Diagnosis support apparatus, diagnosis support program and learned model construction method

Abstract

To reduce troublesomeness in diagnosis of a medical doctor due to the wrong detection of plaque.SOLUTION: A diagnosis support apparatus (3) includes: a classifying part (334) which clusters an object region that is encircled by markers in a medical image projecting the blood vessel attached with markers (B) encircling a region to be a plaque in the blood vessel and classifies the object region into a blood vessel region where the blood vessel is projected and a plaque region where a plaque is projected; a calculation part (335) calculating a ratio of width of the plaque region relative to a width of the blood vessel region; a determination part (336) determining whether or not the ratio is prescribed determination threshold or greater; and an output processing part (338) performing output processing based on a determination result by the determination part.SELECTED DRAWING: Figure 5

Description

[Technical field]

[0001] The present invention relates to a diagnostic support device, a diagnostic support program, and a method for constructing a trained model. [Background technology]

[0002] Conventionally, various techniques for classifying plaque in a blood vessel have been proposed. For example, Patent Document 1 describes a method including a step of providing a characterized tissue component map from a spectral analysis of ultrasound echo information related to a blood vessel cross section by a tissue characterization application, in which different plaque components are discriminated in the characterized tissue map by assigned discrimination values, a step of applying a classification criterion to the spatially arranged data of the characterized tissue map by a plaque classification application, and a step of providing a plaque classification related to the blood vessel cross section according to the applying step.

[0003] Also, a trained model is used for classifying plaque. For example, Patent Document 2 describes a medical image processing device including an image acquisition unit that acquires a medical image, a classifier that classifies the medical image or a region of interest contained in the medical image into one of two or more classes based on a feature amount obtained from the medical image, a reliability calculation unit that calculates the reliability of the classification result of the medical image or the region of interest by the classifier, and a determination unit that determines the classification result of the medical image or the region of interest based on the reliability. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Special Publication No. 2009-530068 [Patent Document 2] International Publication No. 2020 / 170791 Summary of the Invention [Problem to be solved by the invention]

[0005] Conventional devices and methods have sometimes misjudged low-risk plaques, which have a relatively low risk of developing angina pectoris, myocardial infarction, etc., as high-risk plaques, which have a high risk of developing. For this reason, medical images may have more markers indicating the presence of high-risk plaques than the number of high-risk plaques actually present. If a doctor were to use such medical images for diagnosis as is, the doctor would have to perform more diagnoses than necessary, including determining whether the markers are appropriate. [Means for solving the problem]

[0006] A diagnostic support device according to one aspect of the present invention includes a classification unit that performs clustering on a target area surrounded by a marker in a medical image showing a blood vessel, the target area being surrounded by the marker surrounding an area believed to be plaque within the blood vessel, to classify the target area into a vascular area in which a blood vessel is shown and a plaque area in which plaque is shown, a calculation unit that calculates the ratio of the width of the plaque area to the width of the vascular area, a judgment unit that judges whether the ratio is equal to or greater than a predetermined judgment threshold, and an output processing unit that performs output processing based on the judgment result by the judgment unit.

[0007] In addition, a diagnostic assistance program according to another aspect of the present invention causes a processor to execute a classification step of applying clustering to a target area surrounded by a marker in a medical image showing a blood vessel, the target area being surrounded by the marker surrounding an area believed to be plaque within the blood vessel, to classify the target area into a vascular area in which a blood vessel is shown and a plaque area in which plaque is shown, a calculation step of calculating the ratio of the width of the plaque area to the width of the vascular area, a judgment step of judging whether the ratio is equal to or greater than a predetermined judgment threshold, and an output processing step of performing output processing based on the judgment result in the judgment step.

