WO2020218460A1

WO2020218460A1 - Diagnostic assistance device, speculation device, diagnostic assistance system, diagnostic assistance method, diagnostic assistance program, and learned model

Info

Publication number: WO2020218460A1
Application number: PCT/JP2020/017585
Authority: WO
Inventors: 宗司廣田; 純一伊東; 三浦　雄治; 青島　健; 悦子今林
Original assignee: エーザイ・アール・アンド・ディー・マネジメント株式会社
Priority date: 2019-04-26
Filing date: 2020-04-24
Publication date: 2020-10-29
Also published as: JPWO2020218460A1

Abstract

This diagnostic assistance device 10 is provided with: a first acquisition unit 11 for acquiring an image of the brain of a subject; a second acquisition unit 12 for acquiring an speculation result obtained by making, on the basis of an image, speculation regarding dementia of the subject using learned models 21b, 21c; and a display unit 10f for displaying the speculation result and information indicating the area of interest in the brain in the image used as the basis for the speculation result.

Description

Diagnostic support device, guessing device, diagnostic support system, diagnostic support method, diagnostic support program and trained model

The present invention relates to a diagnostic support device, a guessing device, a diagnostic support system, a diagnostic support method, a diagnostic support program, and a trained model.

Conventionally, research has been conducted to estimate the risk of dementia based on MRI (Magnetic Resonance Imaging) images of the brain using machine learning technology.

For example, Patent Document 1 below describes a diagnostic support device that predicts whether or not a subject with mild cognitive impairment will develop Alzheimer's disease within a predetermined period of time. The diagnostic support device divides the brain image acquired from the subject into white matter, white matter, and spinal fluid portions, sets a plurality of regions of interest in each of the divided regions, and determines the volume of each region of interest in each region of interest. Calculate the t-value and p-value, calculate the z-value of each region of interest based on the t-value and p-value, and based on the z-value, according to the machine-learned prediction algorithm, the subject has Alzheimer's disease within a predetermined period. Predict whether or not to develop.

Patent Document 2 below describes a medical image processing apparatus for accurately acquiring the amount of change between brain images of the same patient. The medical image processing device aligns the first brain image of the subject with reference to the standard brain image, divides the first brain image into a plurality of regions, and positions the second brain image of the subject with reference to the first brain image. At the same time, the amount of change for at least one region is acquired. As a method of division, based on Broadman's brain map, the brain is divided into areas that control each function such as movement, language, perception, memory, vision and hearing, and the brain is divided into the cerebrum, diencephalon, and midbrain. A method of dividing into six types of regions, the metencephalon, the cerebellum, and the medulla oblongata, is described. Further, the same document describes that a discriminator is created by machine learning the amount of change for each region, that is, the atrophy rate as teacher data for a plurality of patients in the past.

Non-Patent Document 1 below shows the distribution of SUVR values by the VBM (Voxel Based morphometry) method in which a brain image is divided into voxels and evaluated by PET examination. Figure 2 of the same document shows the orbitofrontal cortex, central prefrontal cortex, middle temporal cortex, temporal-occipital junction, posterior cingulate cortex, angular gyrus, precuneus, putamen, and nucleus accumbens by logistic regression analysis. A map of the disease contribution rate of the SUVR value of each region is shown.

Non-Patent Document 2 below describes that a differential diagnosis of dementia is performed using physical findings, neurological examination, blood test, imaging test, and the like.

Japanese Patent No. 6438890 International release WO2019 / 003749

The result of estimating the risk of dementia based on MRI images etc. using machine learning technology is utilized for diagnosis by doctors.

However, it is not clear why the guess was made simply by presenting the guess result. Therefore, it is difficult for the doctor to judge the validity of the estimation result, and as a result, the estimation result may be difficult to adopt.

In addition, when the hippocampus does not show atrophy but does not correspond to AD (Alzheimer's disease), or when PET examination detects a large accumulation of amyloid β protein, it is a precursor state of AD. It may be difficult to make a highly accurate diagnosis by the conventional method of evaluating a part of the brain, such as when it does not correspond to MCI (Mild Cognitive Impairment).

Therefore, the present invention provides a diagnostic support device, a guessing device, a diagnostic support system, a diagnostic support method, a diagnostic support program, and a diagnostic support device, a guess device, a diagnostic support system, a diagnostic support method, a diagnostic support program, which makes it easy to confirm the validity of the estimation result by the trained model and can improve the diagnostic accuracy. Provide a trained model.

The diagnostic support device according to one aspect of the present invention is a first acquisition unit that acquires an image of the whole brain of a subject, and an estimation in which an image is input to a trained model to make a guess related to dementia of the subject. It includes a second acquisition unit for acquiring the result, and a display unit for displaying the estimation result and information indicating the region of interest of the whole brain in the image on which the estimation result is based.

According to this aspect, the trained model is used to display the guess result of making a guess related to dementia of the subject together with the information showing the region of interest of the whole brain in the image on which the guess result is based. , It becomes easier to confirm the validity of the estimation result by the trained model.

Here, the "whole brain image" refers to an image of all areas surrounded by the skull (in the case of a cross-sectional image, all areas surrounded by the skull in the cross section). Typically, a whole brain image of the horizontal section of the brain (axial section, axial plane) includes the frontal lobe, temporal lobe, occipital lobe, etc., and is a vertical section parallel to the anterior-posterior direction (sagittal section, sagittal plane). The sagittal plane) includes the frontal lobe, parietal lobe, occipital lobe and spinal cord, and the vertical section parallel to the left-right direction (coronary section, coronal plane) includes the temporal lobe, frontal lobe or parietal lobe, spinal cord. ..

Further, the "region of interest of the whole brain in the image on which the estimation result is based" means an region that brings about a different estimation result when the image data of that region is different. For example, there is a possibility of dementia when a trained model is input with different image data of the region of interest displayed for the image data of the whole brain image that resulted in the estimation that there is a high possibility of dementia. Is the area that gives the guess result that is low.

The "region of interest of the whole brain" refers to the region of interest within all the regions surrounded by the skull, and the "region of interest" refers to the region that caused the estimation results related to dementia.

In the above aspect, the first acquisition unit may acquire a three-dimensional image of the whole brain.

According to this aspect, by using a three-dimensional image of the whole brain, it is possible to capture the three-dimensional features of the whole brain and make a dementia-related guess.

In the above aspect, the display unit may display the region of interest of the whole brain in the three-dimensional image that is the basis of the estimation result in an aspect in which the three-dimensional position can be identified.

According to this aspect, the region of interest of the whole brain in the image on which the estimation result is based can be captured three-dimensionally, and the validity of the estimation result by the trained model can be more easily confirmed.

In the above aspect, the second acquisition unit may acquire the estimation result in which the subject estimates the risk of developing dementia within a predetermined period based on the image using the trained model.

