. If the atrophy of the pancreatic parenchyma progresses to an extreme extent, the pancreas 30 may become so thin that it cannot be seen in the medical image G0, resulting in some areas of the pancreas disappearing from the medical image G0. Also, congenital malformations may cause parts of the pancreas 30 to be missing. In such cases, it is not possible to extract all of the subregions of the head 33, body 34, and tail 35 from the medical image G0. For this reason, the second extraction unit 23 extracts only the subregions that can be extracted.

Note that the extraction of the head 33, body 34, and tail 35 as sub-regions from the pancreas 30 is not limited to the above method. For example, by using a segmentation model that has been machine-learned to extract the head 33, body 34, and tail 35 from the pancreas 30, the pancreas 30 may be divided into the head 33, body 34, and tail 35, and the head 33, body 34, and tail 35 may be extracted as sub-regions. In this case, the segmentation model can be trained by preparing multiple sets of training data consisting of pairs of training images including the pancreas and mask images obtained by dividing the pancreas into the head, body, and tail based on the boundary definition described above.

The first determination unit 24 determines the presence or absence of each of the multiple subregions. Specifically, the first determination unit 24 determines the presence or absence of the head 33, body 34, and tail 35. Here, a normal pancreas includes all of the subregions of the head 33, body 34, and tail 35, as shown in Figures 4 and 5. On the other hand, as described above, if the atrophy of the pancreatic parenchyma progresses to an extreme extent, the pancreas becomes so thin that it is not visible in the medical image G0, and the pancreatic parenchyma disappears from the medical image G0. For example, if the tail 35 of the pancreas 30 is extremely atrophied, the tail 35 of the pancreas 30 disappears from the medical image G0, as shown in Figure 6. Note that the missing tail 35 is indicated by a dashed line in Figure 6. In this case, the tail 35 will not be extracted by the second extraction unit 23. Furthermore, if the tail 35 is missing due to a congenital malformation, the tail 35 will not be extracted by the second extraction unit 23.

If the tail 35 is not extracted by the second extraction unit 23 in this way, the first determination unit 24 determines that the head 33 is present, the body 34 is present, and the tail 35 is absent.

The selection unit 25 selects a classifier to be used to determine the presence or absence of pancreatic abnormalities from multiple types of classifiers based on the determination result by the first determination unit 24. The classifier in this embodiment is a classifier for determining the presence or absence of pancreatic abnormalities. In this embodiment, multiple classifiers are prepared according to combinations of the presence or absence of subregions.

In this embodiment, as shown in Figure 7, a first classifier 41, a second classifier 42, and a third classifier 43 are prepared. The first classifier 41 is a classifier for pancreases from which the head has not been extracted, the second classifier 42 is a classifier for pancreases from which the body has not been extracted, and the third classifier 43 is a classifier for pancreases from which the tail has not been extracted.

As described above, if the tail 35 is missing, the first determination unit 24 determines that the head 33 is present, the body 34 is present, and the tail 35 is absent. Therefore, the selection unit 25 selects the third classifier 43 from the first to third classifiers 41 to 43.

Here, the first to third classifiers 41 to 43 are made up of convolutional neural networks that have undergone machine learning using medical images consisting of combinations of the presence and absence of multiple subregions as training data, so as to output an evaluation value indicating the presence or absence of a pancreatic abnormality based on the characteristics of each subregion. The evaluation value indicating the presence or absence of a pancreatic abnormality output by the first to third classifiers 41 to 43 is the probability of indicating a pancreatic abnormality, and is a value between 0 and 1.

Here, the first classifier 41 for the pancreas, where the head has not been extracted, is constructed by training a neural network using multiple training data sets, in which the features of the head, body, and tail of the pancreas extracted from medical images in which the head of the pancreas has disappeared are used as training data, and the presence or absence of abnormalities is used as correct answer data.

