WO2022054711A1 - Computer program, information processing device, terminal device, information processing method, learned model generation method, and image output device - Google Patents

Abstract

Provided are a computer program, information processing device, terminal device, information processing method, learned model generation method and image output device which make it possible to provide the same brain diagnosis information as does a PET image without performing a PET exam.　This computer program causes a computer to execute processing for obtaining one or more MRI images about the brain of an imaging subject and captured at one or more points in time, and when an MRI image about the brain is inputted, inputting the obtained one or more MRI images to a learned model for outputting pseudo-PET images about the brain, and outputting the pseudo-PET images about the brain of the imaging subject at the one or more points in time.

Description

Computer program, information processing device, terminal device, information processing method, trained model generation method and image output device

The present invention relates to a computer program, an information processing device, a terminal device, an information processing method, a trained model generation method, and an image output device.

In recent years, the number of dementia patients and dementia reserves (mild cognitive impairment) is increasing due to the aging of the population. The four major diseases that cause dementia include Alzheimer-type dementia (AD), Lewy body dementia (DLB), cerebrovascular dementia, frontotemporal lobar degeneration (FTLD), and other frequent occurrences. Various neurological symptoms including dementia also appear in sclerosis (MS) and the like. The cause of Alzheimer's disease has not yet been elucidated, but as the condition progresses, peculiar lesions are found in the brain. For example, in Alzheimer-type dementia and frontotemporal lobar degeneration, it is known that amyloid β causes amyloid plaque deposition and accumulation of abnormal proteins mainly in tau on the outside of nerve cells. It has been shown that amyloid plaque deposition occurs in the earliest stages of the onset of Alzheimer's disease and begins long before clinical symptoms appear (eg, a dozen years ago). In addition, α-synuclein, a protein of unknown function, is expressed in nerve cells of the brain, which is considered to be the cause of neurodegenerative diseases such as Parkinson's disease.

Patent Document 1 discloses a device that injects a drug that binds to amyloid β in brain tissue to a subject and uses a PET (Positron Emission Tomography) image showing the concentration distribution of the drug on a cut section across the brain. ing.

International Publication No. WO2014 / 034724

However, taking a PET image using a drug requires ingesting a radioactive substance into the body, even if it is a small amount, and there is a risk of being exposed to radiation, which puts a burden on the patient. In addition, the drug may not be given to patients suffering from certain diseases such as kidney disease. Further, there is also a problem that the cost of PET examination is generally high.

The present invention has been made in view of such circumstances, and is a computer program, an information processing device, a terminal device, an information processing method, and a learned model that can provide brain diagnostic information equivalent to a PET image without performing a PET examination. It is an object of the present invention to provide a generation method and an image output device.

In the computer program according to the embodiment of the present invention, when one or a plurality of MRI images taken at one or a plurality of time points regarding the subject's brain are acquired and the MRI image regarding the brain is input to the computer, the brain. The acquired one or more MRI images are input to the trained model for outputting the pseudo PET image of the subject, and the process of outputting the pseudo PET image at the one or more time points regarding the brain of the subject is executed. ..

The information processing apparatus according to the embodiment of the present invention is the case where the acquisition unit for acquiring one or more MRI images captured at one or a plurality of time points regarding the subject's brain and the MRI image regarding the brain are input. An output unit that inputs the acquired one or more MRI images to the trained model that outputs a pseudo PET image about the brain and outputs the pseudo PET image at the one or a plurality of time points about the subject's brain. Be prepared.

The terminal device according to the embodiment of the present invention displays a pseudo PET image over one or more time points and at least one of the amyloid accumulation over one or more time points for the brain of a subject to whom a particular drug has been administered. do.

The information processing method according to the embodiment of the present invention acquires one or a plurality of MRI images captured at one or a plurality of time points regarding the subject's brain, and when the MRI image regarding the brain is input, the pseudo brain. The acquired one or more MRI images are input to the trained model for outputting the PET image, and the pseudo PET image at the one or more time points regarding the subject's brain is output.

