System, control method, and computer program

Abstract

To provide a system for determining a state of dementia by using an image discrimination technique in cooperation with a cognitive ability test score.SOLUTION: A system (1) has a first input module (10) configured to obtain a first evaluation index based on data related to the physical state of a subject's brain, a second input module (20) configured to obtain a second evaluation index based on data related to the function of the subject's brain, and an estimation module (30) configured to estimate the state of dementia of the subject on the basis of an evaluation value obtained by a first evaluation function having the first evaluation index and the second evaluation index as variables.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a system capable of evaluating dementia. [Background Art]

[0002] Patent Document 1 describes that a driving aptitude diagnosis apparatus and a driving aptitude diagnosis method are provided, which are less susceptible to influences such as the test environment, the physical condition and mental state of the subject, and the arbitrariness of the examiner, and can diagnose the driving aptitude of the subject with high reliability. The driving aptitude diagnosis apparatus includes a white matter lesion testing means for testing the degree of cerebral white matter lesions in a subject, and a driving aptitude determining means for determining the driving aptitude of the subject based on the test result of the white matter lesion testing means, wherein the driving aptitude determining means determines that the driving aptitude of the subject is inappropriate when the degree of white matter lesions tested by the white matter lesion testing means is equal to or greater than a specified value. [Prior Art Documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Unexamined Patent Application Publication No. 2011-206452 [Summary of the Invention] [Problems to be Solved by the Invention]

[0004] Matsuda, Hiroshi et al. disclose a technique for image differentiation between Alzheimer's disease and dementia with Lewy bodies using brain images in "Differentiation Between Dementia With Lewy Bodies And Alzheimer’s Disease Using Voxel-Based Morphometry Of Structural MRI: A Multicenter Study." (Neuropsychiatric Disease and Treatment 15 (2019): 2715.)

[0005] In "Cognitive loss in dementia with Lewy bodies and Alzheimer disease" by Shimomura, Tatsuo et al. (Archives of Neurology 55.12 (1998): 1547-1552), it is disclosed that there is a bias in the distribution of specific cognitive ability test scores (WAIS-R Block Design test score or ADAS delayed recall score) between Alzheimer's disease and Lewy body dementia.

[0006] However, no technology has been disclosed that combines image recognition techniques with cognitive ability test scores to determine the state of dementia. [Means for solving the problem]

[0007] One aspect of the present invention is a system comprising: a first input module configured to acquire a first evaluation index based on data relating to the physical state of a subject's brain; a second input module configured to acquire a second evaluation index based on data relating to the function of the subject's brain; and an estimation module configured to estimate the state of dementia and / or other brain disorders of a subject based on evaluation values ​​obtained by a first evaluation function with the first and second evaluation indices as variables. The evaluation results of the data relating to the physical state of the brain and the evaluation results of the data relating to the function of the brain can be convolved into a single collaborative evaluation result using the evaluation function, enabling evaluation of the state of dementia with higher accuracy. Therefore, a system can be provided to determine or assist in determining the disease state, including the presence or absence of brain disease in a subject. Furthermore, if the subject is included in a group that has ingested pharmaceuticals (including investigational drugs and unapproved drugs), food and beverages, supplements, etc., the estimation module may include a function (unit) to evaluate the effects of these on dementia. The estimation module may also include a function to estimate the disease state of a first causative disease.

[0008] Another aspect of the present invention is a method for controlling a system. The system comprises a first input module configured to acquire a first evaluation index based on data relating to the physical state of a subject's brain, a second input module configured to acquire a second evaluation index based on data relating to the function of the subject's brain, and an estimation module configured to estimate the subject's state of brain impairment, including dementia. The method includes the following steps: 1. The estimation module obtains the first evaluation index and the second evaluation index via the first input module and the second input module. 2. Estimate the dementia status of the subject based on the evaluation value obtained by a first evaluation function with the first and second evaluation indices as variables. The estimated dementia status may include at least one of the pieces of information necessary for the prevention and treatment of dementia, such as the presence or absence of dementia, the stage, and the underlying disease.

[0009] One further different aspect of the present invention is a program. The program (program product) has instructions for a computer to perform the following actions: to obtain a first evaluation index based on data relating to the physical state of the subject's brain; to obtain a second evaluation index based on data relating to the function of the subject's brain; and to estimate the subject's dementia state based on an evaluation value obtained by a first evaluation function with the first and second evaluation indices as variables. The program may be provided on a computer-readable recording medium. [Brief explanation of the drawing]

[0010] [Figure 1] A block diagram illustrating the general outline of the dementia assessment system. [Figure 2] A flowchart illustrating dementia assessment methods. [Figure 3] A diagram showing examples of tests or examinations that measure brain health. [Figure 4] Following Figure 3, this figure shows examples of tests or examinations that measure brain health. [Figure 5] Examples of cognitive ability test scores. [Figure 6] An example of an odds table. [Figure 7] A diagram showing results of evaluation based on odds. [Figure 8] A diagram showing results of logistic regression evaluation. [Figure 9] A diagram showing results of evaluation using Z-scores. [Figure 10] A diagram showing categorization. [Figure 11] It is a schematic diagram showing a first example of the configuration of the system according to the third embodiment. [Figure 12] It is a schematic diagram showing an example of an evaluation result of a hippocampal region. [Figure 13] It is a schematic diagram showing an example of an evaluation result of a middle temporal gyrus region. [Figure 14] It is a schematic diagram showing an example of evaluation values for respective sites. [Figure 15] It is a schematic diagram showing a first example of a site comparison screen. [Figure 16] It is a schematic diagram showing a second example of a site comparison screen. [Figure 17] It is a schematic diagram showing a third example of a site comparison screen. [Figure 18] It is a schematic diagram showing an example of correlation of Z-scores between respective regions of the left brain and the right brain. [Figure 19] It is a schematic diagram showing an example of an aggregation method for the left brain and the right brain. [Figure 20] It is a schematic diagram showing a specific example of whole-brain evaluation values. [Figure 21] It is a schematic diagram showing a first example of a whole-brain evaluation screen. [Figure 22] It is a schematic diagram showing a second example of a whole-brain evaluation screen. [Figure 23] It is a schematic diagram showing a third example of a whole-brain evaluation screen. [Figure 24] It is a schematic diagram showing a fourth example of a whole-brain evaluation screen. [Figure 25] It is a schematic diagram showing a fifth example of a whole-brain evaluation screen. [Figure 26] It is a schematic diagram showing a second example of the configuration of the system according to the third embodiment. [Figure 27A]This is a schematic diagram showing an example of an ROC curve. [Figure 27B] This is a schematic diagram showing an example of an ROC curve. [Figure 28] This is a flowchart illustrating the first example of the procedure for outputting evaluation results for each part of the brain. [Figure 29] This is a flowchart illustrating the procedure for the second example of outputting evaluation results for each part of the brain. [Figure 30] This flowchart shows the procedure for outputting the whole-brain assessment results. [Modes for carrying out the invention]

[0011] (First Embodiment) Figure 1 shows a system 1 for evaluating brain disorders such as dementia according to the present invention (dementia evaluation system). This system 1 includes a first input module 10 configured to acquire a first evaluation index X1 based on data relating to the physical state of the subject's brain, a second input module 20 configured to acquire a second evaluation index X2 based on data relating to the function of the subject's brain, and an estimation module 30 configured to estimate the state of brain disorders, including dementia (dementia and / or other brain disorders), of the subject based on an evaluation value fv obtained by a first evaluation function f1 with the first evaluation index X1 and the second evaluation index X2 as variables. The system 1 further includes a database 19 in which image data including brain images 18 of the subject are stored, and a database 29 in which clinical information 28 including cognitive ability test results of the subject are stored.

[0012] Brain disorders include primarily higher-order brain disorders such as dementia, attention deficit, memory impairment, executive function disorder, social behavior disorder, aphasia, apraxia, and acognition. Dementia includes AD (Alzheimer's disease), DLB (Dementia with Lewy Bodies), and other degenerative dementias, such as frontotemporal dementia, progressive supranuclear palsy, corticobasal degeneration, and argyrophilic grain dementia. The state of brain disorders includes various aspects related to the brain disorders of the subject (patient, examinee, user), such as the presence or absence of brain disorders, their progression, the presence and differentiation of the underlying disease causing the brain disorders (causative illnesses) such as dementia, and the progression of one or more underlying diseases.