According to this aspect, whether the estimation result regarding the risk of developing dementia in the subject within a predetermined period is valid or not is confirmed by the information indicating the region of interest of the whole brain in the image on which the estimation result is based. Can be done.

In the above aspect, the second acquisition unit may acquire the estimation result of estimating the accumulation of dementia-causing proteins such as amyloid β protein and tau protein related to the whole brain based on the image using the trained model. ..

According to this aspect, whether the estimation result of estimating the accumulation of dementia-causing proteins such as amyloid β protein and tau protein related to the whole brain of the subject is valid, or whether the estimation result of the whole brain in the image on which the estimation result is based is valid. It can be confirmed by the information indicating the area of interest.

In the above embodiment, the second acquisition unit uses the trained model to obtain the score of the test regarding the cognitive function of the subject, the age of the subject, the gender of the subject, the height of the subject, the weight of the subject, and the subject. Information on the medical history, the volume of each part of the subject's whole brain, the score representing the degree of atrophy of each part of the subject's whole brain, and at least one or more combinations of these secular data, and images. Based on the above, the guess result of making a guess related to the subject's dementia may be obtained.

According to this aspect, the accuracy of estimation can be improved by making an estimation related to dementia of the subject based on the information about the subject and the image of the whole brain of the subject using the trained model. it can.

The guessing device according to another aspect of the present invention uses a guessing unit that inputs an image of the whole brain of the subject into the trained model and makes a guess related to dementia of the subject, and the trained model. An estimation unit that estimates the region of interest of the whole brain in the image that is the basis of the result, and a provision unit that provides the result of the estimation and information indicating the region of interest of the whole brain in the image that is the basis of the estimation result to the diagnostic support device. , Equipped with.

According to this aspect, the trained model is used to calculate the estimation result of making a dementia-related estimation of the subject and the information indicating the region of interest of the whole brain in the image on which the estimation result is based. By providing it to the diagnostic support device, it becomes easier for the user who uses the diagnostic support device to confirm the validity of the estimation result by the trained model.

The diagnostic support system according to another aspect of the present invention is a diagnostic support system including a diagnostic support device and a guessing device that stores a learned model, and the diagnostic support device is an image of the whole brain of a subject. The first acquisition unit that acquires the estimation result and the second acquisition unit that acquires the estimation result of making an estimation related to dementia of the subject by inputting an image into the trained model, and the estimation result and the basis of the estimation result. It has a display unit for displaying information indicating an area of interest of the whole brain in an image.

In the above aspect, the guessing device uses the trained model to make a guess related to the subject's dementia based on the image, and the trained model to use the trained model to base the guess result on the image. It may have an estimation unit that estimates the region of interest of the whole brain.

According to this aspect, the estimation result of making a dementia-related estimation of the subject using the trained model by the estimation device, and the information indicating the region of interest of the whole brain in the image on which the estimation result is based. Can be calculated and provided to the diagnostic support device.

In the above aspect, the guessing device executes the learning process of the learning model and learns by using the learning data including the images of the whole brains of the plurality of subjects to which the actually measured data related to the dementia of the plurality of subjects are associated. It may have a generator that generates a completed model.

According to this aspect, it is possible to generate a trained model capable of making an appropriate guess using images of the whole brain of a plurality of subjects to which actual measurement data related to dementia of a plurality of subjects are associated.

In the above embodiment, the learning data includes the test score of the subject's cognitive function, the subject's age, the subject's gender, the subject's height, the subject's weight, the subject's history, the subject's volume for each part of the whole brain, and the subject's. Further including subject data including a score representing the degree of atrophy for each part of the whole brain and information of at least one or a combination of a plurality of these secular data, the generator includes images of the whole brain of the plurality of subjects and images of the whole brain of the plurality of subjects. A trained model may be generated based on the subject data.

According to this aspect, by generating a trained model based on subject data on a plurality of subjects and images of the subject's whole brain, a trained model with higher estimation accuracy related to dementia of the subject is generated. can do.

In the above aspect, the generation unit may relearn the trained model based on the image of the whole brain of the subject and the diagnosis result related to the dementia of the subject.

According to this aspect, the trained model can be improved by using the accumulated images of the whole brain of the subject and the diagnosis result.

The diagnostic support method according to another aspect of the present invention is to acquire an image of the whole brain of the subject, input the image into the trained model, and make a guess related to dementia of the subject. It includes acquiring and displaying information indicating the region of interest of the whole brain in the estimation result and the image on which the estimation result is based.

In the diagnostic support program according to another aspect of the present invention, the diagnostic support device acquires an image of the whole brain of the subject, and the image is input to the trained model to make a guess related to dementia of the subject. Acquiring the estimated result and displaying the information indicating the region of interest of the whole brain in the estimated result and the image on which the estimated result is based are executed.

The trained model according to another aspect of the present invention is used for learning processing of a learning model using learning data including images of the whole brains of a plurality of subjects to which actual measurement data related to dementia of a plurality of subjects are associated. To perform an image of the subject's whole brain to make inferences related to the subject's dementia and to estimate the area of interest in the whole brain in the image on which the results of the inference are based. It is being learned.

According to this aspect, the estimation result of the dementia-related estimation of the subject and the information indicating the region of interest of the whole brain in the image on which the estimation result is based are calculated and provided to the diagnostic support device. Can be done.

The diagnostic support device according to another aspect of the present invention relates to the dementia of the subject by inputting the first image into the trained model and the first acquisition unit that acquires the first image of the whole brain of the subject. A second image that is different from the first estimation result when the second acquisition unit that acquires the first estimation result after estimation and the second image whose data is different only in a part of the first image are input to the trained model. (2) A setting unit for acquiring a partial area from which the estimation result is obtained and a display unit for displaying the first estimation result and the partial area are provided.

In the diagnostic support method according to another aspect of the present invention, the first image of the whole brain of the subject is acquired, and the first image is input to the trained model to make a guess related to the dementia of the subject. When the first estimation result is acquired and the second image in which only a part of the first image is different data is input to the trained model, the second estimation result different from the first estimation result is acquired. Includes setting a part of the area and displaying the first estimation result and a part of the area.

Here, some areas correspond to the areas on which the estimation results are based, and differ for each target person. That is, a part of the area for a certain target person and a part of the area for a different target person are different.

According to the present invention, a diagnostic support device, a guessing device, a diagnostic support system, a diagnostic support method, a diagnostic support program, and a diagnostic support device, a guessing device, a diagnostic support system, a diagnostic support method, and a diagnostic support program that make it easy to confirm the validity of the estimation result by the trained model and improve the diagnostic accuracy. A trained model can be provided.