[0008] In addition, a method for constructing a trained model according to another aspect of the invention includes the steps of: binarizing a target area surrounded by a mark in a plurality of medical images showing blood vessels and plaque within the blood vessels, the target area being surrounded by a mark, using a predetermined classification threshold; if a white area is present in the binarized target area, labeling the plaque in the target area as calcified plaque; and constructing a trained model by machine learning using a pair of the medical images and the label attached to the target area in the medical images as training data. [Brief description of the drawings]

[0009] [Figure 1] 1 is a block diagram showing a schematic configuration of a diagnosis support system according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a block diagram showing the functional configuration of a diagnosis support device provided in the system. [Diagram 3] FIG. 2 is a diagram illustrating an example of a process performed by the device. [Figure 4] FIG. 2 is a diagram showing an example of classification of target regions by the device. [Diagram 5] FIG. 2 is a diagram illustrating an example of a process performed by the device. [Figure 6] FIG. 2 is a diagram illustrating an example of a process performed by the device. [Figure 7] FIG. 13 is a block diagram showing a schematic configuration of a diagnosis support system according to a modified example of the embodiment. [Figure 8] 13 is a flowchart showing the flow of a method for constructing a trained model according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] <Diagnosis support system> First, an embodiment of a diagnosis support system 100 (hereinafter, referred to as system 100) according to an aspect of the present invention will be described in detail.

[0011] As shown in FIG. 1, the system 100 includes a modality 1, a server 2, a diagnosis support device 3, and a terminal device 4.

[0012] [Modality] The modality 1 captures an image of a diagnostic target part of a subject, and generates medical image data showing the diagnostic target part. The medical image data generated by the modality 1 according to this embodiment is in a format conforming to the DICOM (Digital Imaging and Communications in Medicine) standard. The modality 1 according to this embodiment is a CT (Computed Tomography) device. The modality 1 transmits the generated medical image data to the server 2. Note that the modality 1 may be an MRI (Magnetic Resonance Imaging) device, a general X-ray device, or the like.

[0013] [server] The server 2 stores data of medical images acquired from the modality 1 or the diagnosis support device 3 in association with information on the subject. The server 2 constitutes a PACS (Picture Archiving and Communications System).

[0014] [Terminal Device] The terminal device 4 acquires medical image data according to the doctor's operation from the server 2 or the diagnosis support device 3. Then, the terminal device 4 displays the medical image based on the acquired data. The terminal device 4 is configured to be a PC (Personal Computer), a smartphone, a tablet terminal, or a monitor. The terminal device 4 is also configured to be able to accept input of a diagnosis result for the displayed medical image. When the terminal device 4 accepts the input of the diagnosis result, it transmits it to the server 2. The server 2 that receives the diagnosis result stores the received diagnosis result in association with the medical image data.

[0015] [Diagnostic support device] The diagnosis support device 3 performs processing for supporting diagnosis on data of medical images acquired from the modality 1 or the server 2. The diagnosis support device 3 is configured with a PC (Personal Computer) or a dedicated device. As shown in FIG. 2, the diagnosis support device 3 according to this embodiment includes a communication unit 31, a storage unit 32, and a control unit 33.

[0016] [Communications Department] The communication unit 31 communicates with the server 2. The communication unit 31 according to this embodiment is configured with a communication module. The communication unit 31 may be configured to communicate with at least one of the modality 1 and the terminal device 4 via the server 2. The communication unit 31 may also be configured to communicate directly with at least one of the modality 1 and the terminal device 4 without going through the server 2.

[0017] [Storage section] The storage unit 32 stores a diagnostic assistance program 321. The storage unit 32 according to this embodiment also stores a trained model 322. Moreover, the storage unit 32 according to this embodiment is configured with a semiconductor memory, a hard disk, and the like. Note that the storage unit 32 may be divided into one that stores the diagnostic assistance program 321 and one that stores the trained model 322. Furthermore, the trained model 322 may be stored in another device with which the diagnostic assistance device 3 can communicate.

[0018] The diagnostic support program 321 describes the operation of the control unit 33 of the diagnostic support device.

[0019] The trained model 322 is constructed by machine learning using as training data a set of a past medical image showing a blood vessel and a mark B, which is attached by a doctor and surrounds an area showing intravascular plaque in the past medical image. The trained model 322 according to the present embodiment is constructed by machine learning using as training data a set of a past medical image showing a coronary artery and a mark B surrounding an area showing intravascular calcified plaque. That is, the trained model 322 according to the present embodiment is for extracting calcified plaque from a medical image showing a coronary artery. The trained model 322 is not limited to the trained model constructed as described above. That is, the trained model 322 may be constructed using as training data a set of a medical image showing a blood vessel other than the coronary artery in which plaque may occur and the mark B. The trained model 322 may also be constructed using as training data a set of a medical image and a mark B surrounding an area showing intravascular lipid plaque.