[0013] One of the objectives of System 1 is to enable numerical quantification of various states of brain disorders in subjects, including dementia, brain disorders including dementia, and brain disorders not including dementia, through a combination of multiple assessments, thereby enabling categorization of these states and providing information for further evaluation and analysis within each category. Categorizing and evaluating subjects with various brain disorders is effective not only in the treatment of subjects but also in other fields such as clinical research. For example, System 1 is also effective in evaluating the effects and impacts of various products consumed by different people, such as pharmaceuticals, foods and beverages, and supplements, on the brain or brain disorders including dementia, and can be used as a stratification marker. Furthermore, System 1 may also be applicable to the evaluation of other applications that may affect the brain or brain disorders, such as information devices, games, and other devices. The following describes an example of evaluating the state of dementia, including other objectives and effects of this system.

[0014] The first input module 10 includes a first evaluation unit 11 that obtains a first evaluation index X1 by statistically evaluating a first type of medical image of at least some of the regions of interest of the subject's brain, and a second evaluation unit 12 that obtains a first evaluation index X1 by evaluating the subject's medical image using a first model that has been machine-trained to evaluate a first disease based on the first type of medical image.

[0015] Various types of tomography devices (modalities) such as CT (Computed Tomography), MRI (Magnetic Resonance Imaging), PET (Positron Emission Tomography), SPECT (Single Photon Emission Computed Tomography), and PET-CT are known as devices for diagnosing the morphology and function of test subjects (subjects), and their modality images (medical images) are used in the diagnosis of various diseases. In particular, modality images (medical images) of a subject's brain 18 are used to acquire data related to the physical state of the subject's brain and are used in the diagnosis of diseases such as dementia and Parkinson's disease. In this specification, the physical state of the brain refers to a state in which the brain can be measured by physical methods, and refers to a state that can be evaluated, measured, and estimated using methods such as statistical processing and learning models based on various modality images, including MRI, PET, and SPECT, which can measure morphology, glucose metabolism, blood flow, etc.

[0016] Examples of medical imaging types include CT and MRI, which can reflect highly accurate morphological information. Other examples of medical imaging types include PET and SPECT, which generate images by administering a radioactive agent into the subject's body via intravenous injection or other means, and imaging the radiation emitted from the agent within the body. Images using drugs allow physicians to understand not only the morphology of various parts of the body, but also how the administered drug is distributed within the body, or the accumulation of substances in the body that react with the drug, thus contributing to improved accuracy in disease diagnosis. For example, by using a radioactive agent (tracer) commonly known as Pittsburgh compound B for PET imaging, and measuring the degree of amyloid-beta protein accumulation in the brain based on the acquired PET images, it can be useful in the differential diagnosis or early diagnosis of Alzheimer's disease.

[0017] One example of SPECT imaging is DatSCAN (Dopamine transporter SCAN), an imaging technique that visualizes the distribution of dopamine transporters (DATs) in SPECT scans administered with the radiopharmaceutical 123I-ioflupane. The purposes of this imaging include early diagnosis of Parkinsonian syndrome (PS) in Parkinson's disease (PD), assistance in the diagnosis of Lewy body dementia (DLB), and decision-making regarding drug treatment, such as levodova, in cases of dopamine neuronal loss in the striatum.

[0018] The first evaluation unit 11 uses statistical processing to evaluate medical images. The first evaluation unit 11 may also perform evaluation by statistically comparing the brain images of the subject with those of a healthy person. As a method for evaluating brain atrophy using brain images, VBM (Voxel Based Morphometry) is known, which is performed by image processing of brain images acquired by imaging the subject's head, using voxels, which are three-dimensional pixels, as the unit. A typical statistical processing is the generation of a Z-score map. Therefore, the first evaluation unit 11 may acquire the Z-score as the first evaluation index X1.

[0019] Taking MR images as an example, the Z-score is created by substituting the values ​​of the data (normal standard brain) created by calculating the mean and standard deviation for each voxel from MR images of normal cases that have undergone brain morphology standardization processing, and the values ​​of the subject's image data (processed image), into the following formula for calculating the Z-score. z=(M(x,y,z)-I(x,y,z)) / SD(x,y,z) M and SD represent the mean and standard deviation images of a normal brain, while I represents the processed image. By using a Z-score map, it is possible to quantitatively analyze which areas have changed and in what ways compared to a normal brain. For example, voxels with positive Z-score values ​​indicate areas of atrophy compared to a normal brain, and a larger value indicates a greater statistical deviation. For example, a Z-score of "2" means that the deviation exceeds twice the standard deviation from the mean, and is considered statistically significant at a significance level of approximately 5%. To quantitatively evaluate atrophy within a region, M, SD, and I are calculated for each region of interest, and the average of all positive Z-scores is then calculated.

[0020] Various statistical methods have been proposed, such as comparing the volume or area of ​​different brain regions, or using t-tests with general linear models (GLMs).

[0021] By using a radiopharmaceutical (tracer) commonly known as Pittsburgh compound B for PET imaging, the degree of amyloid-beta protein accumulation in the brain can be measured based on the acquired PET images, which can be useful in the differential or early diagnosis of Alzheimer's disease. These PET images can be statistically analyzed using the SUVR value (Standardized Uptake Value Ratio, cerebellar ratio SUVR), which represents the ratio of the sum of amyloid-beta protein accumulation (SUV, Standardized Uptake Value) in the cerebral gray matter to the amyloid-beta protein accumulation (SUV) in the cerebellum. SUVR can be defined by the following formula.

number

[0022] For statistical processing of DatSCAN using SPECT images, the Binding Ratio (BR) can be used as the evaluation (index value), and is expressed by the following formula.

number

[0023] The second evaluation unit 12 evaluates the subject's brain images 18 using a first model (learned model) that has been machine-trained to evaluate a first disease, such as AD (Alzheimer's disease) or DLB (Dementia with Lewy Bodies), based on medical images of a type common to or different from that of the first evaluation unit 11.

[0024] Using machine learning models based on medical image information, it is being practiced to differentiate diseases from medical images of subjects. Iizuka, Tomomichi et al.'s "Deep-learning-based imaging-classification identified cingulate island sign in dementia with Lewy bodies." (Scientific reports 9.1 (2019): 1-9.) reported that they achieved an even higher accuracy of 89.32% in an experiment using a convolutional neural network on perfusion SPECT images, and that deep learning focuses on occipital lobe blood flow findings, which have been conventionally used in image interpretation, when making these differentiations. Litjens, Geert et al.'s "A survey on deep learning in medical image analysis." (Medical image analysis 42 (2017): 60-88.) and Wen, Junhao et al.'s "Convolutional Neural Networks for Classification of Alzheimer's Disease: Overview and Reproducible Evaluation." (CoRR abs / 1904.07773 (2019)) report that the application of recent deep learning techniques has shown high accuracy in the differential diagnosis of Alzheimer's Disease.

[0025] The second evaluation unit 12 may obtain the output softmax value Xa of the activation function when estimating the first causative disease using the deep learning differential diagnosis model as the first evaluation index X1. The first input module 10 may output the following value obtained by the first evaluation unit 11 and / or the second evaluation unit 12 as the first evaluation index X1. Xa: The output softmax value of the activation function when estimating the first causative disease using a deep learning differential diagnosis model that takes brain images as input. Xb: The output softmax value of the activation function when estimating the first causative disease using a deep learning differential diagnosis model, based on the region of interest obtained by statistical processing of brain images. Xc: Volume or blood flow of a region of interest obtained through statistical processing of brain images. Xd: Z-score value for evaluating the volume or blood flow of a region of interest through statistical processing of brain images. Xe: The volume value or blood flow rate of the region of interest when estimating the first causative disease using a deep learning differential diagnosis model that takes brain images as input. Xf: Z-score value of the volume or blood flow evaluation of the region of interest when estimating the first causative disease using a deep learning differential diagnosis model that takes brain images as input.

[0026] The second input module 20 includes a configuration for acquiring an evaluation of clinical information 28, including a cognitive ability test, as a second evaluation index X2. This second input module 20 includes a unit 22 that performs the evaluation of the cognitive ability test and a unit 21 that evaluates other clinical information related to the user's attributes.