It is a figure which shows the network configuration of the diagnosis support system which concerns on embodiment of this invention. It is a figure which shows the functional block of the diagnosis support apparatus and the estimation apparatus which concerns on this embodiment. It is a figure which shows the physical structure of the diagnosis support apparatus which concerns on this embodiment. It is a figure which shows the example which displays the estimation result about dementia by the estimation device which concerns on this embodiment and the basis | display on the diagnosis support device. It is a figure which shows the estimation accuracy about dementia by the estimation device which concerns on this embodiment. It is a flowchart about the estimation about dementia by the diagnosis support system which concerns on this embodiment, and the display of the basis. It is a flowchart of the learning process of the 1st trained model by the guessing apparatus which concerns on this embodiment. It is a figure which shows the example which displays the estimation result about the accumulation of amyloid β protein by the estimation apparatus which concerns on this embodiment, and the basis | display on the diagnosis support apparatus. It is a figure which shows the estimation accuracy about the accumulation of amyloid β protein by the estimation apparatus which concerns on this embodiment. It is a flowchart about the estimation about the accumulation of amyloid β protein by the diagnosis support system which concerns on this embodiment, and the display of the basis.

An embodiment of the present invention will be described with reference to the accompanying drawings. In each figure, those having the same reference numerals have the same or similar configurations.

FIG. 1 is a diagram showing a network configuration of the diagnostic support system 100 according to the embodiment of the present invention. The diagnosis support system 100 includes a diagnosis support device 10, an estimation device 20 that stores a learned model, an image management server 30, and an MRI scanner 50. The diagnosis support device 10, the estimation device 20, and the image management server 30 are communicably connected by a communication network N such as the Internet or a LAN (Local Area Network). The image management server 30 and the MRI scanner are communicably connected by DICOM (Digital Imaging and Communications in Medicine) or the like.

The diagnostic support device 10 is composed of a general-purpose computer and is used by, for example, a doctor. The diagnosis support device 10 displays the estimation result of the dementia-related estimation of the subject from the estimation device 20 and the information indicating the region of interest of the brain in the MRI image on which the estimation result is based. A user such as a doctor who uses the diagnosis support device 10 can make a diagnosis related to dementia of the subject by using the displayed estimation result and the information indicating the region of interest of the brain on which the estimation result is based as an aid to the diagnosis. it can.

The guessing device 20 is composed of a general-purpose computer, and uses a trained model to make a guess related to the subject's dementia based on the image of the subject's brain acquired from the image management server 30. In addition, the estimation device 20 estimates the region of interest of the brain in the image on which the estimation result is based, using the trained model. The estimation result calculated by the estimation device 20 and the information regarding the region of interest of the brain on which the estimation result is based are provided to the diagnosis support device 10.

The image management server 30 is composed of a general-purpose computer and has a database of MRI images of the subject's brain measured by the MRI scanner 50. The image management server 30 may store training data including images of the brains of a plurality of subjects used to generate a trained model of the guessing device 20. The image management server 30 includes not only an MRI image of the subject's brain measured by the MRI scanner 50, but also a CT image of the subject's brain measured by a CT (Computed Tomography) scanner and PET (Positron Emission Tomography). ) A medical image of the brain such as a PET image of the subject's brain measured by a scanner may be stored, but considering the simplicity of examination and the like, it is preferable to use an MRI image of the brain.

The MRI scanner 50 captures an image of the inside of the subject's body using nuclear magnetic resonance. The MRI scanner 50 specifically captures an image of the subject's head, that is, an image of the brain. The captured MRI image is stored in the image management server 30, and is acquired by the estimation device 20 and the diagnosis support device 10.

FIG. 2 is a diagram showing functional blocks of the diagnostic support device 10 and the estimation device 20 according to the present embodiment. The diagnosis support device 10 includes a first acquisition unit 11, a second acquisition unit 12, and a display unit 10f. Further, the estimation device 20 includes a storage unit 21, a generation unit 22, an estimation unit 23, an estimation unit 24, and a provision unit 25.

The first acquisition unit 11 acquires an image of the subject's brain from the image management server 30. The first acquisition unit 11 may acquire a three-dimensional image of the subject's brain. Here, the three-dimensional image may be an MRI image of the brain measured by the MRI scanner 50. By using a three-dimensional image of the brain, it is possible to capture the three-dimensional features of the brain and make inferences related to dementia.

The second acquisition unit 12 uses the first trained model 21b or the second trained model 21c to make a guess related to the subject's dementia based on the image of the brain, and the guessing device 20 Get from. In addition, the second acquisition unit 12 acquires information indicating the region of interest of the brain in the image that is the basis of the estimation result from the estimation device 20.

The second acquisition unit 12 may acquire the estimation result of estimating the risk of developing dementia by the subject within a predetermined period based on the image of the brain using the first learned model 21b. The estimation result in which the subject estimates the risk of developing dementia within a predetermined period may be, for example, the result of calculating the probability that the subject develops dementia within 2 years.

Further, the second acquisition unit 12 may acquire the estimation result of estimating the accumulation of amyloid β protein related to the brain based on the image of the brain using the second trained model 21c. The estimation result of estimating the accumulation of amyloid β protein related to the brain is the estimation result of the accumulation amount of amyloid β protein, and whether the accumulation amount of amyloid β protein is above the threshold (presence or absence of accumulation of amyloid β protein). It may be the result of guessing.

The display unit 10f displays the estimation result of the dementia-related estimation of the subject and the information indicating the region of interest of the brain in the image on which the estimation result is based. In this way, the trained model is used to display the guess result of making a guess related to dementia of the subject together with the information indicating the region of interest of the brain in the image on which the guess result is based. It becomes easier to confirm the validity of the estimation result by the model.

The display unit 10f may display the region of interest of the brain in the three-dimensional image that is the basis of the estimation result in a manner in which the three-dimensional position can be identified. The display unit 10f may display the region of interest of the brain in the three-dimensional image on which the estimation result is based, for example, by a heat map superimposed on the MRI image of the brain. Further, the display unit 10f may display a three-dimensional image of the brain in a rotation-operable manner, and display a region of interest of the brain in the three-dimensional image on which the estimation result is based by a heat map superimposed on the three-dimensional image. As a result, the region of interest of the brain in the image on which the estimation result is based can be captured three-dimensionally, and the validity of the estimation result by the trained model can be more easily confirmed.

The display unit 10f may display the estimation result in which the subject estimates the risk of developing dementia within a predetermined period and the information indicating the region of interest of the brain in the image on which the estimation result is based. As a result, it is possible to confirm whether the estimation result of the risk that the subject develops dementia within a predetermined period is appropriate by the information indicating the region of interest of the brain in the image on which the estimation result is based.