[0020] The teacher data used to construct the trained model 322 according to this embodiment is a set of a past medical image, a landmark B, and plaque group data. The plaque group data is data indicating which of a plurality of groups set according to the size of the plaque, a plaque shown in an area surrounded by the landmark B, is classified into. In this embodiment, the plaque shown in the past medical image included in the teacher data (surrounded by the landmark B) is classified into one of the following three groups (1) to (3). Of these, the plaque classified into the group (2) is a high-risk plaque having a high risk of developing angina pectoris, myocardial infarction, etc. The plaque classified into the group (3) is a low-risk plaque having a low risk of development. The plaque classified into the group (1) is a medium-risk plaque having a risk of development lower than (2) and higher than (3). (1) A group containing plaques whose width perpendicular to the blood vessel is less than a certain value. (2) This group contains plaques whose width perpendicular to the blood vessel is so large that they cause stenosis of the blood vessel. (3) This group includes plaques whose width perpendicular to the blood vessel is larger than that of (1), but not large enough to cause stenosis in the blood vessel.

[0021] When the target to be extracted is lipid plaque, plaques classified into group (2) are high-risk plaques, and plaques classified into groups (1) and (3) are medium-risk plaques. Note that plaques shown in medical images may be classified into either of two groups, a high-risk group and a low-risk group, or into four or more groups with different risks of developing.

[0022] [Control Unit] The control unit 33 includes an acquisition unit 331, a plaque extraction unit 332, a binarization unit 333, a classification unit 334, a calculation unit 335, a judgment unit 336, a second judgment unit 337, and an output processing unit 338. The control unit 33 according to this embodiment is configured with a processor. When this processor executes the diagnostic assistance program stored in the storage unit, the diagnostic assistance program 321 causes the control unit 33 to function as each of the control blocks 331 to 338.

[0023] (Acquisition Department) The acquisition unit 331 executes an acquisition step. In the acquisition step, the acquisition unit 331 acquires data of a medical image before extraction, which is a medical image showing a blood vessel and does not have the mark B. The acquisition unit 331 according to the present embodiment acquires data of the medical image via the communication unit 31.

[0024] (Plaque Extraction Section) The plaque extraction unit 332 executes a plaque extraction step. In the plaque extraction step, the plaque extraction unit 332 inputs data of a medical image before extraction to the trained model 322, and attaches a mark B surrounding a region that is considered to be a plaque in a blood vessel to the medical image before extraction. As described above, the trained model 322 according to this embodiment is constructed by machine learning using a set of a past medical image in which a coronary artery is captured and a mark B surrounding a region in which a calcified plaque in a blood vessel is captured as training data. That is, the plaque to be extracted by the plaque extraction unit 332 according to this embodiment is a calcified plaque. The mark B according to this embodiment is a rectangular frame (bounding box) whose four sides are parallel to the four sides of the medical image, as shown on the left side of FIG. 3. Note that the mark B is not limited to a rectangle as long as it has a shape that can surround a region in which a calcified plaque is captured. That is, the mark B may be, for example, a circle, a triangle, or the like.

[0025] The plaque extraction unit 332 according to this embodiment outputs information based on plaque group data corresponding to the plaque when attaching a mark B surrounding an area considered to be plaque in a blood vessel. The "information based on plaque group data" may be, for example, direct information on which group the plaque shown in the past medical image belongs to (wording such as "group (1)", "meets the requirements to be classified into group (2)"), or may be information on the level of risk of onset (wording such as "high-risk plaque"). In addition, the "information based on plaque group data" may include the reason why the plaque is classified into that group (wording such as "because the size of the plaque is ...").

[0026] Depending on the degree of learning of the trained model 322, the plaque extraction unit 332 may erroneously determine that a non-plaque, a low-risk plaque, or the like, is a calcified plaque and mark it with the mark B. For this reason, a medical image based on data immediately after the plaque extraction unit 332 marks the mark B may have more marks B marked than the number of calcified plaques that actually exist.

[0027] (Binarization section) The binarization unit 333 executes a binarization step. In the binarization step, the binarization unit 333 binarizes the target region with a predetermined classification threshold, as shown in the center and right of FIG. 3. The target region is an area surrounded by marks B. When multiple marks B are attached to one medical image, the binarization unit 333 sequentially extracts the target region surrounded by each mark B, and binarizes each target region. Note that the control unit 33 does not need to include the binarization unit 333.