The second classifier 42 for pancreases where the body has not been extracted is constructed by training a neural network using multiple training data sets, each containing the characteristics of the head, body, and tail of the pancreas extracted from a medical image in which the body of the pancreas has disappeared, as training data, and the presence or absence of abnormalities as correct answer data.

The third classifier 43 for pancreases where the tail has not been extracted is constructed by training a neural network using multiple training data sets, in which the features of the head, body, and tail of the pancreas extracted from medical images in which the tail of the pancreas has disappeared are used as training data, and the presence or absence of abnormalities is used as correct answer data.

Features of a subregion include at least one of the diameter, size, and texture of the subregion. The diameter of the subregion can be the diameter of a cross section intersecting the long axis of the pancreas. Because the diameter of the pancreas varies at each position along the long axis of the pancreas, multiple cross sections intersecting the long axis of the pancreas are set at predetermined intervals along the long axis of the pancreas, and a representative value of the diameters in the multiple cross sections (e.g., maximum, minimum, median, and average) can be used as the diameter of the subregion. Because the cross sections intersecting the long axis of the pancreas are not circular, a representative value of the diameters in multiple directions intersecting the long axis of the pancreas (e.g., maximum, minimum, median, and average) can be used as the diameter of the subregion. The size of the subregion can be calculated from the number of voxels in the subregion and the spacing between voxels in the medical image G0. The texture of the subregion is the pixel value of each pixel in the subregion in the medical image G0 (CT value in the case of a CT image).

In medical image G0, there are two situations in which a subregion of the pancreas has disappeared: (1) it has atrophied and disappeared as a result of the progression of a pancreatic disease (pathological disappearance), and (2) it has disappeared due to a congenital malformation (i.e., it was never present in the first place) (congenital disappearance). Considering these two cases, two patterns are possible: either only pathological disappearance is considered to be abnormal, or both pathological disappearance and congenital disappearance are considered to be abnormal. For this reason, when training classifiers 41-43, it is necessary to decide which of the two patterns to consider to be abnormal and train the neural network accordingly.

When training is performed using training data that considers only pathological loss to be abnormal, the first to third classifiers 41 to 43 determine whether a missing subregion is pathological or congenital based on the characteristics of the multiple subregions, and determine whether an abnormality exists. On the other hand, when training is performed using training data that considers both pathological and congenital loss to be abnormal, the first to third classifiers 41 to 43 will determine, based on the characteristics of the multiple input subregions, that the subregion has disappeared and is therefore abnormal, outputting an evaluation value close to 1.

Note that the first to third classifiers 41 to 43 may determine abnormal regions in the pancreas in addition to evaluating the presence or absence of abnormalities. The abnormal region determination result is an evaluation value that represents the probability that each pixel in the medical image G0 is an abnormal region, and this too is a value between 0 and 1.

Furthermore, the first to third classifiers 41 to 43 are not limited to those that determine the presence or absence of pancreatic abnormalities based on the characteristics of the subregions. The first to third classifiers 41 to 43 may be configured to extract the characteristics of each subregion from the medical image G0 when the medical image G0 is input and determine the presence or absence of pancreatic abnormalities. Furthermore, the first to third classifiers 41 to 43 may determine that there is an abnormality in the pancreas if at least one of the subregions is absent.

On the other hand, there may be cases where the determination result by the first determination unit 24 indicates that all subregions are present. For this reason, a fourth classifier for the pancreas from which all subregions have been extracted may be prepared separately, and when the determination result by the first determination unit 24 indicates that the head, body, and tail are present, the selection unit 25 may select the fourth classifier.

The second determination unit 26 determines the presence or absence of pancreatic abnormalities using the classifier selected by the selection unit 25. To do this, the second determination unit 26 derives features of each subregion from the medical image G0 and inputs the derived features into the selected classifier to determine the presence or absence of pancreatic abnormalities.