Further, the display unit 10f may display the estimation result of estimating the accumulation of amyloid β protein related to the brain and the information indicating the region of interest of the brain in the image on which the estimation result is based. As a result, it is possible to confirm whether the estimation result of estimating the accumulation of amyloid β protein in the subject's brain is valid or not by the information indicating the region of interest of the brain in the image on which the estimation result is based. According to the diagnostic support system 100 according to the present embodiment, the accumulation of amyloid β protein can be estimated from the MRI image, and the brain image obtained by the MRI scanner 50 can be used without measurement by the PET scanner or collection of cerebrospinal fluid. The risk of dementia can be estimated by a relatively simple method of imaging.

The storage unit 21 of the estimation device 20 stores the training data 21a, the first trained model 21b, and the second trained model 21c. The learning data 21a includes images of the brains of a plurality of subjects to which actual measurement data related to dementia of the plurality of subjects are associated. The learning data 21a is an MRI image of the brains of a plurality of subjects or amyloid β related to the brains of a plurality of subjects, to which actual measurement data indicating whether or not the cognitive function of the plurality of subjects has deteriorated within 2 years is associated. It may be an MRI image of the brains of a plurality of subjects associated with measured data (accumulation amount or presence / absence of accumulation) indicating protein accumulation.

The generation unit 22 executes the learning process of the learning model using the learning data 21a, and generates the first trained model 21b and the second trained model 21c. The generation unit 22 uses a convolutional neural network using learning data 21a including MRI images of the brains of a plurality of subjects, which is associated with actual measurement data indicating whether or not the cognitive function of the plurality of subjects has deteriorated within two years. The learning process of the learning model composed of is executed, and the first trained model 21b is generated. In addition, the generation unit 22 is configured by a convolutional neural network using learning data 21a including MRI images of the brains of a plurality of subjects to which actual measurement data showing the accumulation of amyloid β protein related to the brains of the plurality of subjects is associated. The learning process of the training model is executed to generate the second trained model 21c. The learning process may be, for example, a process of updating the parameters of the neural network by the error back propagation method so as to minimize the loss function for evaluating the error between the guess result and the correct answer. The first trained model 21b and the second trained model 21c may be configured by, for example, a 3D CNN (Convolutional Neural Network) that accepts voxel data of a brain image as an input. The first trained model 21b and the second trained model 21c may both be configured by a convolutional neural network, but the network structures may be different. In this way, it is possible to generate a trained model capable of making an appropriate guess by using images of the brains of a plurality of subjects to which actual measurement data related to dementia of a plurality of subjects are associated.

The guessing unit 23 uses the first trained model 21b or the second trained model 21c to make a guess related to the subject's dementia based on the image of the brain. The guessing unit 23 uses the first trained model 21b to estimate the risk of the subject developing dementia within a predetermined period based on the image of the brain, or uses the second trained model 21c to estimate the risk. , The accumulation of amyloid β protein related to the brain may be inferred based on the image of the brain.

The learning data includes the test score of the subject's cognitive function, the subject's age, the subject's gender, the subject's height, the subject's weight, the subject's medical history, the volume of each part of the subject's brain, and the degree of atrophy of the subject's brain. Subject data may further include the score to be represented and information on at least one or a combination of these secular data. The subject data may be transmitted from the diagnosis support device 10 to the estimation device 20 and stored in the storage unit 21, but a part of the subject data may be taken into the estimation device 20 from an external database. Here, the volume of each part of the brain may include, for example, the volume of the forebrain basal ganglia, the volume of the posterior cingulate gyrus, and the volume of the medial temporal lobe. Further, the score representing the degree of atrophy of the brain may be represented by a Z value (a value converted so that the average is 0 and the standard deviation is 1). Then, the generation unit 22 converts the subject data into numerical values, synthesizes the feature map of the brain image calculated by the convolutional neural network and the numerical values representing the subject data, and performs, for example, a calculation by the fully connected layer to recognize the subject. A first trained model 21b and a second trained model 21c may be generated to make inferences about the disease. In this way, by generating a trained model based on the subject data of a plurality of subjects and the image of the subject's brain, it is possible to generate a trained model with higher estimation accuracy related to the subject's dementia. it can.

In this case, the guessing unit 23 makes a guess related to the subject's dementia based on the brain image and the subject data. Using the first trained model 21b or the second trained model 21c, the guessing unit 23 uses the test score for the cognitive function of the subject, the age of the subject, the gender of the subject, the height of the subject, and the subject. Information on weight, subject's medical history, volume of each part of the subject's brain, scores representing the degree of atrophy of the subject's brain, and at least one or a combination of these secular data, and brain Make inferences related to dementia in the subject based on the images. In this way, the accuracy of estimation can be further improved by making an estimation related to dementia of the subject based on the information about the subject and the image of the brain of the subject using the trained model.

The estimation unit 24 uses the first trained model 21b or the second trained model 21c to estimate the region of interest of the brain in the image on which the estimation result is based. When the first trained model 21b or the second trained model 21c is configured by a neural network, the estimation unit 24 uses, for example, LRP (Layer-wise Relevance Propagation) to describe the brain in the image on which the estimation result is based. The region of interest may be estimated.

The generation unit 22 may relearn the first trained model 21b or the second trained model 21c based on the image of the subject's brain and the diagnosis result related to the subject's dementia. Images of the subject's brain and diagnosis results related to the subject's dementia are accumulated as the doctor diagnoses the subject. The generation unit 22 may add the newly measured image of the subject's brain to the training data and relearn the first trained model 21b or the second trained model 21c. As a result, the trained model can be improved by using the accumulated images of the subject's brain and the diagnosis results.

The providing unit 25 provides the diagnosis support device 10 with the estimation result calculated by the estimation unit 23 and the information indicating the region of interest of the brain in the image on which the estimation result estimated by the estimation unit 24 is based. The providing unit 25 may transmit the estimation result and the information indicating the region of interest of the brain in the image on which the estimation result is based to the diagnosis support device 10 via the communication network N.

As described above, according to the estimation device 20 according to the present embodiment, the estimation result of the estimation related to the dementia of the subject using the trained model and the interest of the brain in the image on which the estimation result is based. Information indicating the area can be calculated and provided to the diagnosis support device 10.

FIG. 3 is a diagram showing a physical configuration of the diagnostic support device 10 according to the present embodiment. The diagnostic support device 10 includes a CPU (Central Processing Unit) 10a corresponding to a calculation unit, a RAM (Random Access Memory) 10b corresponding to a storage unit, a ROM (Read only Memory) 10c corresponding to a storage unit, and a communication unit. It has a 10d, an input unit 10e, and a display unit 10f. Each of these configurations is connected to each other via a bus so that data can be transmitted and received. In this example, the case where the diagnosis support device 10 is composed of one computer will be described, but the diagnosis support device 10 may be realized by combining a plurality of computers. Further, the configuration shown in FIG. 3 is an example, and the diagnosis support device 10 may have configurations other than these, or may not have a part of these configurations.