[0028] (Second Judgment Department) The second determination unit 337 executes a second determination step. In the second determination step, the second determination unit 337 determines whether or not a white area exists in the binarized target area. For example, the binarized target area shown on the right side of FIG. 3 is determined to have a white area. Note that if the control unit 33 does not include the binarization unit 333, the control unit 33 does not need to include the second determination unit 337.

[0029] (Classification Department) The classification unit 334 executes a classification step. In the classification step, the classification unit 334 performs clustering on a target region in a medical image, and classifies the target region into a blood vessel region in which a blood vessel is captured, and a plaque region in which a plaque is captured. The classification unit 334 according to the present embodiment classifies a target region in which the second judgment unit 337 judges that a white region is present, into a blood vessel region and a plaque region. This makes it possible to narrow down in advance the number of target regions for which the presence or absence of stenosis is judged, thereby reducing the processing load.

[0030] The classification unit 334 according to this embodiment classifies the target region into a blood vessel region in which a blood vessel is captured and a calcified plaque region in which a calcified plaque is captured. The classification unit 334 according to this embodiment classifies the target region into a blood vessel region and a plaque region using the K-means method. In particular, the classification unit 334 according to this embodiment classifies the target region into a plaque region, an external region, and a blood vessel region as shown in FIG. 4 using the K-means method with K=4. The plaque region is the region showing the highest brightness. The external region is the region showing the lowest brightness in which external tissue is captured. The blood vessel region is the region showing the second and third highest brightness from the top. If the value of K is 3 or less, the plaque region will be crushed, and if the value of K is 5 or more, it will be difficult to draw a line between the regions, but by setting K=4, the plaque region can be extracted in a natural shape. The classification unit 334 may be configured to classify the target region using a clustering method other than the K-means method.

[0031] (Calculation section) The calculation unit 335 executes a calculation step. In the calculation step, the calculation unit 335 calculates the ratio of the width of the plaque region to the width of the blood vessel region. The width of the blood vessel may not be uniform. Similarly, the width of the plaque in the blood vessel may not be uniform. For this reason, the calculation unit 335 according to the present embodiment scans the blood vessel region and the plaque region along the extension direction of the blood vessel. Specifically, as shown in FIG. 5, the calculation unit 335 calculates the coordinates of three intersection points P1, P2, and P3 of a straight line L drawn on the target region so as to be perpendicular to the extension direction of the blood vessel region (the up-down direction in FIG. 5). The intersection point P1 is an intersection point between the straight line L and a line indicating the inner wall on the side where plaque does not occur in the blood vessel region. The intersection point P2 is an intersection point between the straight line L and a line indicating the inner wall on the side where plaque occurs in the blood vessel region. The intersection point P3 is an intersection point between the straight line L and an end of the plaque region on the side that is not in contact with the inner wall of the blood vessel. Then, the calculation unit 335 calculates the ratio of the distance between the intersection points P3 and P2 to the distance between the intersection points P1 and P2 (the sum of the distance between the intersection points P1 and P3 and the distance between the intersection points P3 and P2). The above operation is repeated by moving the position of the straight line L a predetermined distance in the extension direction of the blood vessel, thereby calculating the ratio for each part of the blood vessel region. Note that when the plaque region is branched, the calculation unit 335 calculates the ratio by taking the sum of the width of the main part of the plaque region and the width of the branched part as the width of the plaque region.

[0032] The calculation unit 335 according to this embodiment checks the extension direction of the vascular region before scanning the vascular region and the plaque region. If the extension direction of the vascular region is not parallel to any of the frame lines of the target region as shown in the left side of FIG. 6, the calculation unit 335 rotates the target region so that the extension direction of the vascular region is parallel to any of the frame lines of the target region as shown in the right side of FIG. 6. The calculation unit 335 then scans the vascular region and the plaque region along the extension direction of the frame lines. This makes it easy to scan the vascular region and the plaque region.