The second determination unit 26 then compares the evaluation value output by the selected classifier with a predetermined threshold, and outputs a determination result indicating the presence of an abnormality if the evaluation value is equal to or greater than the threshold. Furthermore, if the classifiers 41-43 are used to extract abnormal regions in the pancreas, they determine as abnormal regions any region for which the evaluation value indicating an abnormal region is equal to or greater than the threshold.

The display control unit 27 displays the determination result of whether or not an abnormality is present on the display 14. Figure 8 is a diagram showing the display screen of the determination result. As shown in Figure 8, the display screen 50 displays a medical image G0 when it is determined that an abnormality is present. The displayed medical image G0 is, for example, a tomographic image including an area where an abnormality is found. In Figure 8, the abnormal area 51 determined in the medical image G0 is hatched. The determination result 52 of whether or not an abnormality is present is also displayed. In Figure 8, in addition to "abnormality present," the value of 0.9, which is the evaluation value output by the selected classifier, is also displayed.

Next, the processing performed in this embodiment will be described. Figure 9 is a flowchart showing the processing performed in this embodiment. First, the image acquisition unit 21 acquires the medical image G0 from the storage 13 (step ST1), and the first extraction unit 22 extracts a pancreas region from the medical image G0 (step ST2). Next, the second extraction unit 23 extracts multiple sub-regions from the pancreas extracted by the first extraction unit 22 (step ST3). Then, the first determination unit 24 determines the presence or absence of each of the multiple sub-regions (step ST4).

Then, the selection unit 25 selects a classifier from the multiple types of classifiers to be used to determine the presence or absence of pancreatic abnormality based on the determination result by the first determination unit 24 (step ST5). Next, the second determination unit 26 determines the presence or absence of pancreatic abnormality using the classifier selected by the selection unit 25 (step ST6). Then, the display control unit 27 displays the determination result of the presence or absence of abnormality (step ST7), and the process ends.

In this way, in this embodiment, the presence or absence of pancreatic abnormalities is determined using a classifier selected depending on the presence or absence of subregions. Therefore, even if part of the pancreas is missing in the medical image G0, the presence or absence of pancreatic abnormalities can be determined.

In the above embodiment, a head classifier, a body classifier, and a tail classifier may be provided instead of the first classifier 41, the second classifier 42, and the third classifier 43, and the selection unit 25 may select one of the head classifier, the body classifier, and the tail classifier depending on whether a sub-region is present. In this case, if the first determination unit 24 determines that a head 33 is present, a body 34 is present, and a tail 35 is absent, the selection unit 25 selects the head classifier and the body classifier. The second determination unit 26 uses the head classifier and the body classifier selected by the selection unit 25 to determine the presence or absence of an abnormality in the pancreas using the features of the head and body of the pancreas region extracted from the medical image G0.

Here, the head classifier, body classifier, and tail classifier are constructed by machine learning to determine the presence or absence of pancreatic abnormalities from features of only the head, body, and tail of the pancreas, respectively.

In addition, in the above embodiment, the sub-regions extracted from the pancreas are the head, body, and tail, but this is not limited to this. For example, as shown in Figure 10, sub-regions may be extracted by dividing the pancreas at equal intervals along the long axis 36 of the pancreas.

Furthermore, while the above embodiment describes processing when only one of the head, body, and tail of the pancreas is missing, this is not limited to this. A classifier for cases when two of the head, body, and tail of the pancreas are missing may be prepared, and when two of the head, body, and tail of the pancreas are missing, this classifier may be selected to determine whether or not there is an abnormality in the pancreas.

Furthermore, in the above embodiment, a CT image is used as the medical image G0, but this is not limited to this. Any image can be used as the medical image G0, including three-dimensional images such as MRI images, as well as radiological images obtained by simple imaging.