The trained model generation method according to the embodiment of the present invention acquires MRI images and PET images captured at a first time point regarding the brain, and relates to the brain imaged at a second time point after the first time point. An MRI image and a PET image are acquired, and the MRI image and the PET image taken at the first time point and the second time point are used to obtain the MRI image and the PET image about the brain taken at the first time point. , And when an MRI image of the brain captured at the second time point is input, a trained model that outputs a pseudo PET image of the brain at the second time point is generated.

The image output device according to the embodiment of the present invention inputs an MRI image and a PET image taken at the first time point regarding the brain and an MRI image about the brain taken at the second time point after the first time point. In this case, a pseudo PET image relating to the brain at the second time point is output.

According to the present invention, it is possible to provide brain diagnostic information equivalent to a PET image without performing a PET examination.

It is a schematic diagram which shows an example of the structure of the information processing system of this embodiment. It is a schematic diagram which shows the 1st example of the PET image estimation method by the estimation part. It is a schematic diagram which shows the 2nd example of the PET image estimation method by the estimation part. It is a schematic diagram which shows the 1st example of the generation method of the trained model in a learning processing part. It is a schematic diagram which shows the 2nd example of the generation method of the trained model in a learning processing part. It is a schematic diagram which shows the 1st example of the diagnostic information to be displayed on the terminal apparatus of a doctor or the like. It is a schematic diagram which shows the 2nd example of the diagnostic information to be displayed on the terminal device of a doctor or the like. It is a schematic diagram which shows the 3rd example of the diagnostic information to be displayed on the terminal device of a doctor or the like. It is a schematic diagram which shows the 4th example of the diagnostic information to be displayed on the terminal device of a doctor or the like. It is a schematic diagram which shows the 1st example of the diagnostic information to be displayed on the terminal apparatus of a patient. It is a schematic diagram which shows the 2nd example of the diagnostic information to be displayed on the terminal apparatus of a patient. It is a schematic diagram which shows the 3rd example of the diagnostic information to be displayed on the terminal apparatus of a patient. It is a flowchart which shows an example of the procedure of the generation process of the brain diagnosis information. It is a flowchart which shows an example of the procedure of the trained model generation process.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a schematic diagram showing an example of the configuration of the information processing system of the present embodiment. The information processing system includes an information processing device 50. The information processing device 50 is connected to the communication network 1. A terminal device 10 used by a medical worker, a researcher, or the like, and a terminal device 20 used by a subject (patient) are connected to the communication network 1. As the terminal device 10, for example, a personal computer, a smartphone, a tablet, or the like can be used. As the terminal device 20, for example, a smartphone, a tablet, a personal computer, or the like can be used. An image DB 61, a PET device 71, and an MRI device 72 are connected to the information processing device 50. The information processing device 50, the image DB 61, the PET device 71, and the MRI device 72 are installed in a medical institution such as a hospital, for example. The information processing device 50 and the image DB 61 may be installed in a place other than a facility such as a medical institution.

The information processing device 50 includes a control unit 51, a communication unit 52, a storage unit 53, a diagnostic information generation unit 54, an estimation unit 55, a learning processing unit 56, and an interface unit 57 that control the entire information processing device 50. The control unit 51 can be configured by a CPU (Central Processing Unit), a ROM (Read Only Memory), a RAM (Random Access Memory), and the like. Each function of the information processing device 50 may be distributed to a plurality of devices. For example, the estimation unit 55 may be provided in the information processing device 50, and the learning processing unit 56 may be provided in another device.

The communication unit 52 is composed of a required communication module or the like, and provides a communication function between the terminal device 10 and the terminal device 20 via the communication network 1. The communication unit 52 can transmit diagnostic information to, for example, the terminal device 10 and the terminal device 20. The terminal device 10 and the terminal device 20 can display the received diagnostic information.

The storage unit 53 can be composed of a hard disk, a semiconductor memory, or the like, and can store required data such as data obtained as a result of processing in the information processing device 50.

The diagnostic information generation unit 54 generates diagnostic information of the brain, particularly diagnostic information based on the transition of the state of the brain over a plurality of time points. Details of the diagnostic information will be described later.