[0027] Cognitive ability tests are used as a means of obtaining data related to the function of a subject's brain, particularly as a test to check the cognitive function of the brain. The content of cognitive ability tests can include, but are not limited to, calculations such as addition and subtraction, Stroop tests, N-Back tests, and word recall tests. Specific examples of cognitive ability tests are disclosed, for example, in Japanese Patent Publication No. 2019-75071 of the applicant of this application, and include tests related to "forward recitation" and "backward recitation," tests related to "Stroop tests," tests related to "addition" and "subtraction," tests related to N-Back (e.g., 1-Back), and tests related to "immediate recall" (word retrieval). It should be noted that cognitive ability tests are not limited to these, and other tests or examinations of similar forms may be employed, such as those listed in Figures 3 and 4, which measure the health of the brain (including the state of cognitive function and the presence and degree of brain and mental diseases).

[0028] By combining these tests, either individually or in whole, it is possible to provide cognitive ability tests suitable for assessing the overall picture of brain function and the state of being affected by specific underlying diseases (causative illnesses). In this specification, brain function refers to abilities that can be judged by artificial actions involving the brain, such as expression and comprehension, rather than the physical state of the brain. Typically, brain function may be appropriately judged by the results of appropriate cognitive ability tests.

[0029] The results of a cognitive ability test can be used to score the state of brain function based on, for example, the subject's reaction time (response time) and the number of correct answers to the cognitive ability test (hereinafter, this scored value will be referred to as the "cognitive ability score"). The cognitive ability score information may also be used as a second evaluation index X2 to assess the subject's brain function. By representing the results of the cognitive ability test (cognitive ability score) as a normal distribution, brain age can be estimated, and the evaluation index may be corrected using the subject's age. The evaluation index for assessing the state of brain function can be determined based on clinical information, including the cognitive ability test score. In addition to the cognitive ability test score, clinical information may include any of the following: age, sex, educational history, work history, genes (ApoE, etc.), blood test results, or interview results (ADL interview, etc.).

[0030] The second assessment index X2, based on data related to brain function, is also effective as an indicator that accurately shows the estimated range of dementia risk, from pre-MCI (pre-Mild Cognitive Impairment) to MCI (Mild Cognitive Impairment) and then to AD (Alzheimer's Disease).

[0031] A typical example of an estimation module 30 that estimates the dementia state of a subject based on an evaluation value fv obtained by a first evaluation function f1 with the first evaluation index X1 and the second evaluation index X2 described above as variables is a function (disease estimation function, unit) 35 that estimates the prevalence of a first causative disease, such as AD or DLB. The first input module 10 includes a configuration for obtaining a first evaluation index X1 related to the differentiation of the first causative disease, and the second input module 20 includes a configuration for obtaining a second evaluation index X2 related to the differentiation of the first causative disease. The second input module 20 may also include a configuration for obtaining a second evaluation index X2 that includes the results 28 of a cognitive ability test suitable for differentiating the first causative disease.

[0032] A different example of the estimation module 30 is a function (clinical evaluation function, clinical evaluation unit) 36 that evaluates the effect of ingested substances on dementia when a subject is included in a group that has ingested at least one of the following: a drug, food, or supplement. The evaluation results in the evaluation function 36 can be used as stratification markers corresponding to biomarkers in stratified medicine. Therefore, System 1 may include a module that provides information as stratification markers based on the estimations of the estimation module 30. In this System 1, it is possible to provide quantified stratification markers through the collaboration of two or more factors, as will be explained in more detail below, making it possible to classify patients into purpose-specific categories using quantified markers, or to identify categories necessary for evaluation in research or medical care. Further explanation will be given below using a unit 35 that estimates disease status as an example.

[0033] The disease estimation unit 35 includes an odds determination unit 31 that determines whether or not a person is diseased based on odds, and a probability determination unit 32 that determines the probability of disease. The odds determination unit 31 includes a configuration in which the evaluation value s is calculated by the following first evaluation function f1a, with the first evaluation index X1 being the odds x1 of the first causative disease, and the second evaluation index X2 being the odds x2 of the first causative disease. s = x1 × x2 ... (f1a)

[0034] The probability determination unit 32 includes a configuration in which, when the first evaluation index X1 and the second evaluation index X2 are xi, and the weight coefficients of their respective values ​​are wi, the disease probability p is calculated as an evaluation value using the following first evaluation function f1b.

number

[0035] According to the logistic regression model, the log odds of the disease probability p for one of the two classes, i.e., one of the causative diseases A (AD or DLB), can be expressed by the following formula.

number

[0036] According to the evaluation function f1b, if the probability of disease p is greater than 0.5, it is determined that the patient is disease A, and if it is 0.5 or less, it is determined that the patient is not disease A. When evaluating multiple causative diseases (multiple classes), multiple pairs of one vs. all (one vs. rest) can be created, and the causative disease (class) with the highest p value may be determined to be the relevant one. The desired first evaluation function f1c for causative diseases based on multivalued logistic regression may be as follows.

number

[0037] Figure 2 shows a flowchart illustrating the evaluation method using the dementia assessment system 1. The dementia assessment system 1 can be provided as an information processing device equipped with computer resources including memory and CPU, and the control method can be provided as a program having instructions that can be executed on a computer. The program (program product) may be provided by recording it on a computer-readable recording medium, or it may be provided in a form that can be downloaded from the internet, etc. Furthermore, the dementia assessment system 1 may be provided as a service via the internet (SaaS (Software as a Service)).

[0038] In a system 1 having a first input module 10, a second input module 20, and an estimation module 30 configured to estimate the dementia state of a subject, in step 51, the estimation module 30 obtains a first evaluation index X1 from the first input module 10, and in step 52, the estimation module 30 obtains a second evaluation index X2 via the second input module 20. Furthermore, in step 53, the estimation module 30 estimates the dementia state of the subject based on evaluation values ​​obtained by a first evaluation function, for example, the evaluation function f1a or f1b described above, with the first evaluation index X1 and the second evaluation index X2 as variables.

[0039] In step 53, the estimation module 30 may perform a process 54 to estimate the prevalence of the first causative disease. If the subject is included in the group that has ingested at least one of the following: a drug, food or beverage, or supplement, the estimation module 30 may also perform a process 58 to evaluate the effect of the ingested substance on dementia.

[0040] In step 51, the estimation module 30 may obtain a value obtained by the first evaluation unit 11 from the statistical evaluation of the medical image via the first input module 10 as the first evaluation index X1, or a value obtained by the second evaluation unit 12 from the evaluation of the medical image as the first evaluation index X1, or, as shown in values ​​Xa to Xf, obtain a first evaluation index X1 that reflects both the value obtained from the statistical evaluation of the medical image and the value obtained when the subject's medical image was evaluated using a machine learning model (first model).

[0041] Furthermore, in step 52, the estimation module 30 may obtain an evaluation of clinical information, including a cognitive ability test, as a second evaluation index X2. Also, in steps 51 and 52, the estimation module 30 may obtain a first evaluation index X1 for the differentiation of the first causative disease and a second evaluation index X2 for the differentiation of the first causative disease, and in step 53, the prevalence of the first causative disease may be estimated using the first evaluation function.

[0042] In step 53, the estimation module 30 may perform a process to estimate the first causative disease if the evaluation value from the evaluation function exceeds a first threshold. An example of an evaluation function is the process 55 that performs odds determination and the process 56 that performs probability determination using a logistic regression model, as described above. The process 55 that performs odds determination will be explained further below.

[0043] Figure 5 shows an example of the distribution (number of errors) of delayed recall scores for the cognitive ability test (ADAS-Jcog) by disease group, obtained by the second input module 20. Specifically, it shows the number of errors in delayed recall for subjects whose underlying disease is AD and for subjects whose underlying disease is DLB.

[0044] Figure 6 shows an odds table created from the distribution of delayed recall scores for each disease group in the cognitive ability test (ADAS-Jcog). Simultaneously, the first input module 10 obtains the output value of the activation function (softmax function) of a deep learning differential model that takes brain images of each subject as input. In the estimation module 30, the softmax value obtained from the first input module 10 is used as the first evaluation index X1, and the disease odds score, which takes the clinical information (cognitive ability test) result score of the same subject (patient) as input, is used as the second evaluation index X2. The first evaluation function f1a multiplies these indices X1 and X2 to obtain an evaluation value S indicating that the causative disease is AD.