The CPU 10a is a control unit that controls execution of a program stored in the RAM 10b or ROM 10c, calculates data, and processes data. The CPU 10a is a calculation unit that executes a program (diagnosis support program) that displays information indicating an area of interest in the brain in an image that is the basis of the estimation result and the estimation result that made an estimation related to dementia of the subject. The CPU 10a receives various data from the input unit 10e and the communication unit 10d, displays the calculation result of the data on the display unit 10f, and stores the data in the RAM 10b.

The RAM 10b is a storage unit in which data can be rewritten, and may be composed of, for example, a semiconductor storage element. The RAM 10b may store data such as a program executed by the CPU 10a and an image of the subject's brain. It should be noted that these are examples, and data other than these may be stored in the RAM 10b, or a part of these may not be stored.

The ROM 10c is a storage unit capable of reading data, and may be composed of, for example, a semiconductor storage element. The ROM 10c may store, for example, a diagnostic support program or data that is not rewritten.

The communication unit 10d is an interface for connecting the diagnosis support device 10 to another device. The communication unit 10d may be connected to a communication network N such as the Internet.

The input unit 10e receives data input from the user, and may include, for example, a keyboard and a touch panel.

The display unit 10f visually displays the calculation result by the CPU 10a, and may be configured by, for example, an LCD (Liquid Crystal Display). The display unit 10f may display the estimation result of the dementia-related estimation of the subject and the information indicating the region of interest of the brain in the image on which the estimation result is based.

The diagnosis support program may be stored in a storage medium readable by a computer such as RAM 10b or ROM 10c and provided, or may be provided via a communication network connected by the communication unit 10d. In the diagnosis support device 10, the CPU 10a executes the diagnosis support program to realize various operations described with reference to FIG. It should be noted that these physical configurations are examples and do not necessarily have to be independent configurations. For example, the diagnosis support device 10 may include an LSI (Large-Scale Integration) in which the CPU 10a and the RAM 10b or ROM 10c are integrated.

The physical configuration of the estimation device 20 may be the same as that of the diagnosis support device 10, but it may be provided with a GPU (Graphical Processing Unit) and may not necessarily be the same. The trained model stored in the guessing device 20 is generated by a learning process of a learning model using learning data including images of the brains of a plurality of subjects to which actual measurement data related to dementia of a plurality of subjects are associated. To. The trained model performs inferences related to dementia in the subject based on images of the subject's brain and estimates the region of interest in the brain in the image on which the inference results are based. It is learned and processed as follows. Using the trained model, it is possible to calculate the estimation result of making an estimation related to the subject's dementia and the information indicating the region of interest of the brain in the image on which the estimation result is based, and provide it to the diagnosis support device 10. it can.

FIG. 4 is a diagram showing an example in which the estimation result regarding dementia by the estimation device 20 according to the present embodiment and the basis thereof are displayed on the diagnosis support device 10. In the figure, the first image IMG1, the second image IMG2, and the first estimation result PD1 displayed on the display unit 10f of the diagnosis support device 10 are shown.

The first image IMG1 and the second image IMG2 are images showing the MRI image of the subject's brain and the region of interest of the brain in the image on which the estimation result by the first trained model 21b is based. The first image IMG1 is a cross-sectional image of the brain in the sagittal plane, and the first region R1 and the second region R2 are shown as regions of interest. Further, the second image IMG2 is a cross-sectional image of the brain on the coronal plane, and the third region R3 is shown as a region of interest. In this example, the region of interest is displayed by a heat map in which pixels having a greater contribution to the estimation result calculated by the first trained model 21b are shown brighter.

The first estimation result PD1 is "probability of developing dementia within 2 years (estimation): 90%", and the first trained model 21b based on the first image IMG1 and the second image IMG2 causes dementia. The result of calculating the onset probability is shown.

The doctor using the diagnosis support device 10 diagnoses the first estimation result PD1 while confirming the validity of the first estimation result PD1 by the first trained model 21b with reference to the first image IMG1 and the second image IMG2. Can be used for.

In this example, an example is shown in which the region of interest is displayed by a heat map in which the pixel having a greater contribution to the estimation result calculated by the first trained model 21b is displayed brighter, but the region of interest is the first trained model 21b. It may be displayed by marking a part of the brain that contributes greatly to the estimation result calculated by, or may be displayed as text information separately from the brain image.

The first image IMG1, the second image IMG2, the first area R1, the second area R2, the third area R3, and the first estimation result PD1 are displayed on a terminal such as a smartphone or a PC (Personal Computer) used by the subject. May be good. In that case, the providing unit 25 transmits the estimation result and the information indicating the region of interest of the brain in the image on which the estimation result is based to the terminal of the target person. As a result, the subject can confirm his / her own state by referring to the first image IMG1, the second image IMG2, the first region R1, the second region R2, the third region R3, and the first estimation result PD1. it can. The providing unit 25 provides only the first estimation result PD1 to the terminal of the target person, and if necessary, the first image IMG1, the second image IMG2, the first area R1, the second area R2, and the third area R3. May be provided to the target person's terminal.

FIG. 5 is a diagram showing the estimation accuracy regarding dementia by the estimation device 20 according to the present embodiment. In the figure, when a part of the training data is separated as test data in advance and the first trained model 21b is generated by the estimation device 20, the estimation accuracy of the first trained model 21b is calculated using the test data. The result is shown. The "guess result" in the table of the figure is whether or not the probability that the subject develops dementia within a predetermined period is equal to or higher than the threshold value based on the image of the subject's brain according to the first trained model 21b. The result of guessing is shown. If the probability that the subject develops dementia within the predetermined period is equal to or higher than the threshold value, it is counted as the "AD progression group", and if the probability that the subject develops dementia within the predetermined period is less than the threshold value, " It is counted as "AD non-progressive group". In addition, "correct answer" indicates actual measurement data indicating whether or not the subject's cognitive function deteriorated within a predetermined period.

The count of the "guess result" of the first trained model 21b shown in FIG. 5 is the case where the score of the test related to the cognitive function of the subject is combined with the feature map of the brain image and the final judgment is made. Is going.

In the example shown in FIG. 5, as an estimation result, the AD progress group is 144 and the AD non-progress group is 77. Of the 144 cases presumed to be in the AD advanced group, 140 cases were correct answers, and 4 cases were false positives (actually, the AD non-progressive group). In addition, 74 of the 77 cases presumed to be in the AD non-progressive group were correct answers, and 3 cases were false negatives (actually, the AD advanced group). Therefore, the accuracy of the first trained model 21b is (140 + 74) / (140 + 4 + 74 + 3) × 100% = 96.8%, and the sensitivity is 140 / (140 + 3) × 100% = 97. It is 9.9%, and the specificity is 74 / (74 + 4) × 100% = 94.9%. As described above, the first trained model 21b according to the present embodiment shows high performance for all the indexes, can suppress the occurrence of false positives and false negatives to a low level, and can appropriately evaluate the risk of dementia. ..