In addition, in the above embodiment, the hardware structure of the processing units that perform various processes, such as the image acquisition unit 21, first extraction unit 22, second extraction unit 23, first determination unit 24, selection unit 25, second determination unit 26, and display control unit 27, can be the various processors listed below. As mentioned above, the various processors include a CPU, which is a general-purpose processor that executes software (programs) and functions as various processing units, as well as dedicated electrical circuits, such as programmable logic devices (PLDs) whose circuit configuration can be changed after manufacture, such as FPGAs (Field Programmable Gate Arrays), and application-specific integrated circuits (ASICs), which are processors with a circuit configuration designed specifically to perform specific processes.

A single processing unit may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs or a combination of a CPU and an FPGA). Also, multiple processing units may be configured with a single processor.

Examples of configuring multiple processing units with a single processor include, first, a form in which one processor is configured with a combination of one or more CPUs and software, and this processor functions as multiple processing units, as is typical of client and server computers. Second, a form in which a processor is used to realize the functions of an entire system including multiple processing units on a single IC (Integrated Circuit) chip, as is typical of systems on chips (SoCs). In this way, the various processing units are configured as a hardware structure using one or more of the various processors listed above.

More specifically, the hardware structure of these various processors can be an electrical circuit that combines circuit elements such as semiconductor devices.

REFERENCE SIGNS LIST 1 Computer 2 Imaging device 3 Image storage server 4 Network 11 CPU
12 Image processing program 13 Storage 14 Display 15 Input device 16 Memory 17 Network I/F
18 Bus 20 Image processing device 21 Image acquisition unit 22 First extraction unit 23 Second extraction unit 24 First determination unit 25 Selection unit 26 Second determination unit 27 Display control unit 30 Pancreas 31 Vein 32 Artery 33 Head 34 Body 35 Tail 36 Long axis 41 First classifier 42 Second classifier 43 Third classifier 50 Display screen 51 Abnormal region 52 Determination result D0 Tomographic image G0 Medical image

Claims (10)

at least one processor;
The processor:
Extract the pancreas region from the medical image,
Extracting a plurality of sub-regions from the extracted pancreas;
determining the presence or absence of each of the plurality of sub-regions;
selecting a classifier to be used for determining the presence or absence of an abnormality in the pancreas from a plurality of types of classifiers based on the result of the determination;
An image processing device that determines the presence or absence of an abnormality in the pancreas using the selected classifier. The image processing apparatus according to claim 1 , wherein the plurality of types of classifiers include a classifier that determines the presence or absence of the abnormality depending on a combination of the presence or absence of the plurality of sub-regions. The image processing device described in claim 2, wherein the classifier is trained using images consisting of combinations of the presence and absence of the multiple subregions as training data. The image processing apparatus according to claim 1 , wherein the plurality of types of classifiers include a classifier that determines the presence or absence of the abnormality based on the sub -regions determined to have an abnormality. The image processing apparatus according to claim 2 , wherein the plurality of types of classifiers determine the presence or absence of an abnormality based on features of the sub-regions determined to have an abnormality. The image processing device according to claim 5 , wherein the feature of the sub-region determined to be present is at least one of the diameter, size, and texture of the sub-region. The image processing device according to claim 1 , wherein the plurality of types of classifiers include a classifier that determines that there is an abnormality in the pancreas when at least one of the plurality of sub-regions is absent. The image processing device of claim 1, wherein the multiple sub-regions are the head, body, and tail of the pancreas. A computer extracts a region of the pancreas from the medical image,
Extracting a plurality of sub-regions from the extracted pancreas;
determining the presence or absence of each of the plurality of sub-regions;
selecting a classifier to be used for determining the presence or absence of an abnormality in the pancreas from a plurality of types of classifiers based on the result of the determination;
An image processing method for determining the presence or absence of an abnormality in the pancreas using the selected classifier. A procedure for extracting a pancreas region from a medical image;
extracting a plurality of sub-regions from the extracted pancreas;
determining the presence or absence of each of the plurality of sub-regions;
a step of selecting a classifier to be used for determining the presence or absence of an abnormality in the pancreas from a plurality of types of classifiers based on the result of the determination;
and a procedure for determining the presence or absence of an abnormality in the pancreas using the selected classifier.