[0045] Figure 7 shows the first evaluation index X1, the second evaluation index X2, and the evaluation value S obtained as described above. If the evaluation value S is equal to or greater than the cutoff value (threshold, 50% in this example), it is possible to determine whether or not it is a brain disease, in this example, AD. In this example, the discrimination accuracy, which was 72% with the image alone, increased to 83% by using the first evaluation index X1 and the second evaluation index X2 in combination to differentiate diseases.

[0046] The odds table can be constructed by taking the ratio of the proportion of cognitive ability scores of subjects in the disease group (used as training data) to the total, and similarly, the proportion of those scores in the control group. In this example, Alzheimer's disease and Lewy body dementia were evaluated using MRI images and ADAS-Jcog delayed recall scores, but this is not the only method.

[0047] Figure 8 shows the first evaluation index X1, the second evaluation index X2, and the evaluation value py obtained by logistic regression as described above. If the evaluation value py is greater than or equal to the cutoff value (threshold, 50% in this example), it is possible to determine whether or not it is a brain disease, in this example, AD. In this example, the discrimination accuracy, which was 72% with the image alone, increased to 83% by using the first evaluation index X1 and the second evaluation index X2 in combination to differentiate diseases.

[0048] In this example, the input to the logistic regression model was evaluated using MRI images and the ADAS-Jcog delayed recall score to determine whether Alzheimer's disease or Lewy body dementia was present, but this is not limited to this.

[0049] Figure 9 shows the evaluation value py obtained by inputting the Z-scores of the degree of atrophy for three different brain regions (X1, X2, and X3) into a logistic regression along with the second evaluation index X4. If the evaluated evaluation value py is above the cutoff value (threshold, 50% in this example), it can be determined whether or not there is a brain disease, in this example, Alzheimer's disease (AD). In this example, the discrimination accuracy, which was 56% with the image alone, increased to 78% by using the first evaluation indices X1-X3 and the second evaluation index X4 in combination to differentiate diseases.

[0050] In this example, the input to the logistic regression model was evaluated using the Z-score and the ADAS-Jcog delayed recall score to determine Alzheimer's disease and Lewy body dementia, but this is not limited to these methods.

[0051] Figure 10 shows how to determine the negative / positive aspects of both image-based assessment and cognitive test scores for a given disease or its severity. By determining the negative / positive aspects of both image-based assessment and cognitive test scores for a given disease or its severity, the following four categories can be obtained. Category A: Image (Positive) and Cognitive Ability Test (Positive) Category B: Images (Negative) and Cognitive Tests (Positive) Category C: Image (Positive) and Cognitive Ability Test (Negative) Category D: Image (Negative) and Cognitive Ability Test (Negative)

[0052] There are methods for determining diseases using cutoffs for brain image evaluation indices. For example, a method is known in which the degree of reduced blood flow in a specific area of ​​the occipital lobe (CIScore) is evaluated for subjects with Alzheimer's disease and Lewy body dementia, and differentiation is performed using the cutoff value. In addition to setting cutoffs, a method may also be provided for determining attributes according to the magnitude of the brain image evaluation indices, separate from the cutoffs. Regarding the evaluation of brain diseases using cognitive ability test scores, it is known that healthy individuals, mild cognitive impairment, and Alzheimer's disease patients can be differentiated using cutoff values ​​for CDR or MMSE, for example. By using the evaluation values ​​obtained by the dementia evaluation system 1 and evaluation method described above, it becomes possible to further refine the categories D and D' in the category concept diagram 10 with greater precision. Furthermore, it becomes possible to categorize subjects using the evaluation of brain images alone, clinical information (e.g., cognitive ability test) result scores, and the type and progression stage of brain disease evaluated by integrating both, to determine subjects for drug efficacy evaluation, and to use this information for evaluation analysis for each category.

[0053] System 1 may include a module that outputs the estimation (evaluation) of the estimation module 30 described above and / or the process leading to the estimation, the first evaluation index X1, the second evaluation index X2, and other information xi to an output medium including a smartphone, PC, tablet, or paper. The output format may be text, graphics, images, encrypted information such as a QR code (registered trademark), or information indicating the access destination of the information.

[0054] Furthermore, System 1 may include a module that classifies subjects into predetermined categories based on the estimation and / or the process leading to the estimation of the estimation module 30, a first evaluation index X1, a second evaluation index X2, and other information xi. In drug discovery research and stratified medicine, it is known that patients belonging to a certain disease are classified into several subgroups using biomarkers, and appropriate treatment and evaluation are performed for each subgroup. The estimated evaluation of estimation module 30 can also be used as a stratification marker.

[0055] (Second Embodiment) In the second embodiment, the first evaluation function used was the formula (f1b) which converts the logid {log(p / (1-p))} into a probability using the sigmoid function. That is, the disease probability p output from the logistic regression model was used as the evaluation value. In the second embodiment, a configuration using ridge regression instead of logistic regression will be described. In ridge regression, the output evaluation value is a scalar, not a probability.

[0056] The estimation module 30 can calculate the estimated value y (hat) output as the evaluation value using equation (1) as the first evaluation function, where xi is the explanatory variable and wi is the weight coefficient of each explanatory variable xi. Note that x0 = 1 and w0 is the intercept.

[0057]

number

[0058] Let's designate the disease labels we want to separate as -1 and 1. For example, label -1 represents healthy individuals, and label 1 represents individuals with dementia. For unknown data xi, we can classify it based on whether y (hat) is greater than or less than 0.

[0059] The weight coefficient wi in equation (1) can be obtained by solving the optimization problem that minimizes the loss function F shown in equation (2). In equation (2), k=1, ..., n are the number of data samples, and yk are the measured values. β is a parameter that can be set in advance and determines the magnitude of the influence of the regularization term, which is expressed as the square of the L2 norm of wi.

[0060]

number

[0061] As described above, the estimation module 30 can estimate the state of brain damage, including dementia, of a subject based on evaluation values ​​obtained by a first evaluation function with a first evaluation index and a second evaluation index as explanatory variables. Here, the first evaluation function is expressed as a linear combination of explanatory variables, with weight coefficients corresponding to each explanatory variable, predicted by ridge regression using the training data, as shown in equation (1). Note that the explanatory variables may consist of only one of the first evaluation index or the second evaluation index.

[0062] Ridge regression can resolve the problem of multicollinearity, which occurs when there are many explanatory variables, for example, when the explanatory variables are correlated with each other, resulting in unstable estimations.

[0063] (Third embodiment) In the third embodiment, a system for evaluating the risk of degenerative brain disease is described, which evaluates each brain region (region of interest) of a subject's brain images (MR, SPECT, PET, etc.), and a system for evaluating the risk of brain disease targeting the entire brain based on the evaluation values ​​of characteristic regions of the subject. Note that MR images are also called MRI images. MRI images include, for example, T1-weighted images, T2-weighted images, diffusion-weighted images, flare images, diffusion tensor images, QSM images, pseudo-PET images, pseudo-SPECT images, etc.

[0064] Brain diseases include dementia (including AD, DLB, frontotemporal lobar degeneration (FTLD), normal pressure hydrocephalus (NPH), etc.), brain tumors, mental disorders (also called mental illnesses, including schizophrenia, epilepsy, mood disorders, addiction disorders, higher-order cognitive dysfunction, etc.), Parkinson's disease, Asperger's syndrome, attention deficit hyperactivity disorder (ADHD), sleep disorders, pediatric diseases, ischemic brain injury, mood disorders (including depression, etc.), etc. Furthermore, brain disorders include brain-related diseases such as dementia and multiple sclerosis, and diseases related to amyloid-beta, such as mild cognitive impairment (MCI), mild cognitive impairment due to Alzheimer's disease (MCI due to AD), prodromal AD, preclinical AD, Parkinson's disease, multiple sclerosis, insomnia, sleep disorders, cognitive decline, cognitive impairment, and neurodegenerative diseases related to amyloid-positive / negative conditions.