When the first trained model 21b is generated, the guessing device 20 displays a table showing the guessing accuracy as shown in FIG. 5 on the diagnosis support device 10, and indexes such as accuracy, sensitivity, and specificity. May be displayed on the diagnostic support device 10. Further, an index such as a ROC (Receiver Operating Characteristic) curve or an AUC (Area Under the Curve) may be displayed on the diagnosis support device 10.

FIG. 6 is a flowchart relating to the estimation of dementia by the diagnostic support system 100 according to the present embodiment and the display of the basis thereof. First, the brain image of the subject is photographed by the MRI scanner 50 (S10).

After that, the estimation device 20 estimates the progression of dementia in the subject after a lapse of a predetermined period based on the brain image and the subject data using the first trained model 21b (S11). Here, the subject data is the score of the test regarding the cognitive function of the subject, the age of the subject, the gender of the subject, the height of the subject, the weight of the subject, the medical history of the subject, and the brain of the subject. Information on the volume of each site, the score representing the degree of atrophy of the subject's brain, and at least one or a combination of these secular data.

Further, the estimation device 20 estimates the region of interest of the brain image that is the basis of the estimation result by using the first trained model 21b (S12).

Finally, the diagnostic support device 10 displays the brain image, the estimation result, and the region of interest of the brain image on which the estimation result is based (S13).

FIG. 7 is a flowchart of the learning process of the first trained model 21b by the guessing device 20 according to the present embodiment. First, the guessing device 20 accumulates learning data 21a including brain images of a plurality of subjects, subject data, and information on the progression of dementia after a lapse of a predetermined period (S20).

The estimation device 20 executes estimation by the learning model using the learning data 21a (S21), and calculates an error between the estimation result and the correct answer (S22). Here, the correct answer is information on the progress of dementia after a lapse of a predetermined period, and may be information indicating whether or not dementia has progressed after the lapse of a predetermined period.

The guessing device 20 updates the parameters of the learning model so as to reduce the error (S23). After that, when the learning end condition is not satisfied (S24: NO), the guessing device 20 executes the processes S21 to S23 again. Here, the learning end condition may be that the error between the estimation result and the correct answer is not more than a predetermined value, or that the number of executions of the processes S21 to S23 is not more than a predetermined value.

On the other hand, when the learning end condition is satisfied (S24: YES), the guessing device 20 saves the generated first trained model 21b (S25). As a result, the learning process of the first trained model 21b is completed.

The guessing device 20 uses the training data 21a including brain images of a plurality of subjects, subject data, and information on the accumulation of amyloid β protein by the same processing as in the case of generating the first trained model 21b. The learning process of the completed model 21c may be performed.

FIG. 8 is a diagram showing an example in which the estimation result regarding the accumulation of amyloid β protein by the estimation device 20 according to the present embodiment and the basis thereof are displayed on the diagnosis support device 10. In the figure, the third image IMG3, the fourth image IMG4, and the second estimation result PD2 displayed on the display unit 10f of the diagnosis support device 10 are shown.

The third image IMG3 and the fourth image IMG4 are images showing the MRI image of the subject's brain and the region of interest of the brain in the image on which the estimation result by the second trained model 21c is based. The third image IMG3 is a cross-sectional image of the brain in the sagittal plane, and the fourth region R4 is shown as the region of interest. Further, the fourth image IMG4 is a cross-sectional image of the brain on the coronal plane, and the fifth region R5 is shown as a region of interest. In this example, the region of interest is displayed by a heat map in which pixels having a greater contribution to the estimation result calculated by the second trained model 21c are shown brighter.

The second estimation result PD2 is "probability (estimation) that amyloid β protein is accumulated above the threshold value (estimation): 95%", and amyloid β is determined by the second trained model 21c based on the third image IMG3 and the fourth image IMG4. The result of calculating the probability of the presence or absence of protein accumulation is shown.

The doctor using the diagnosis support device 10 diagnoses the second estimation result PD2 while confirming the validity of the second estimation result PD2 by the second trained model 21c with reference to the third image IMG3 and the fourth image IMG4. Can be used for.

In this example, an example is shown in which the region of interest is displayed by a heat map in which the pixel having a greater contribution to the estimation result calculated by the second trained model 21c is displayed brighter, but the region of interest is the second trained model 21c. It may be displayed by marking a part of the brain that contributes greatly to the estimation result calculated by, or may be displayed as text information separately from the brain image.

The third image IMG3, the fourth image IMG4, the fourth area R4, the fifth area R5, and the second estimation result PD2 may be displayed on a terminal such as a smartphone or PC used by the target person. In that case, the providing unit 25 transmits the estimation result and the information indicating the region of interest of the brain in the image on which the estimation result is based to the terminal of the target person. As a result, the subject can confirm his / her own state by referring to the third image IMG3, the fourth image IMG4, the fourth region R4, the fifth region R5, and the second estimation result PD2. In addition, the providing unit 25 provides only the second estimation result PD2 to the target person's terminal, and if necessary, provides the third image IMG3, the fourth image IMG4, the fourth area R4, and the fifth area R5 to the target person's terminal. It may be provided to.

FIG. 9 is a diagram showing the estimation accuracy regarding the accumulation of amyloid β protein by the estimation device 20 according to the present embodiment. In the figure, when a part of the training data is separated as test data in advance and the second trained model 21c is generated by the estimation device 20, the estimation accuracy of the second trained model 21c is calculated using the test data. The result is shown. The "guess result" in the table of the figure shows the result of guessing whether or not the amyloid β protein is accumulated in the brain above the threshold value by the second trained model 21c based on the image of the subject's brain. There is. If the probability that amyloid β protein is accumulated above the threshold value in the subject's brain is greater than or equal to a predetermined value, it is counted as "Positive", and the probability that amyloid β protein is accumulated above the threshold value in the subject's brain is predetermined. If it is less than the value, it is counted as "Negative". In addition, the "correct answer" indicates the result of measuring whether or not the amyloid β protein is accumulated in the subject's brain above the threshold value by collecting PET or cerebrospinal fluid.

The count of the "guess result" of the second trained model 21c shown in FIG. 9 is the case where the score of the test related to the cognitive function of the subject is combined with the feature map of the brain image to make the final judgment. Is going.