The interface unit 57 provides an interface function between the image DB 61, the PET (Positron Emission Tomography) device 71, and the MRI (Magnetic Resonance Imaging) device 72. The interface unit 57 can acquire a PET image obtained as a result of the PET inspection from the PET device 71. As used herein, PET images include both amyloid PET and tau PET. The interface unit 57 can acquire an MRI image obtained as a result of an MRI examination from the MRI apparatus 72. The MRI image is also referred to as an MR image.

The estimation unit 55 has a function of estimating the brain state, and includes an input data generation unit 551 and a model unit 552. The model unit 552 is composed of a semiconductor memory, a hard disk, or the like, and stores a model (learned model) generated by machine learning. As the trained model, for example, the trained model can be configured by a neural network, and for example, GAN (Generative Adversarial Network), U-net, or a combination thereof can be used. The input data generation unit 551 generates data to be input to the learning model when performing the brain state estimation process.

The learning processing unit 56 has a function of generating a trained model by machine learning, and includes a learning data generation unit 561, a model unit 562, and a parameter determination unit 563. The model unit 562 is composed of a semiconductor memory, a hard disk, or the like, and stores a model before machine learning. The trained model generated by performing machine learning in the learning processing unit 56 can be stored in the model unit 552 of the estimation unit 55. The model unit 562 may store a model in the middle of machine learning, a model for re-learning, and a model that has been trained. Further, the learning processing unit 56 is not an indispensable configuration, and may be configured to include the learning processing unit 56 in another learning device. The training data generation unit 561 generates training input data and teacher data when generating a trained model. The parameter determination unit 563 adjusts the parameters of the neural network (for example, weights, biases, filter parameters, etc.) when generating the trained model, and finally determines the parameters.

The learning processing unit 56 is, for example, a hardware such as a CPU (for example, a multi-processor having a plurality of processor cores mounted), a GPU (Graphics Processing Units), a DSP (Digital Signal Processors), an FPGA (Field-Programmable Gate Arrays), or the like. It can be configured by combining wear.

The image DB 61 can record various images (for example, MRI image, PET image, etc.) used in machine learning when generating a trained model. Further, the image DB 61 can record a pseudo PET image generated by the information processing apparatus 50. In addition, in this specification, a pseudo PET image is also literally referred to as a pseudo PET image.

The information processing device 50 includes patient information (subject information), MRI image, PET image (including presence / absence of PET image), and initial time, which are necessary for generating diagnostic information of the brain from a doctor or the like who operates the terminal device 10. Whether or not it is a medical examination, date of medical examination, adverse events (including side effects, such as nausea, vomiting, headache, slow pulse, angina, myocardial infarction, cerebrovascular attack, spasm attack, amyloid-related imaging abnormality (ARIA) , ARIA-E, ARIA-H, etc.), falls, etc.), biomarkers (FDG-PET, tau protein, blood test results, genetic information, etc.), cognitive test results, and the like can be obtained. The cognitive test is a test for checking the cognitive function of the brain. The content of the cognitive test includes, for example, calculations such as addition and subtraction, Stroop, N-Back, and word prescription, but is not limited thereto. Specific examples of the cognitive test include, for example, a test for "forward" and "reverse", a test for "Stroop", a test for "addition" and "subtraction", and a test for N-Back (for example, 1-Back). , Can be mentioned as a test for "immediate reproduction" (word recall). The cognitive ability test is not limited to these, and a test or test for measuring the state of brain health (including the state of cognitive function and the presence and degree of brain disease and mental illness) or other similar type of test or test is performed. It may be adopted. The above information can be displayed on either the terminal device 10 used by a doctor or the like or the terminal device 20 used by a subject (patient).