[0065] Figure 11 is a schematic diagram showing a first example of the system configuration of the third embodiment. The system (also referred to as the "evaluation system" or "dementia evaluation system") comprises a first input module 10 and an estimation module 40. The first input module 10 has the same configuration as in the first embodiment. The estimation module 40 comprises an evaluation value calculation unit 41 and an output unit 42. The first input module 10 can access necessary information by referring to the subject DB 60. The subject DB 60 may be the database 19 of the first embodiment. The estimation module 40 can access necessary information by referring to the healthy person DB 61 and the brain disease patient DB 62, respectively.

[0066] The first input module 10 comprises a first evaluation unit 11 and a second evaluation unit 12. The first evaluation unit 11 calculates a first evaluation index X1 by statistically evaluating medical images of each region (region of interest) of the subject's brain. The second evaluation unit 12 outputs the first evaluation index X1 for each region (region of interest) of the subject's brain using a pre-trained machine learning model that has been trained to output the first evaluation index X1 for each region (region of interest) of the subject's brain based on medical images. Note that either the first evaluation unit 11 or the second evaluation unit 12 may be used alone, or both may be used. The first input module 10 outputs the first evaluation index X1 for each region (region of interest) of the subject's brain to the estimation module 40. In the following, the first evaluation index X1 will be described as the Z-score value of the gray matter volume value of the region of interest in anatomical standard space, but the first evaluation index X1 is not limited to the Z-score.

[0067] The Z-score for a specific area of ​​the brain can be calculated for each pixel of that area using the following formula. Z-score = (Pixel value of brain region in subject - mean value of brain region in healthy individual) / (Standard deviation of brain region in healthy individual) The Z-score for a brain region can be calculated as the average of the positive Z-scores for each pixel in that region. The Z-score represents the degree to which the pixel values ​​of a subject's brain region deviate from those of a healthy brain region. In the case of MR images, a higher Z-score indicates greater atrophy compared to a healthy brain. Brain regions (regions of interest) include, but are not limited to, the diencephalon, superior parietal lobule, inferior parietal lobule, globus pallidus, cerebellum, paracentral lobule, hippocampus, parahippocampal gyrus, precuneus, lateral ventricles, amygdala, entorhinal cortex, and brainstem.

[0068] The first input module 10 can calculate Extent and Ratio for each region. Extent represents the ratio of the number of voxels with a Z score of 2 or higher within a region to the total number of voxels within that region. When the Z score is 2 or higher, it is more than twice the standard deviation from the mean pixel value, indicating a statistically significant difference. Ratio is the value obtained by dividing the average Z score within a region by the average Z score of the entire brain. The first input module 10 outputs the calculated Extent and Ratio to the estimation module 40.

[0069] The output unit 42 functions as an output unit and outputs display data for display on a display device (not shown). The display device may be integrated into the system or be an external device.

[0070] Figure 12 is a schematic diagram showing an example of the evaluation results for the hippocampal region. Figure 12 schematically shows an image of the subject's brain viewed from the front, and may differ from the actual image. In the figure, the patterned areas are the left and right hippocampi, and the color or pattern is changed according to the value of the first evaluation index (Z score) to display them in a map-like manner (evaluation index map). In the example in Figure 12, the Z score is 0.7, the Extent is 20%, and the Ratio is 1.5 times. Note that the Z score is calculated by treating each of the left and right hippocampal regions as a single region. Also, the numerical values ​​shown in Figure 12 are for convenience only and may differ from the actual values.

[0071] Figure 13 is a schematic diagram showing an example of the evaluation results for the middle temporal gyrus region. Figure 13 schematically shows a cross-sectional image of the subject's middle temporal gyrus and may differ from the actual image. In the figure, the patterned areas are the left and right middle temporal gyri, and are displayed in a map-like manner with changes in color or pattern according to the value of the first evaluation index (Z score). In the example in Figure 13, the Z score is 0.41, Extent is 0%, and Ratio is 0. The Z score is calculated by treating each of the left and right middle temporal gyrus regions as a single region. Also, the values ​​shown in Figure 13 are for convenience only and may differ from the actual values.

[0072] Other regions besides the hippocampus and middle temporal gyrus can also be displayed in a similar manner. By providing evaluation results like those shown in Figures 12 and 13 to physicians, they can use them as material for diagnosing various brain diseases, including dementia, in subjects.

[0073] The evaluation value calculation unit 41 functions as a calculation unit and calculates a region evaluation value (also called "region evaluation value") for each of the multiple regions of interest (regions of interest) of the subject's brain based on a first evaluation index for each of the multiple regions of interest and a weighting coefficient corresponding to each first evaluation index.

[0074] Figure 14 is a schematic diagram showing an example of evaluation values ​​for each part. As shown in Figure 14, let j be the index of the part and m be the number of parts. Each part can be represented by j=1, 2, ..., m. Let xdj be the evaluation index for each part j, and let wdj be the weight coefficient of the evaluation coefficient xdj. The evaluation value Ej for each part can be represented as E1=wd1·xd1, E2=wd2·xd2, ..., Em=wdm·xdm.

[0075] The output unit 42 can output evaluation values ​​for the brain regions calculated by the evaluation value calculation unit 41. The output unit 42 may also output display data for displaying the evaluation values ​​for each of the multiple brain regions of the subject in a predetermined order (for example, in descending order of evaluation value). When the evaluation value calculation unit 41 receives the selection of a required subject from among multiple subjects, it can calculate the evaluation values ​​for each of the multiple brain regions of the selected subject.

[0076] Figure 15 is a schematic diagram showing the first example of a region comparison screen. The region comparison screen has an area that displays a list of subjects and an area that displays the evaluation value of each region (ROI) for each subject. In the subject list, the IDs and names of multiple subjects are displayed in a list, and the target subject can be selected. For example, a doctor can select the subject to be diagnosed from the list. In the example in Figure 15, subject OOOX, enclosed by a dashed line, is selected. The evaluation values ​​of the subjects' regions can be displayed in descending order of evaluation value. In the example in Figure 15, the evaluation value of the diencephalon is 0.5, which is the largest value, so it is displayed at the top. Below that, the ROI and the evaluation value of the region are displayed in descending order of evaluation value. This can be used to determine which part of the brain is contributing when a subject is suffering from a particular brain disease. Note that the numerical values ​​exemplified in Figure 15 are for convenience only and may differ from actual values.

[0077] In the screen shown in Figure 15, by selecting subjects as appropriate, it is possible to dynamically switch between multiple subjects and display evaluation values ​​for each body part.

[0078] Figure 16 is a schematic diagram showing a second example of a body part comparison screen. The difference from the first example illustrated in Figure 15 is that the evaluation values ​​for each body part of healthy individuals and evaluation values ​​for each body part of patients with a specific brain disease are displayed. The estimation module 40 functions as a healthy individual evaluation value acquisition unit and can acquire evaluation values ​​for multiple body parts of healthy individuals (for example, the average of evaluation values ​​of many healthy individuals) from the healthy individual DB 61. The estimation module 40 also functions as a brain disease patient evaluation value acquisition unit and can acquire evaluation values ​​for multiple body parts of patients with a specific brain disease (for example, the average of evaluation values ​​of many patients with a specific brain disease) from the brain disease patient DB 62. The estimation module 40 can selectively acquire evaluation values ​​for multiple body parts of specific brain disease patients from among multiple types of brain diseases.

[0079] Figure 16 shows a configuration that displays evaluation values ​​for each part of both healthy individuals and patients with brain disease, but it is also possible to display evaluation values ​​for each part of either healthy individuals or patients with brain disease. In other words, the output unit 42 can output display data for displaying evaluation values ​​for multiple parts of subjects and healthy individuals in a comparative display manner. The output unit 42 can also output display data for displaying evaluation values ​​for multiple parts of subjects and patients with brain disease in a comparative display manner. Although not shown in the figure, in the comparison screen of parts exemplified in Figure 16, multiple types of brain diseases may be displayed in a list, and the required brain disease may be selected from the listed brain diseases. Each time a brain disease is selected, the evaluation values ​​for each part of patients with the selected brain disease may be displayed.

[0080] As shown in Figure 16, by displaying the evaluation values ​​for each part of the subject and the evaluation values ​​for each part of at least one of the healthy individuals and patients with brain disease in a comparative manner, it is possible to determine the state of the subject's brain disease compared to a healthy individual or to patients with brain disease.