[0033] (Judgment Department) The determination unit 336 executes a determination step. In the determination step, the determination unit 336 determines whether or not the ratio calculated by the calculation unit 335 is equal to or greater than a predetermined determination threshold (e.g., 50%). If the ratio is equal to or greater than (less than) the determination threshold, the determination unit 336 determines that stenosis occurs (does not occur) in a portion of the blood vessel corresponding to the target region. As described above, the calculation unit 335 calculates the ratio at each of a plurality of locations in the blood vessel region. Therefore, the determination unit 336 according to this embodiment determines whether or not the highest ratio among the plurality of ratios obtained by scanning is equal to or greater than the determination threshold. This makes it possible to accurately determine whether or not stenosis occurs even when the width of the blood vessel or plaque is not uniform.

[0034] (Output processing section) The output processing unit 338 executes an output processing step. In the output processing step, the output processing unit 338 performs output processing based on the judgment result by the judgment unit 336. First, when the judgment unit 336 judges that the ratio is not equal to or greater than the judgment threshold, the output processing unit 338 according to this embodiment deletes the mark B that surrounded the corresponding target region from the medical image. This makes the medical image easy to view, with the mark B only attached to the region that truly requires diagnosis. Note that the output processing unit 338 may be configured to display the judgment result of whether or not the mark B is equal to or greater than the judgment threshold, for each mark B, instead of deleting the mark B from the medical image. Next, the output processing unit 338 according to this embodiment controls the communication unit 31 so that the communication unit 31 transmits data of the medical image from which the unnecessary mark B has been deleted to the server 2 or the terminal device 4.

[0035] [Diagnosis support system variant 1] As shown in FIG. 7, the system 100 may include, instead of the diagnosis support device 3, an extraction device 3A and a diagnosis support device 3B.

[0036] In this case, the extraction device 3A includes a second storage unit (not shown) and a second control unit (not shown) in addition to the communication unit 31 similar to that of the diagnosis support device 3. The second control unit includes an acquisition unit 331 and a plaque extraction unit 332 similar to those of the diagnosis support device 3, as well as a second output processing unit (not shown). The second output processing unit controls the communication unit 31 to output data of a medical image showing a blood vessel, the medical image having a marker B attached thereto surrounding an area considered to be plaque in the blood vessel, to the diagnosis support device 3B.

[0037] Further, the diagnosis support device 3B includes a third storage unit (not shown) and a third control unit (not shown) in addition to the communication unit 31 similar to that of the diagnosis support device 3. The third control unit includes a second acquisition unit (not shown) in addition to the binarization unit 333, classification unit 334, calculation unit 335, judgment unit 336, second judgment unit 337, and output processing unit 338 similar to those of the diagnosis support device 3. The second acquisition unit acquires data of a medical image showing a blood vessel, which is a medical image with a mark B surrounding a region considered to be a plaque in the blood vessel, from the extraction device 3A via the communication unit of the diagnosis support device 3B. The second acquisition unit acquires data of a medical image with a mark B surrounding a region considered to be a calcified plaque in the blood vessel. The second acquisition unit may be configured to acquire data of a medical image with a mark B added by a person (such as an engineer).

[0038] [Modification 2 of the diagnostic support system] In addition, the system 100 may be configured such that the diagnosis support device 3 is integrated with the terminal device 4. In addition, the system 100 may be configured such that the diagnosis support device 3 is integrated with the server 2.

[0039] [Diagnosis support system variant 3] The system 100 may also include a model construction device (not shown). The model construction device constructs a trained model 322 used by the plaque extraction unit 332 of the diagnosis support device 3. This model construction device includes a binarization unit, a judgment unit, and a learning unit. The binarization unit binarizes a target area of ​​a medical image with a predetermined classification threshold. The judgment unit judges whether or not a white area exists in the binarized target area. The learning unit constructs a trained model by machine learning using pairs of a medical image and a landmark, in which a label indicating that the target area is a calcified plaque, as training data. The labeled medical image used for learning may be processed by the diagnosis support device 3 to support diagnosis.

[0040] [Diagnosis Support System Variation 4] In addition, some or all of the functions of each control block of at least one of the devices (e.g., diagnosis support device 3) constituting system 100 can be realized by a logic circuit. For example, an integrated circuit in which a logic circuit that functions as each of the above control blocks is formed is also included in the scope of the present invention. In addition, the functions of each of the above control blocks can be realized by, for example, a quantum computer.