In the example shown in FIG. 9, as the estimation result, Positive is 865 and Negative is 467. Of the 865 cases presumed to be Positive, 807 cases were correct and 58 cases were false positives (actually Negative). Of the 467 cases presumed to be Negative, 427 cases were correct answers, and 40 cases were false negatives (actually Positive). Therefore, the accuracy of the second trained model 21c is (807 + 427) / (807 + 58 + 427 + 40) × 100% = 92.6%, and the sensitivity is 807 / (807 + 40) × 100% = 95. It is 3.3%, and the specificity is 427 / (427 + 58) × 100% = 88.0%. As described above, the second trained model 21c according to the present embodiment shows high performance for all the indexes, suppresses the occurrence of false positives and false negatives to a low level, and appropriately determines the presence or absence of accumulation of amyloid β protein. Can be evaluated.

When the second trained model 21c is generated, the estimation device 20 displays a table showing the estimation accuracy as shown in FIG. 9 on the diagnosis support device 10, and indexes such as accuracy, sensitivity, and specificity. May be displayed on the diagnostic support device 10. Further, an index such as an ROC curve or AUC may be displayed on the diagnosis support device 10.

FIG. 10 is a flowchart regarding the estimation of the accumulation of amyloid β protein by the diagnostic support system 100 according to the present embodiment and the display of the basis thereof. First, the brain image of the subject is taken by the MRI scanner 50 (S30).

Then, the estimation device 20 estimates the accumulation of amyloid β protein related to the subject's brain based on the brain image and the subject data using the second trained model 21c (S31). Here, the subject data is the score of the test regarding the cognitive function of the subject, the age of the subject, the gender of the subject, the height of the subject, the weight of the subject, the medical history of the subject, and the brain of the subject. Information on the volume of each site, the score representing the degree of atrophy of the subject's brain, and at least one or a combination of these secular data.

Further, the estimation device 20 estimates the region of interest of the brain image that is the basis of the estimation result using the second trained model 21c (S32).

Finally, the diagnostic support device 10 displays the brain image, the estimation result, and the region of interest of the brain image on which the estimation result is based (S33).
The embodiments described above are for facilitating the understanding of the present invention, and are not for limiting and interpreting the present invention. Each element included in the embodiment and its arrangement, material, condition, shape, size, etc. are not limited to those exemplified, and can be changed as appropriate. In addition, the configurations shown in different embodiments can be partially replaced or combined.

In the above-mentioned diagnostic support device 10, the first image IMG1 (FIG. 4), which is the sagittal plane of the subject, shows a brain image of the entire region surrounded by the skull. Equivalent to. In addition, the second image IMG2 (FIG. 4), which is the coronal plane of the subject, shows a brain image of the entire region surrounded by the skull, and thus corresponds to an image of the whole brain.

Therefore, the learning data 21a stored in the storage unit 21 of the estimation device 20 includes images of the whole brains of a plurality of subjects. In addition, the generation unit 22 executes the learning process of the learning model composed of the convolutional neural network that inputs the learning data 21a including the MRI images of the whole brains of a plurality of subjects to execute the first trained model 21b and the first trained model 21b. 2 Generate a trained model 21c. The guessing unit 23 uses an image of the whole brain of the subject as an input to make a dementia-related guess, and the guessing unit 24 estimates the region of interest of the whole brain in the image on which the guess result is based. Here, the first region R1 of the first image IMG1 indicates a region near the occipital lobe in the whole brain, the second region R2 indicates a region near the frontal lobe, and the third image IMG3 is temporal. It shows the area near the leaves. Therefore, the estimation unit 24 is configured to be able to estimate the region of interest within all the regions surrounded by the skull. Further, the providing unit 25 provides the estimated region of interest of the whole brain to the diagnosis support device 10.

Further, the first acquisition unit 11 of the diagnosis support device 10 acquires an image of the whole brain of the subject as an input. The second acquisition unit 12 acquires from the estimation device 20 the estimation result of making an estimation related to the dementia of the subject based on the image of the whole brain of the subject. Further, the display unit 10f displays the estimation result of the dementia-related estimation of the subject and the information indicating the region of interest of the whole brain in the image on which the estimation result is based.

According to such a configuration, it is possible to acquire an image of the whole brain of the subject and input this image into the trained model to obtain the estimation result of making a guess related to the subject's dementia. It will be possible. For this reason, it becomes possible to perform evaluation based on the relationship between a plurality of regions, which was difficult to evaluate by the method of dividing and evaluating brain images for each voxel. For example, when there is a certain amount of dementia-causing protein accumulation in each of the frontal lobe, occipital lobe, and temporal lobe, the method of dividing and evaluating brain images has a small accumulation amount of dementia-causing protein in each region. Even when the evaluation result that the possibility of dementia is low is obtained for some reason, according to the diagnostic support device 10 according to the present invention, evaluation based on the relationship between a plurality of areas and a plurality of areas can be obtained. Since it is possible to make an evaluation due to a cause that crosses over, for example, dementia is caused by a certain amount of dementia-causing protein accumulation in each of the frontal lobe, occipital lobe, and temporal lobe. It is possible to obtain a speculative result that is likely to be. Furthermore, it is possible to expect the discovery of a new region of the brain related to dementia or a new correlation, which has not been recognized by conventional technical common sense. In addition, even a doctor who specializes in neurosurgery or the like who does not have sufficient expertise in identifying individual parts of the brain can easily make a diagnosis.

Further, the estimation unit 24 estimates the region of interest in all the regions surrounded by the skull, and the display unit 10f displays the region of interest in all the regions surrounded by the skull. For example, the display unit 10f displays related regions covering a plurality of regions such as the frontal lobe, the occipital lobe, and the temporal lobe, as shown in the first image IMG1 to the third image IMG3. Therefore, the user of the diagnostic support device 10 can obtain new knowledge that the relationship between a plurality of areas can cause dementia. It is difficult to obtain the effect brought about by such a configuration by VBM (Voxel Based Morphometry) in which a brain image is divided and evaluated for each voxel, or a PET examination in which only a site that absorbs a drug is displayed.

Further, since the region of interest of the whole brain displayed by the display unit 10f is the basis of the estimation result, it usually differs depending on the subject. For example, in the case of a subject with frontotemporal dementia, the frontal lobe or the temporal lobe may be displayed as an area of interest, but in the case of a subject with vascular dementia, the site where the blood vessel is damaged may be displayed. It may be displayed as an area of interest. The region of interest is represented by highlighting the target pixel, for example, increasing the brightness of the target pixel. Therefore, even within the same frontal lobe, different regions of interest are usually displayed depending on the subject. Such a configuration is quite different from simply highlighting areas previously known to be likely to cause dementia (eg, the frontal lobe).