[0081] Figure 17 is a schematic diagram showing a third example of a comparison screen of brain regions. The example in Figure 17 demonstrates that it is possible to estimate a subject's brain disease based on their evaluation values ​​for each brain region. In subjects with core symptoms of dementia, there is a suggestive characteristic in imaging findings that "medial temporal lobe atrophy is relatively milder in Lewy body dementia (DLB) compared to Alzheimer's disease (AD)." This suggestive characteristic can be judged based on the evaluation values ​​related to atrophy for each brain region of the subject. Here, the medial temporal lobe refers to the gray matter region including the parahippocampal gyrus and the hippocampus. As shown in Figure 17, in the case of subject A, the hippocampal atrophy evaluation is high, but compared to the evaluation values ​​of patients with brain diseases, subject A's evaluation value is small, and the atrophy can be judged to be mild. On the other hand, in the case of subject B, the evaluation values ​​of the hippocampus and parahippocampal gyrus are high, and it can be judged that the atrophy is advanced. In such cases, subject A can be estimated to have Lewy body dementia, and subject B can be estimated to have Alzheimer's disease.

[0082] Next, we will explain the assessment of disease risk for the entire brain of the subject, based on evaluation values ​​for each region of the subject's brain.

[0083] The estimation module 40 functions as an estimation unit and estimates the state of brain damage, including dementia, of the subject based on a whole-brain evaluation value obtained by a second evaluation function that uses a first evaluation index for each of several parts of the subject's brain as a variable. Specifically, the evaluation value calculation unit 41 can calculate a whole-brain evaluation value using the second evaluation function.

[0084] The second evaluation function can be expressed by equation (3).

[0085]

number

number

[0086] In equation (3), EA is the predicted value of the whole brain evaluation score, xdj is the evaluation index for region j, and wdj is the weight coefficient of the evaluation coefficient xdj. m is the number of regions, and ε is a constant for evaluating the error.

[0087] The weight coefficient wdj in equation (3) can be obtained by solving the optimization problem that minimizes the loss function L shown in equation (4). In equation (4), i=1, ..., n are the number of training data samples, and E is the measured value of the whole-brain evaluation score. α is a parameter that can be set in advance and determines the magnitude of the influence of the regularization term, which is expressed as the square of the L2 norm of wdj. That is, the second evaluation function expressed in equation (3) is expressed as a linear combination of the first evaluation index xdj, each of which is predicted by ridge regression using the training data, and the weight coefficient wdj is the coefficient of each of the first evaluation indices xdj.

[0088] Let's assign labels -1 and 1 to the classes we want to separate. For example, label -1 represents healthy individuals and label 1 represents individuals with dementia. Classification can be performed based on whether the whole-brain evaluation score EA is greater than or less than 0 for an unknown evaluation index xdj. If there are three or more classes to separate (for example, let's say the three classes are healthy individuals, brain disease B1, and brain disease B2), the brain disease can be classified by majority vote on all possible combinations of two classes. Specifically, if there are three combinations of healthy individuals and brain disease B1, brain disease B1 and brain disease B2, and brain disease B2 and healthy individuals, and brain disease B1 is classified twice, brain disease B2 is classified once, and healthy individuals are classified zero times, then the most frequently classified group is brain disease B1.

[0089] As mentioned above, the problem of multicollinearity can be solved by using a ridge regression model. This point will be explained below.

[0090] Figure 18 is a schematic diagram illustrating an example of the correlation of Z-scores between different regions of the left and right hemispheres of the brain. 'j' is an index representing a specific region. Regions of the right brain are represented by j=1 to 51, and regions of the left brain by j=52 to 102. Here, a region j in the right brain corresponds to a region (j+51) in the left brain. For example, if region j in the right brain is the hippocampus, then the region (j+51) in the left brain is also the hippocampus. In Figure 18, the dashed lines represent regions with a high correlation of Z-scores. That is, the same regions in both the left and right hemispheres tend to have a high correlation of Z-scores.

[0091] If a function based on a logistic regression model is used as the second evaluation function, the variables within the logistic regression model will correlate with other variables, leading to multicollinearity and unstable calculation of the estimated values. Therefore, by using a ridge regression model as shown in equation (3), a regularization term is added, which solves the multicollinearity problem. Specifically, when using a function based on a logistic regression model, many areas in the same part of the right and left hemispheres will have one side having a positive weight coefficient and the other side having a negative weight coefficient, making it impossible to calculate the whole-brain evaluation value with high accuracy. Ideally, all weight coefficients should be positive. By using a function based on a ridge regression model, the number of areas with negative weight coefficients can be significantly reduced, and for example, the whole-brain evaluation value can be obtained with an accuracy of about 82%.

[0092] Next, we will explain a method for further improving the estimation accuracy of whole-brain evaluation values ​​when using a ridge regression model.

[0093] Figure 19 is a schematic diagram illustrating an example of a method for aggregating the left and right hemispheres. As shown in Figure 19, the right hemisphere region is represented by j, and the same region in the left hemisphere as the right hemisphere is represented by (j+51). The first aggregation method is to unify the right hemisphere region j and the left hemisphere region (j+51) into a single region using the right hemisphere index j. If the number of regions in the right and left hemispheres is 51 each, and the total number of regions in the brain is 102, then the first aggregation method reduces the number of regions from 102 to 51. Ridge regression using the first aggregation method allowed us to achieve an estimation accuracy of approximately 96% for the whole-brain evaluation value. Note that the total number of regions in the brain is not limited to 102; other numbers may also be used.

[0094] The second aggregation method involves unifying the weight coefficients of right brain region j and left brain region (j+51) into a right brain region index. If the number of regions in the right and left brains is 51, and the total number of regions in the brain is 102, then the second aggregation method reduces the number of regions from 102 to 51. Ridge regression using the second aggregation method allowed us to achieve an estimation accuracy of approximately 91% for the whole brain evaluation value.

[0095] The third aggregation method involves unifying the weight coefficient of the right brain region j with the larger of the weight coefficient of the left brain region (j+51) to determine the right brain region index. If the number of regions in the right and left brains is 51, and the total number of regions in the brain is 102, then the third aggregation method reduces the number of regions from 102 to 51. Ridge regression using the third aggregation method allowed us to achieve an estimation accuracy of approximately 82% for the whole brain evaluation value.

[0096] Next, we will explain a specific example of the whole-brain evaluation value EA calculated by the evaluation value calculation unit 41.

[0097] Figure 20 is a schematic diagram showing a specific example of the whole-brain evaluation value. As mentioned above, the whole-brain evaluation value EA can be calculated using equation (3). When using the aggregation method described above, the number of regions j can be, for example, 51, and the number of evaluation values ​​for regions will also be 51. In Figure 20, for simplicity, the meaning of the whole-brain evaluation value is explained on a two-dimensional plane by setting the number of regions to 2. As shown in Figure 20, the region evaluation values ​​are E1 and E2. The distance from the class division line (plane) that divides the classes (healthy individuals, patients with brain disease) on the two-dimensional plane becomes the whole-brain evaluation value. For example, for subject S1, -d1 represents the whole-brain evaluation value, and for subject S2, d2 represents the whole-brain evaluation value. Here, the healthy side is assigned a negative sign, and the patients with brain disease side is assigned a positive sign. By using the subject's whole-brain evaluation value, it is possible to visually represent whether the subject is closer to a healthy individual or a patient with brain disease, as shown below.

[0098] Figure 21 is a schematic diagram showing the first example of a whole-brain assessment screen. The whole-brain assessment screen has an area for displaying a list of subjects and an area for displaying the subjects' whole-brain assessment values. In the subject list, the IDs and names of multiple subjects are displayed in a list, and the target subject can be selected. For example, a doctor can select the subject to be diagnosed from the list. In the example in Figure 21, subject OOOO, enclosed by a dashed line, has been selected.

[0099] The estimation module 40 functions as a whole-brain evaluation value acquisition unit, and can acquire whole-brain evaluation values ​​of healthy individuals from the healthy individual DB 61 and whole-brain evaluation values ​​related to the required brain disease from the brain disease patient DB 62. The subject's whole-brain evaluation values ​​can be displayed in a way that allows comparison with the whole-brain evaluation values ​​of healthy individuals and brain disease patients (for example, the average value of whole-brain evaluation values). In the example in Figure 21, one end of the horizontal bar graph represents healthy individuals and the other end represents brain disease patients, and the subject is represented by its position on the horizontal bar graph. This makes it possible to visually represent whether the subject is closer to a healthy individual or a brain disease patient.