[0041] [Effects of diagnostic support devices] The diagnosis support device 3, 3A described above judges whether the mark B attached to the medical image is based on a false detection or surrounds an area that truly requires diagnosis, and performs output based on the judgment result (deletion of the mark B based on the false detection, direct display of the judgment result, etc.). Therefore, the medical image output by the diagnosis support device 3, 3A is one in which the mark B based on the false detection can be distinguished from the mark B surrounding the area that truly requires diagnosis, or the mark B is attached only to the area that truly requires diagnosis. As a result, the doctor can narrow down the target and make a diagnosis. That is, the diagnosis support device 3, 3A can reduce the doctor's effort in diagnosis due to the false detection of plaque. In particular, the diagnosis support device 3 includes a plaque extraction unit 332. Therefore, the diagnosis support device 3 can obtain a medical image in which the mark B is attached only to the area that truly requires diagnosis, simply by inputting the data of the medical image before plaque extraction.

[0042] <How to build a trained model> Next, an embodiment of a trained model construction method according to another invention will be described.

[0043] [Flow of how to build a trained model] The method for constructing a trained model is to construct a trained model for extracting plaque in blood vessels from medical images showing blood vessels. As shown in Fig. 8, the method for constructing a trained model includes a binarization step S1, an assignment step S2, and a learning step S3.

[0044] (Binarization step) In the first binarization step S1, in a plurality of medical images showing blood vessels and plaques in the blood vessels, with the mark B surrounding the plaque, a target region surrounded by the mark B is binarized using a predetermined classification threshold. For the binarization, first, the medical image data needs to be converted from the DICOM standard to, for example, a png format. In this case, when the window level is set to, for example, 120 and the window width is set to, for example, 1200, it is preferable that the classification threshold (png pixel value) is set within a range of 200 to 230. The target region is binarized using, for example, a model construction device or the like described in the above "Modification 3 of the diagnosis support system."

[0045] (Granting step) After the target region is binarized, the process proceeds to the labeling step S2. In the labeling step S2, if a white region exists in the binarized target region, the plaque in the target region is labeled as a calcified plaque. The labeling of the target region is performed, for example, by using the above-mentioned model construction device. On the other hand, in the construction of a trained model for extracting lipidic plaque, in the labeling step, if a white region does not exist in the binarized target region, the plaque in the target region is labeled as a lipidic plaque.

[0046] (Learning Steps) After the labels are attached, the process proceeds to a learning step S3. In the learning step S3, a trained model 322 is constructed by machine learning using a set of a medical image and a label attached to a target region in the medical image as training data. The trained model 322 is constructed, for example, using the above-mentioned model construction device or the like. In constructing a trained model for extracting calcified plaque, a trained model is constructed using a set of a medical image and a label indicating that it is calcified plaque as training data. On the other hand, in constructing a trained model for extracting lipidic plaque, a trained model is constructed using a set of a medical image and a label indicating that it is lipidic plaque as training data.

[0047] [Effects of the trained model construction method] The method for constructing a trained model described above classifies plaques in medical images into either calcified plaques or lipid plaques by binarization. Therefore, this method for constructing a trained model makes it easy to construct a trained model that extracts plaques from medical images that show blood vessels.

[0048] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention.

[0049] 〔summary〕 A diagnostic support device according to a first aspect of the present invention includes a classification unit that performs clustering on a target area surrounded by a marker in a medical image showing a blood vessel, the target area being surrounded by the marker surrounding an area believed to be plaque within the blood vessel, to classify the target area into a vascular area in which a blood vessel is shown and a plaque area in which plaque is shown; a calculation unit that calculates the ratio of the width of the plaque area to the width of the vascular area; a judgment unit that judges whether the ratio is equal to or greater than a predetermined judgment threshold; and an output processing unit that performs output processing based on the judgment result by the judgment unit.

[0050] A diagnostic support device according to aspect 2 of the present invention may be configured in the above-mentioned aspect 1 such that, when the judgment unit judges that the ratio is not greater than or equal to the judgment threshold, the output processing unit deletes the landmark that surrounded the corresponding target region from the medical image.