Next, the details of the method of generating the brain diagnostic information by the information processing apparatus 50 will be described. The brain diagnostic information is diagnostic information regarding a disease related to the brain, amyloid β, tau protein, or a disease related to α-synuclein (also referred to as α-synuclein aggregate). For example, diseases related to the brain include dementia and multiple sclerosis. Diseases related to amyloid β or tau protein include, for example, mild cognitive impairment (MCI: Mild cognitive impairment), mild cognitive impairment due to Alzheimer's disease (MCIdue to AD), prodromal AD, and Alzheimer's disease. Presymptomatic stage / preclinical AD, Parkinson's disease, multiple sclerosis, insomnia, sleep disorders, cognitive decline, cognitive impairment, amyloid positive / negative disorders, motor disorders, motor disorders, etc. Including neurodegenerative diseases. Diseases associated with alpha-synuclein include, for example, Alzheimer's disease (amyloid plaque), dementia (Lewy bodies), synucleinopathy (protein aggregates in neurons, nerve fibers, glial cells), multiple system atrophy, Parkinson's disease. (Lewy bodies), including motor dysfunction. Specifically, based on each MRI image obtained at the time of one or a plurality of MRI examinations, the PET image is estimated by generating a pseudo PET image at each MRI examination time point, and the diagnostic information of the brain is obtained. offer. The MRI examination time point is the MRI examination time point which is continuously performed after the first MRI examination based on the instruction of the doctor.

The MRI image is obtained by performing a predetermined calculation on the MRI signal as well as the MRI image (for example, T1 weighted image, T2 weighted image, diffusion weighted image, flare image, diffusion tensor image) obtained by the MRI apparatus 72. It also includes processed images (eg, QSM images) and the like. The term "pseudo-PET image" is used to distinguish it from an actual PET image, but in the present specification, it means not only an image similar to a PET image but also numerical information that can be estimated from the image. .. That is, "pseudo" means that not only the image itself but also information (numerical information, etc.) obtained from the image is included. For example, the pseudo PET image may include an image similar to an amyloid PET image, a tau PET image, or an α-synuclein aggregate image. Further, the pseudo PET image may include a centiroid value. In amyloid PET, the value of composite SUVR differs depending on the PET drug even if the accumulation of the same amyloid is the same, so that this is standardized by the value that does not depend on the drug. For example, the degree of accumulation of amyloid in the brain can be expressed by a numerical value from 0 to 100, where the average of healthy young people is 0 and the average of reliable AD is 100. Sentiloid can be estimated directly from the MRI image, or can be estimated via the distribution information of amyloid β estimated from the MRI image. That is, the pseudo PET image also includes, for example, numerical information obtained by quantifying an amyloid PET image with an index called SUVR and correcting it by a standardization method called a centiroid method. Further, the pseudo PET image includes distribution information (for example, UVVR value for each voxel) that visualizes the distribution of amyloid β, tau protein or α-synuclein in the brain, as in the PET image. The SUVR (Standardized Uptake Value Ratio) generally includes a SUVR (voxelUVR) for each boxel, a SUVR (regional SUVR) for each site, a SUVR (composite UVR) within a region of interest, and the like, but in the present specification. , Any SUVR may be used. That is, in the present specification, the description of SUVR shall include any of the above-mentioned SUVRs. For example, composite SUVR is the sum of SUVs (Standardized Uptake Value: amyloid β protein accumulation) of four sites of cerebellar gray matter (prefrontal cortex, anterior and posterior cingulate cortex, parietal lobe, and lateral temporal lobe) as regions of interest. Can be determined by dividing by the SUV of a particular reference region (eg, cerebellum, etc.). The degree of accumulation (SUV) of amyloid β, tau protein or α-synuclein can be determined, for example, by counting the number of voxels in which the value indicating the brightness of each voxel constituting the required site is equal to or higher than a predetermined threshold value. Accumulation (for example, OO%) can be calculated from the ratio of the count value to the total number of voxels of the required part. The SUVR for each voxel is a value obtained by dividing the SUV of a certain voxel by the average value of the SUV in the reference region (for example, the cerebellum). The SUVR for each site is a value obtained by dividing the SUV of a certain site by the average value of the SUV in the reference region (for example, the cerebellum).

FIG. 2 is a schematic diagram showing a first example of the PET image estimation method by the estimation unit 55. The model unit 552 is a trained model, and the model is composed of a neural network (for example, CNN: Convolutional Neural Network), and GAN (Generative Adversarial Network), U-net, or a combination thereof can be used. The model unit 552 includes an input layer, an intermediate layer, and an output layer, and neural network parameters (weights, biases, filter parameters, etc.) are determined by machine learning. The input data generation unit 551 inputs the MRI image acquired from the MRI apparatus 72 via the interface unit 57 to the model unit 552 as input data.