[0100] In the screen shown in Figure 21, by appropriately selecting subjects, it is possible to dynamically switch between multiple subjects and display whole-brain evaluation values. Although not shown in the figure, in the whole-brain evaluation screen exemplified in Figure 21, multiple types of brain diseases may be displayed in a list, and the required brain disease may be selected from the displayed list. Each time a brain disease is selected, a horizontal bar graph showing the number of brain disease patients affected by the selected brain disease may be displayed on the other end.

[0101] Figure 22 is a schematic diagram showing a second example of a whole-brain assessment screen. As shown in Figure 22, the whole-brain assessment values ​​of the subject, as well as those of healthy individuals and patients with brain diseases, can be represented on a radar chart. The axes of multiple brain diseases are arranged in a regular polygonal shape from the center. In the example in Figure 22, three brain diseases, brain diseases 1, 2, and 3, are represented by an equilateral triangular radar chart. The position of healthy individuals on the radar chart is shown with a dashed line, and the position of the subject is shown with a solid line. This allows for a visual representation of whether the subject is closer to a healthy individual or a patient with a brain disease for each brain disease.

[0102] In the first example of the system illustrated in Figure 11, a first evaluation index X1 was used, but a second evaluation index X2 may also be used in addition to the first evaluation index.

[0103] Figure 23 is a schematic diagram showing a third example of a whole-brain assessment screen. As shown in Figure 23, the likelihood of each brain disease (referred to as brain diseases 1, 2, 3, and 4 in the example in Figure 23) based on the subject's whole-brain assessment value is visually represented by the size of the circles, like a bubble chart. For example, for each brain disease, the closer the subject's whole-brain assessment value is to the average whole-brain assessment value of healthy individuals, the smaller the circle becomes. Conversely, the closer the subject's whole-brain assessment value is to the average whole-brain assessment value of patients with brain diseases, the larger the circle becomes. In the example in Figure 23, it can be seen that subject OOOO is highly likely to be a patient with brain disease 1. It can also be seen that subject OOOO may have, or cannot rule out, brain diseases 2 and 3. Furthermore, it can be seen that subject OOOO's brain disease 4 is, for example, at a healthy level. Note that the types of brain diseases are not limited to the four shown in Figure 23.

[0104] Figure 24 is a schematic diagram showing a fourth example of a whole-brain assessment screen. In Figure 24, the symbols A to F in each cell can represent the type of brain disease or the brain region. The numbers in the table can represent the strength of the association with each brain disease or region. The numbers may also be the whole-brain assessment values ​​of the subject. Alternatively, as shown in Figure 24, the strength of the association with each brain disease or region may be represented by patterns or colors attached to each cell. In the example in Figure 24, it can be seen that subject OOOO has a strong association with brain disease (or region) C.

[0105] Figure 25 is a schematic diagram showing the fifth example of a whole-brain assessment screen. As shown in Figure 25, the likelihood of each brain disease (brain disease 1 and 2 in the example in Figure 23) based on the subject's whole-brain assessment value is visually represented in a matrix chart. The matrix has two axes and is divided into four regions S1 to S4. Region S1 indicates a high probability of brain disease 2 and a low probability of brain disease 1 (i.e., it can be determined that the subject has brain disease 2). Region S2 indicates a high probability of brain disease 2 and a high probability of brain disease 1 (i.e., it can be determined that the subject has both brain disease 1 and 2). Region S3 indicates a low probability of brain disease 2 and a low probability of brain disease 1 (i.e., it can be determined that the subject does not have either brain disease 1 or 2). Region S4 indicates a low probability of brain disease 2 and a high probability of brain disease 1 (i.e., it can be determined that the subject has brain disease 1). In the example in Figure 25, subject OOOO can be determined to have both brain disease 1 and 2.

[0106] Figure 26 is a schematic diagram showing a second example of the system configuration of the third embodiment. The second example differs from the first example in that it includes a second input module 20. The second input module 20 is the same as in the first embodiment shown in Figure 1, so its description is omitted. The estimation module 40 can estimate the state of brain damage, including dementia, of a subject based on a whole-brain evaluation value obtained by a second evaluation function in which the first evaluation index and the second evaluation index X2 for each of several parts of the subject's brain are variables. Specifically, the evaluation value calculation unit 41 can calculate the whole-brain evaluation value using the second evaluation function. The second evaluation function can be expressed by the above-described equation (3), where the evaluation index xdj includes the first evaluation index and the second evaluation index X2.

[0107] Figures 27A and 27B are schematic diagrams illustrating examples of ROC curves. Figure 27A shows the ROC (Receiver Operating Characteristic) curve when AD and DLB are differentiated using only the first evaluation index (Z-scores for 51 sites), while Figure 27B shows the ROC curve when differentiation is performed using the second evaluation index (delayed recall score in the ADASCog cognitive test) in addition to the first evaluation index. In Figures 27A and 27B, the vertical axis represents the true positive rate (TPR), and the horizontal axis represents the false positive rate (FRP). The ROC curve plots the true positive rate and false positive rate for each cutoff point that distinguishes between normal and abnormal in AD diagnosis and DLB diagnosis. The area under the plotted graph is called the AUC (Area Under the Curve), and the AUC can take values ​​from 0 to 1. A value closer to 1 indicates higher differentiation accuracy. In Figure 27A, the AUC is 0.62, while in Figure 27B, the AUC is 0.74, indicating that Figure 27B has a higher AUC. In the example in Figure 27, it can be seen that the accuracy of differentiation improves by adding a delayed recall score that assesses the decline in long-term memory (which is stronger in AD cases).

[0108] Next, the operation of the system according to the third embodiment will be described.

[0109] Figure 28 is a flowchart illustrating the procedure for the first example of outputting evaluation results for each part of the brain. The system acquires image data of the subject's brain (S11) and sets the subject's regions of interest (parts) (S12). The system calculates an evaluation index (e.g., Z-score) for each region of interest (S13) and calculates Extent and Ratio for each region of interest (S14).

[0110] The system generates an evaluation index map for each area of ​​interest based on the calculated evaluation index (S15), and outputs the subject's evaluation index map, evaluation index, Extent, and Ratio (S16). The system determines whether there are other areas of interest (S17), and if there are other areas of interest (YES in S17), it continues processing from step S12 onwards.

[0111] If there are no other areas of interest (NO in S17), the system determines whether there are other subjects (S18). If there are other subjects (YES in S18), the system continues processing from step S11 onwards; if there are no other subjects (NO in S18), the process ends.

[0112] Figure 29 is a flowchart showing the procedure for a second example of outputting evaluation results for each part of the brain. The system acquires image data of the subject's brain (S31) and sets the subject's regions of interest (parts) (S32). The system calculates an evaluation index (e.g., Z score) for each region of interest (S33) and calculates a region evaluation value Ej for each region of interest of the subject (S34). The region evaluation value Ej can be calculated as Ej = wdj·xdj, where xdj is the evaluation index (e.g., Z score) for region j and wdj is the weighting coefficient.

[0113] The system obtains region evaluation values ​​for each area of ​​interest of healthy individuals from the healthy individuals DB61 (S35), and obtains region evaluation values ​​for each area of ​​interest of patients with brain diseases from the brain disease patient DB62 (S36). The system determines whether or not to display the data in order of region evaluation value (for example, in descending order) (S37), and if it decides to display the data in order of region evaluation value (YES in S37), it sorts the data by the region evaluation value of the subjects (S38), and then performs the process described in step S39 below.

[0114] If the system does not display the results in order of the body part evaluation values ​​(NO in S37), it determines whether to display the subject's body part evaluation values ​​in comparison to those of healthy individuals and patients with brain disease (S39). If it chooses to display them in comparison to healthy individuals and patients with brain disease (YES in S39), the system outputs the body part evaluation values ​​for the subject, healthy individuals, and patients with brain disease (S40), and then performs the process described in step S42 below.

[0115] If the data is not displayed in comparison to healthy individuals or patients with brain disease (NO in S39), the system outputs the subject's region evaluation values ​​(S41) and determines whether there are other subjects (S42). If there are other subjects (YES in S42), the system continues processing from step S31 onwards; if there are no other subjects (NO in S42), the process ends.