As a method of extracting the region of interest, that is, the region of the whole brain that is the basis of the estimation result, the above-mentioned LRP is a method of calculating the influence of the input value on the output value. For example, after propagating a neural network such as CNN to a certain layer, it is possible to identify a part having a large influence by setting a value other than the part to be examined to zero and back-propagating. In addition to LRP, a method of changing the input information to create a plurality of input information, inputting them into the same trained model, and comparing the results is considered to acquire the area on which the estimation result is based. Be done. Specifically, first, the estimation unit 23 inputs an image of the entire brain of the subject (an example of the "first image") into the first trained model 21b or the second trained model 21c, and the estimation result ("" An example of "first estimation result") is acquired. Next, the estimation unit 24 generates a plurality of images of the whole brain (an example of the "second image") in which only a part of the first image is different. For example, a second image in which a part of the pixel values of the frontal lobe is masked, a second image in which a part of the pixel values of the occipital lobe are masked and changed to different values, and the like are generated. Then, the guessing unit 23 inputs the generated second image into the first trained model 21b or the second trained model 21c to acquire a plurality of guessing results (an example of the "second guessing result"). .. Then, the estimation unit 24 compares the first estimation result and the second estimation result, and determines a second image in which different estimation results are obtained. The estimation unit 24 acquires the masked region as the region of interest on which the estimation result is based in the second image that brings about the second estimation result different from the first estimation result, and provides the masked region to the providing unit 25. The display unit 10f increases the brightness of the pixels corresponding to the region of interest and superimposes them on the image of the entire brain of the subject for display. When three or more types of estimation results are selected, the region of interest can be configured to include two or more types of regions (a grounded first region of interest and a more grounded second region of interest). Become.

According to the above configuration, it is easy to confirm the validity of the estimation result by the trained model, and it is possible to improve the diagnostic accuracy. Diagnosis support device, estimation device, diagnosis support system, diagnosis support method, diagnosis. It will be possible to provide support programs and trained models.

10 ... Diagnosis support device, 10a ... CPU, 10b ... RAM, 10c ... ROM, 10d ... Communication unit, 10e ... Input unit, 10f ... Display unit, 11 ... First acquisition unit, 12 ... Second acquisition unit, 20 ... Guess Device, 21 ... storage unit, 21a ... learning data, 21b ... first trained model, 21c ... second trained model, 22 ... generation unit, 23 ... estimation unit, 24 ... estimation unit, 25 ... providing unit, 30 ... Image management server, 50 ... MRI scanner, 100 ... diagnostic support system

Claims

The first acquisition unit that acquires an image of the entire brain of the subject,
A second acquisition unit that inputs the image into the trained model and acquires the estimation result of making an estimation related to the dementia of the subject, and
A display unit that displays information indicating the estimation result and the region of interest of the whole brain in the image that is the basis of the estimation result, and
Diagnostic support device equipped with.
The first acquisition unit acquires a three-dimensional image of the whole brain.
The diagnostic support device according to claim 1.
The display unit displays the region of interest of the whole brain in the three-dimensional image, which is the basis of the estimation result, in a manner in which the three-dimensional position can be identified.
The diagnostic support device according to claim 2.
The second acquisition unit acquires the estimation result of estimating the risk of the subject developing dementia within a predetermined period based on the image using the learned model.
The diagnostic support device according to any one of claims 1 to 3.
The second acquisition unit acquires the estimation result of estimating the accumulation of the dementia-causing protein related to the whole brain based on the image using the trained model.
The diagnostic support device according to any one of claims 1 to 3.
The second acquisition unit acquires the estimation result of estimating the accumulation of amyloid β protein with respect to the whole brain based on the image using the trained model.
The diagnostic support device according to any one of claims 1 to 3.
Using the trained model, the second acquisition unit uses the trained model to obtain a test score for the cognitive function of the subject, the age of the subject, the gender of the subject, the height of the subject, the weight of the subject, and the like. At least one or more of the medical history of the subject, the volume of the subject for each part of the whole brain, the score representing the degree of atrophy of the subject for each part of the whole brain, and these secular data. Based on the information of the combination of the above and the image, the estimation result of making the estimation related to the dementia of the subject is acquired.
The diagnostic support device according to any one of claims 1 to 6.
An inference unit that inputs an image of the subject's entire brain into the trained model and makes inferences related to the subject's dementia.
Using the trained model, an estimation unit that estimates the region of interest of the whole brain in the image that is the basis of the estimation result, and an estimation unit.
A providing unit that provides the diagnostic support device with information indicating the result of the estimation and the region of interest of the whole brain in the image on which the result of the estimation is based.
A guessing device equipped with.
It is a diagnostic support system including a diagnostic support device and a guessing device that stores a trained model.
The diagnostic support device is
The first acquisition unit that acquires an image of the entire brain of the subject,
A second acquisition unit that inputs the image into the trained model and acquires the estimation result of making an estimation related to the dementia of the subject.
It has a display unit that displays information indicating the region of interest of the whole brain in the image, which is the basis of the estimation result and the estimation result.
Diagnostic support system.
The guessing device
A guessing unit that makes a dementia-related guess for the subject based on the image using the trained model.
Using the trained model, it has an estimation unit that estimates the region of interest of the whole brain in the image, which is the basis of the estimation result.
The diagnostic support system according to claim 9.
The guessing device
Generation that executes the learning process of the learning model and generates the trained model by using the learning data including the images of the whole brains of the plurality of subjects to which the measured data related to the dementia of a plurality of subjects are associated. Have a part,
The diagnostic support system according to claim 10.
The learning data includes the score of the test related to the cognitive function of the subject, the age of the subject, the gender of the subject, the height of the subject, the weight of the subject, the medical history of the subject, and each part of the whole brain of the subject. Further includes subject data including information on the volume of the subject, a score representing the degree of atrophy of the whole brain of the subject, and information on at least one or a combination of these secular data.
The generation unit generates the trained model based on the images of the whole brains of the plurality of subjects and the subject data.
The diagnostic support system according to claim 11.
The generation unit relearns the trained model based on the image of the whole brain of the subject and the diagnosis result related to the dementia of the subject.
The diagnostic support system according to claim 11 or 12.
Acquiring an image of the subject's whole brain and
By inputting the image into the trained model and obtaining the estimation result of making the estimation related to the dementia of the subject,
To display the information indicating the estimation result and the region of interest of the whole brain in the image on which the estimation result is based.
Diagnostic support methods including.
For diagnostic support equipment
Acquiring an image of the subject's whole brain and
By inputting the image into the trained model and obtaining the estimation result of making the estimation related to the dementia of the subject,
To display the information indicating the estimation result and the region of interest of the whole brain in the image on which the estimation result is based.
Diagnostic support program to execute.
By the learning process of the learning model using the learning data including the images of the whole brains of the plurality of subjects to which the measured data related to the dementia of the plurality of subjects are associated.
By inputting an image of the subject's whole brain and making inferences related to the subject's dementia,
To estimate the region of interest of the whole brain in the image, which is the basis of the result of the estimation,
A trained model that has been trained to run.