As shown in FIG. 2, when the MRI image at the time point t1 is input to the model unit 552, the model unit 552 outputs a pseudo PET image at the time point t1. Similarly, when the MRI image at the time point t2 is input to the model unit 552, the model unit 552 outputs a pseudo PET image at the time point t2. Hereinafter, similarly, when the MRI image at the time point tN is input to the model unit 552, the model unit 552 outputs the pseudo PET image at the time point tN. Time points t1, t2, t3, t4, ..., TN can be the time points of the MRI examination of the subject (patient). The time interval between adjacent time points is the time interval of a normal MRI examination, and can be, for example, 6 months or 1 year. In FIG. 2, it is expressed that the MRI images at the time points t1, t2, t3, t4, ..., TN are simultaneously input to the model unit 552, but it is not necessary and the MRI image is not necessary. The MRI image may be input to the model unit 552 each time.

The pseudo PET image is an image that estimates the actual PET image. PET images are images of the distribution of amyloid β, tau protein or α-synuclein in the brain. Amyloid β and tau protein are mild cognitive impairment (MCI: Mild cognitive impairment), mild cognitive impairment due to Alzheimer's disease (MCIdue to AD), prodromal AD, pre-symptomatic stage of Alzheimer's disease / preclinical AD (preclinical AD). Abnormalities found in the brain of patients such as preclinical AD), Parkinson's disease, multiple sclerosis, insomnia, sleep disorders, cognitive decline, cognitive impairment, and neurodegenerative diseases such as amyloid positive / negative disorders. It is a protein. In the PET test performed for the purpose of confirming the accumulation of amyloid β, tau protein or α-synuclein, the test drug is administered to the patient by intravenous injection, and the drug specifically adsorbs amyloid β, tau protein or α-synuclein. This makes it possible to observe the distribution of amyloid β, tau protein or α-synuclein in the brain.

The diagnostic information generation unit 54 generates diagnostic information based on the pseudo PET images at the time points t1, t2, t3, t4, ..., TN output by the model unit 552. Since pseudo-PET images at multiple time points are obtained, not only the state of the brain at a specific time point (specifically, the distribution state of amyloid β, tau protein or α-synuclein) but also the state of the brain over multiple time points. Diagnostic information based on changes in state can be obtained. This makes it possible to provide brain diagnostic information equivalent to a PET image without performing a PET examination. The details of the diagnostic information will be described later.

FIG. 3 is a schematic diagram showing a second example of the PET image estimation method by the estimation unit 55. The difference from the first example shown in FIG. 2 is that, in addition to the MRI image, the PET image at the time point t1 is input to the model unit 552 as input data. The PET image at time point t1 is obtained by a PET examination performed at the same timing as the MRI examination at time point t1. The time point t1 of the MRI examination and the time point t1 of the PET examination do not necessarily have to be the same timing, and there may be a time difference of about several days, about several weeks, or about several months. Further, in the example of FIG. 3, the PET image is at the time point t1, but is not limited to the time point t1, and may be a time point t2, a time point t3, or the like. This is because the PET examination is not always performed at the time of the first MRI examination, and the PET examination may be performed at the second or third MRI examination depending on the patient's symptoms.

The model unit 552 includes MRI and PET images captured at the first time point (time point t1 in the example of the figure) and the second time point (time point t2, t3, t4 in the example of the figure) after the first time point with respect to the brain. , ..., When an MRI image of the brain imaged at tN) is input, in the example of the second time point diagram, a pseudo PET image of the brain at time points t2, t3, t4, ..., TN) can be output. can. For example, when the MRI image and PET image obtained by the MRI examination and the PET examination performed at the first examination and the MRI image obtained by the second and subsequent MRI examinations are input to the model unit 552, the model unit 552 It is possible to output a pseudo PET image related to the brain at the time of the second and subsequent MRI examinations.