[0051] A diagnosis support device according to aspect 3 of the present invention may be configured in the above-mentioned aspect 1 or 2 such that the classification unit uses a K-means algorithm to classify the target region into the vascular region and the plaque region.

[0052] A diagnostic support device according to aspect 4 of the present invention may be configured in the above-mentioned aspect 3 such that the classification unit uses a K-means algorithm with K=4 to classify the target region into the plaque region showing the highest brightness, an external region showing the lowest brightness and including external tissue, and the plaque region showing the second and third highest brightness from the top.

[0053] A diagnosis support device according to aspect 5 of the present invention may be configured in any one of aspects 1 to 4 above, such that the calculation unit scans the vascular region and the plaque region along the extension direction of the blood vessel, and the judgment unit judges whether the highest ratio is equal to or greater than the judgment threshold value.

[0054] A diagnosis support device according to a sixth aspect of the present invention may be configured in the above-mentioned fifth aspect such that the landmark is a rectangular frame whose four sides are parallel to the four sides of the medical image, and the calculation unit rotates the target region so that the extension direction of the vascular region is parallel to one of the frame lines of the target region, and scans the vascular region and the plaque region along the extension direction of the frame lines.

[0055] A diagnosis support device according to aspect 7 of the present invention may be configured in any one of aspects 1 to 6 above, further comprising an acquisition unit that acquires data of a pre-extraction medical image that shows a blood vessel and that does not have the marker attached, and a plaque extraction unit that adds a marker surrounding an area that is considered to be intravascular plaque to the pre-extraction medical image by inputting the data of the pre-extraction medical image into a trained model constructed by machine learning using as training data a set of a past medical image showing a blood vessel and a marker attached by a doctor to surround an area showing intravascular plaque in the past medical image.

[0056] A diagnostic support device according to aspect 8 of the present invention, in accordance with aspect 7 above, may be configured such that the teacher data is a set of the past medical image, the landmark, and plaque group data indicating which of a plurality of groups set according to the size of the plaque a plaque depicted in the area surrounded by the landmark is classified into, and the plaque extraction unit, when adding a landmark surrounding an area considered to be plaque in the blood vessel, outputs information based on the plaque group data corresponding to the plaque.

[0057] A diagnostic support device according to aspect 9 of the present invention, in any of aspects 1 to 8 above, may include a binarization unit that binarizes the target region using a predetermined classification threshold, and a second judgment unit that determines whether or not a white region is present in the binarized target region, and the classification unit may be configured to classify the target region in which the second judgment unit has determined that a white region is present, into the vascular region and the plaque region.

[0058] A diagnostic support device according to aspect 10 of the present invention may be configured in any one of aspects 1 to 9 above, wherein the plaque is calcified plaque, and the classification unit classifies the target region into a vascular region in which a blood vessel is depicted, and a calcified plaque region in which a calcified plaque is depicted.

[0059] A diagnostic assistance program according to aspect 11 of the present invention is configured to cause a processor to execute a classification step of applying clustering to a target area surrounded by a marker in a medical image showing a blood vessel, the target area being surrounded by the marker surrounding an area believed to be plaque within the blood vessel, to classify the target area into a vascular area in which a blood vessel is shown and a plaque area in which plaque is shown; a calculation step of calculating the ratio of the width of the plaque area to the width of the vascular area; a judgment step of judging whether the ratio is greater than or equal to a predetermined judgment threshold; and an output processing step of performing output processing based on the judgment result in the judgment step.

[0060] A method for constructing a trained model according to aspect 12 of the present invention is a method comprising: a binarization step of binarizing, using a predetermined classification threshold, a target region surrounded by a mark in a plurality of medical images showing blood vessels and plaque within the blood vessels, the target region being surrounded by a mark surrounding the plaque; an assignment step of assigning a label indicating that the plaque within the target region is a calcified plaque if a white region is present in the binarized target region; and a learning step of constructing a trained model by machine learning using as training data a pair of the medical images and the label assigned to the target region in the medical images. [Explanation of symbols]

[0061] 100 Diagnostic Support System 1. Modality 2 Server 3, 3B Diagnostic support device 3A extraction device 31 Communications Department 32 Storage section 321 Diagnostic Support Program 322 trained models 33 Control Unit 331 Acquisition Department 332 Plaque Extraction Section 333 Binarization section 334 Classification Department 335 Calculation Department 336 Judgment Department 337 Second Judgment Department 338 Output Processing Section 4 Terminal Equipment