[0116] Figure 30 is a flowchart showing the procedure for outputting whole-brain evaluation results. The system acquires image data of the subject's brain (S51) and sets the subject's regions of interest (regions) (S52). The system calculates an evaluation index (e.g., Z-score) for each region of interest (S53) and calculates a region evaluation value Ej for each region of interest of the subject (S54). The region evaluation value Ej can be calculated as Ej = wdj·xdj, where xdj is the evaluation index (e.g., Z-score) for region j and wdj is the weighting coefficient.

[0117] The system calculates the whole-brain evaluation score EA (S55). The whole-brain evaluation score EA can be calculated using equation (3). The system obtains the whole-brain evaluation scores of healthy individuals from healthy individuals DB61 (S56) and the whole-brain evaluation scores of patients with brain diseases from patients with brain diseases DB62 (S57). The system calculates the position of the subject's whole-brain evaluation score within the range of whole-brain evaluation scores for healthy individuals and patients with brain diseases, respectively (S58). For example, if the average whole-brain evaluation score of healthy individuals is a, the average whole-brain evaluation score of patients with brain diseases is b, and the subject's whole-brain evaluation score is c, then the position can be determined by where the value c lies within the range from a to b.

[0118] The system outputs the subject's whole-brain evaluation score in a display format that allows comparison with the whole-brain evaluation scores of healthy individuals and patients with brain disease, respectively (see, for example, Figure 21) (S59). The system determines whether there are other subjects (S60). If there are other subjects (YES in S60), the system continues processing from step S51 onwards; if there are no other subjects (NO in S60), the process ends.

[0119] In the third embodiment described above, the first evaluation index X1 was configured to use the Z-score value of the gray matter volume of the region of interest on an anatomical standard space. However, the first evaluation index is not limited to the Z-score. For example, physical quantities such as blood flow in the region of interest or the accumulation of malignant proteins (e.g., amyloid-beta, tau protein, etc.) in the region of interest may be used. When using such physical quantities, normalization may be performed using the proportion of the physical quantity within the region of interest that exceeds a predetermined threshold (for example, the proportion of voxels exceeding the threshold to the total number of voxels in the region of interest). This makes it possible to compare evaluation indices between regions of interest, regardless of the size of the region of interest.

[0120] Furthermore, at least one of SUVR and BR may be used as the first evaluation index X1. In this case, in equation (3), for example, w SUVR ·SUVR, w BR • Add at least one of the terms in BR. Here, w SUVR w is the weighting coefficient of the evaluation index SUVR, BR This is the weighting coefficient for the evaluation index BR. Furthermore, when using the evaluation index SUVR or the evaluation index BR, if these index values ​​are standardized from the distribution of the healthy control database, it is possible to evaluate the importance of each subject being judged as having a disease, similar to when each body part is evaluated using only the Z score. For example, the standardized value of SUVR is SUVR. Z , SUVR Z It can be calculated using the formula: ={(Average SUVR of healthy individuals - SUVR of the subject) / Standard deviation of SUVR of healthy individuals}. [Explanation of Symbols]

[0121] 1. Dementia Assessment System (System) 10. First input module 11. First evaluation unit 12. Second evaluation unit 20 Second input module 21, 22 units 30 Estimated Modules 31 Odds Determination Unit 32 Probability Determination Units 35 Estimated Units of Infection 36. Clinical Evaluation Unit (Clinical Evaluation Function, Function) 40 Estimated Modules 41 Evaluation Value Calculation Unit 42 Output Units 60 Subject Database 61 Healthy Individual Database 62. Database of patients with brain diseases

Claims

1. A first input module configured to acquire a first evaluation index based on data relating to the physical state of the subject's brain, A second input module configured to acquire a second evaluation index based on data relating to the brain function of the subject, the second evaluation index being the odds x2 of the first causative disease, A system comprising: an estimation module configured to estimate the state of brain damage, including dementia, of the subject based on an evaluation value obtained by a first evaluation function with the first evaluation index and the second evaluation index as variables.

2. In claim 1, The system includes at least one of the following: a first evaluation unit that obtains a first evaluation index by statistically evaluating a first type of medical image of at least a portion of a region of interest in the subject's brain; and a second evaluation unit that obtains a first evaluation index by evaluating the subject's medical image using a first model that has been machine-trained to evaluate a first disease based on the first type of medical image.

3. In claim 1 or 2, The system includes a configuration in which the second input module acquires an evaluation of clinical information, including a cognitive ability test, as the second evaluation index.

4. In any of claims 1 to 3, The aforementioned subjects were included in the group that ingested at least one of the following: pharmaceuticals, food and beverages, or supplements. The aforementioned estimation module is a system that includes a function to evaluate the effect of ingested substances on dementia.

5. In any of claims 1 to 3, The estimation module is a system that includes a function for estimating the disease state of a first causative disease.

6. In claim 5, The first input module includes a configuration for obtaining the first evaluation index for differentiating the first causative disease, The second input module includes a configuration for obtaining a second evaluation index for differentiating the first causative disease, The estimation module is a system that includes a first evaluation function for estimating the disease state of the first causative disease.

7. In claim 6, The system includes a configuration in which the first input module acquires at least one of the following values ​​as the first evaluation index. a: The output softmax value of the activation function when estimating the first causative disease using a deep learning differential diagnosis model that takes brain images as input. b: The softmax output of the activation function when estimating the first causative disease using a deep learning differential diagnosis model, based on the region of interest obtained by statistical processing of brain images. c: Volume value or blood flow of the region of interest obtained by statistical processing of brain images. d: Z-score value for volume or blood flow assessment of region of interest using statistical processing of brain images. e: Volume value or blood flow of the region of interest when estimating the first causative disease using a deep learning differential diagnosis model that takes brain images as input. f: Z-score value of the volume or blood flow evaluation of the region of interest when estimating the first causative disease using a deep learning differential diagnosis model that takes brain images as input.

8. In claim 6 or 7, The system includes a configuration in which the second input module acquires the results of a cognitive ability test suitable for differentiating the first causative disease as the second evaluation index.

9. In any of claims 6 to 8, The estimation module is a system that includes a first evaluation function in which a first causative disease is estimated when the evaluation value exceeds a first threshold.

10. In any of claims 6 to 9, The system comprises a configuration in which the first evaluation index is the odds x1 of the first causative disease, and the estimation module is configured to determine the evaluation value s by the following first evaluation function. s = x¹ × x²

11. In any of claims 6 to 9, The estimation module is a system that includes a configuration in which, when the first and second evaluation indices are xi, the probability of contracting the causative disease y* is calculated as the evaluation value using the following first evaluation function. [Math 1] The set wyi represents the weight coefficients for each evaluation index of the respective causative disease, where i is an integer.

12. In any of claims 1 to 11, The estimation module is a system that includes a configuration that provides information as a stratification marker.

13. In any of claims 1 to 12, A system including a module that outputs the estimation and / or the process leading to the estimation, the first evaluation index, the second evaluation index, and other information to an output medium including a smartphone, PC, tablet, or paper.

14. In any of claims 1 to 13, A system including a module for classifying subjects into predetermined categories based on the estimation and / or the process leading to the estimation by the estimation module, the first evaluation index, the second evaluation index, and other information.

15. A method for controlling a system, The system includes a first input module configured to acquire a first evaluation index based on data relating to the physical state of the subject's brain, A second input module configured to acquire a second evaluation index based on data relating to the brain function of the subject, the second evaluation index being the odds x2 of the first causative disease, It includes an estimation module configured to estimate the dementia state of the subject, The control method described above is The estimation module obtains the first evaluation index and the second evaluation index via the first input module and the second input module, The method involves estimating the state of brain damage, including dementia, of the subject based on evaluation values ​​obtained by a first evaluation function with the first and second evaluation indices as variables. Control method.

16. To obtain a first evaluation index based on data relating to the physical state of the subject's brain, Obtaining a second evaluation index based on data relating to the brain function of the subject, wherein the second evaluation index is the odds x2 of the first causative disease, A computer program having instructions to perform the following actions on a computer: estimate the state of brain damage, including dementia, of the subject based on an evaluation value obtained by a first evaluation function with the first evaluation index and the second evaluation index as variables.

Images