Compared to the case where the MRI image at the time points t1 to tN is simply input to the model unit 552 as shown in FIG. 2, the information of the actual PET image at the time point t1 is also used, so that the subject is in the initial stage. Considering the distribution state of amyloid β, tau protein or α-synuclein in the brain, it is expected that the accuracy of the pseudo PET image output by the model unit 552 will be improved.

Next, the method of generating the trained model will be explained.

FIG. 4 is a schematic diagram showing a first example of a method of generating a trained model in the learning processing unit 56. The method of generating the trained model shown in FIG. 4 corresponds to the method of generating the model unit 552 of FIG. As shown in FIG. 4, as learning input data, MRI at time point nt regarding the brain of many subjects including healthy subjects and patients with diseases related to brain and amyloid β, tau protein or α-synuclein-related diseases. The image is used, and the PET image at the time point nt of the relevant brain is used as the teacher data. Here, n can be n = 1, 2, ..., N. The time point nt may be different for each target person, but the time points of the MRI image and the PET image for one target person are the same time point. However, the time point nt of the MRI image and the time point nt of the PET image do not necessarily have to be at the same timing, and there may be a time difference of about several days, about several weeks, or about several months.

An MRI image at the time point nt regarding the brain is input to the model part 562, and the pseudo PET image output by the model part 562 approaches the actual PET image at the time point nt as teacher data (pseudo PET image and actual PET). A trained model can be generated by adjusting the parameters of the model unit 562 (eg, weight wij, bias bkl, filter parameters, etc.) so that the value of the loss function based on the image is minimized). ..

FIG. 5 is a schematic diagram showing a second example of a method of generating a trained model in the learning processing unit 56. The method of generating the trained model shown in FIG. 5 corresponds to the method of generating the model unit 552 of FIG. The difference from the first example shown in FIG. 4 is that in the second example, the PET image at the time point t1 is used as the input data for learning.

As shown in FIG. 5, as learning input data, MRI images and PET images taken at the first time point (time point t1 in the example of the figure) regarding the brains of many subjects including healthy subjects and dementia patients, and Using the MRI image of the brain taken at the second time point (in the example of the figure, time point nt, n = 2, 3, ..., N) after the first time point, the second time point (in the figure) is used as teacher data. In the example, the actual PET image obtained by the PET examination at time points nt, n = 2, 3, ..., N) is used.

The MRI image and PET image at the first time point and the MRI image at the second time point regarding the brain are input to the model unit 562 so that the pseudo PET image output by the model unit 562 approaches the actual PET image as the teacher data. Learning by adjusting the parameters of the model unit 562 (eg, weight wij, bias bkl, filter parameters, etc.) (so that the value of the loss function based on the pseudo PET image and the actual PET image is minimized). You can generate a finished model.

Compared to the case where the MRI image at the second time point (in the example of the figure, time point tn, n = 2, 3, ..., N) is simply input to the model unit 562, the first time point before the second time point (in the example of the figure). In the example of the figure, the information of the MRI image and the PET image at the time point t1) is also used, so the model unit 562 is set in consideration of the distribution state of amyloid β, tau protein or α-synuclein in the brain at the initial stage. It can be machine-learned, and it is expected that the accuracy of the pseudo PET image output by the trained model will be improved.

Next, the details of the diagnostic information about the brain will be explained. Hereinafter, the diagnostic information displayed on each of the terminal device 10 and the terminal device 20 will be described. First, diagnostic information displayed on the terminal device 10 used by medical professionals, researchers, and the like (hereinafter, also referred to as “doctors and the like”) will be described.