Claims

1. a classification unit that performs clustering on a target region surrounded by a mark in a medical image showing a blood vessel, the target region being surrounded by a mark surrounding a region considered to be plaque in the blood vessel, and classifies the target region into a blood vessel region showing the blood vessel and a plaque region showing the plaque; A calculation unit that calculates a ratio of a width of the plaque region to a width of the blood vessel region; A determination unit that determines whether the ratio is equal to or greater than a predetermined determination threshold value; an output processing unit that performs output processing based on a result of the determination by the determination unit; A diagnosis support device comprising:

2. When the determination unit determines that the ratio is not equal to or greater than the determination threshold, the output processing unit deletes the mark that surrounded the corresponding target region from the medical image.

2. The diagnosis support device according to claim 1.

3. The classification unit classifies the target region into the blood vessel region and the plaque region using a K-means algorithm.

2. The diagnosis support device according to claim 1.

4. The classification unit classifies the target region using a K-means method with K=4. the plaque region exhibiting the highest brightness; The outer region, which shows the lowest brightness of the outer tissue, the blood vessel region showing the second and third highest brightness from the top; Classify into 4. The diagnosis support device according to claim 3.

5. The calculation unit is Scanning the blood vessel region and the plaque region along an elongation direction of the blood vessel; The determination unit determines whether the highest ratio is equal to or greater than the determination threshold value.

2. The diagnosis support device according to claim 1.

6. the mark is a rectangular frame whose four sides are parallel to the four sides of the medical image, The calculation unit is Rotating the target region so that the extension direction of the blood vessel region is parallel to any of the border lines of the target region; scanning the blood vessel region and the plaque region along an extension direction of the frame line; 6. The diagnosis support device according to claim 5.

7. an acquisition unit that acquires data of a medical image showing a blood vessel, the medical image having no markers and before extraction; a plaque extraction unit that adds a mark surrounding an area considered to be intravascular plaque to the medical image before extraction by inputting data of the medical image before extraction into a trained model constructed by machine learning using a set of a past medical image showing a blood vessel and a mark surrounding an area showing intravascular plaque in the past medical image, the mark having been added by a doctor, as training data; Further comprising:

2. The diagnosis support device according to claim 1.

8. the teacher data is a set of the past medical image, the landmark, and plaque group data indicating to which group a plaque shown in an area surrounded by the landmark is classified among a plurality of groups set according to the size of the plaque, the plaque extraction unit outputs information based on the plaque group data corresponding to the plaque when attaching a mark surrounding a region that is determined to be a plaque in a blood vessel.

8. The diagnosis support device according to claim 7.

9. a binarization unit that binarizes the target region using a predetermined classification threshold; a second determination unit that determines whether or not a white area exists in the binarized target area; Equipped with The classification unit classifies the target region in which the second judgment unit judges that a white region is present into the blood vessel region and the plaque region.

2. The diagnosis support device according to claim 1.

10. the plaque is a calcified plaque; The classification unit classifies the target region into a blood vessel region in which a blood vessel is captured and a calcified plaque region in which a calcified plaque is captured.

10. The diagnosis support device according to claim 1,

11. The processor: a classification step of performing clustering on a target region surrounded by a mark in a medical image showing a blood vessel, the target region being surrounded by a mark surrounding a region considered to be plaque in the blood vessel, and classifying the target region into a blood vessel region showing a blood vessel and a plaque region showing a plaque; A calculation step of calculating a ratio of a width of the plaque region to a width of the blood vessel region; a determination step of determining whether the ratio is equal to or greater than a predetermined determination threshold; an output processing step of performing output processing based on a result of the determination in the determination step; A diagnostic assistance program characterized by executing the above.

12. a binarization step of binarizing a target region surrounded by a mark in a plurality of medical images showing a blood vessel and plaque in the blood vessel, the target region being surrounded by a mark, using a predetermined classification threshold; an assignment step of labeling plaque in the target region as calcified plaque when a white region is present in the binarized target region; A learning step of constructing a trained model by machine learning using a pair of the medical image and a label attached to the target region in the medical image as training data; A method for constructing a trained model, comprising:

Images