FIG. 6 is a schematic diagram showing a first example of diagnostic information displayed on a terminal device 10 of a doctor or the like. As shown in FIG. 6, the diagnostic information 101 includes first, second, third, and fourth pseudo PET images for a patient. When the actual PET image is obtained by performing the PET examination for the first time, the actual PET image may be displayed instead of the pseudo PET image for the first time. In this case, an identifier (mark) for identifying whether the image is an actual PET image or a pseudo PET image can be displayed. Further, the number of pseudo PET images is not limited to four. The time interval between the first time and the second time may be, for example, two years, and the interval after the second time may be one year, and the like, but the time interval is not limited thereto. In each pseudo PET image, the distribution state of amyloid β, tau protein or α-synuclein at a required site in the brain is visualized. The diagnostic information 101 includes the dementia diagnosis result for the patient. The dementia diagnosis result is a comprehensive judgment of the result of a cognitive ability test or the like by a doctor, and is not necessarily judged only by a pseudo PET image. In the example of the figure, it is judged as NC (Normal Control, healthy person, healthy elderly person, etc.) at the first time, but it is judged as MCI (mild cognitive impairment) after the second time. Diagnostic information 101 includes a state of amyloid accumulation for the patient. As shown in the figure, the state of amyloid accumulation may be the presence or absence of amyloid β accumulation at a required site in the brain (for example, cerebral cortex), or may be the degree of amyloid β accumulation (SUV). Although not shown, the presence / absence and degree of accumulation of abnormal proteins such as tau protein or α-synuclein may be displayed instead of or in addition to amyloid accumulation.

Amyloid β-induced amyloid plaque deposition is thought to be a pathological change that occurs from the earliest stages of the onset of Alzheimer's disease, and is said to begin more than a dozen years before clinical symptoms appear. As shown in FIG. 6, even in the case of NC, the distribution of amyloid β is likely to be remarkable (in the example of the figure, amyloid accumulation is “present”). Thus, early diagnosis of MCI is facilitated at an early stage when there is no subjective symptom of dementia.

In addition, amyloid plaque begins to accumulate from the base of the neocortex and gradually spreads over the entire cerebral cortex. Therefore, in MCI (mild cognitive impairment), AD (Alzheimer's disease) develops by observing the time-series display of pseudo-PET images or the transition of the distribution state of amyloid β based on the pseudo-PET images obtained in chronological order. Is likely to be foreseen. In the example of FIG. 6, since the accumulation degree of amyloid β in the cerebral cortex obtained from the pseudo PET image at the 4th time is higher than a predetermined threshold value, the possibility of developing AD is diagnosed as “possible”. In this way, it is possible to infer the presence or absence of future AD onset at the stage of MCI before the onset of Alzheimer's disease. In addition, the accumulation of tau protein in the brain is considered to be more closely related to the loss of nerve cells and memory impairment than to amyloid β, and it is expected that the severity in the early stage of dementia can be evaluated. In addition, excessive α-synuclein eventually causes death of nerve cells and causes symptoms such as motor dysfunction. Therefore, overexpression of α-synuclein in nerve cells occurs, for example, in the early stage of Parkinson's disease. It is believed that.

As described above, the diagnostic information includes a pseudo PET image at a plurality of time points and a PET diagnostic index specified based on each pseudo PET image. The PET diagnostic index is an index that can be obtained based on a pseudo PET image, and may be, for example, an accumulation degree (SUV) indicating the accumulation of amyloid β, tau protein, or α-synuclein, or SUVR. The degree of accumulation of amyloid β, tau protein or α-synuclein may be the degree of accumulation of each part of the brain, the total degree of accumulation of a plurality of parts, or the degree of accumulation in the entire brain. Since it is based on pseudo PET images at multiple time points, it is possible to generate diagnostic information considering the transition of accumulation of amyloid β, tau protein or α-synuclein.

Further, the diagnostic information includes the brain disease diagnosis result and the PET diagnostic index for the subject for each pseudo PET image. Brain disease diagnosis results include, for example, NC (Normal Control, healthy person, healthy elderly person, etc.), MCI (mild cognitive impairment) and the like. By including the brain disease diagnosis result and the transition of the PET diagnostic index, for example, the possibility of developing AD can be estimated.

FIG. 7 is a schematic diagram showing a second example of diagnostic information displayed on a terminal device 10 of a doctor or the like. As shown in FIG. 7, the diagnostic information 102 includes an MRI image and a pseudo PET image in June 2018, an MRI image and a pseudo PET image in January 2019, and an MRI image and a pseudo PET image in June 2019. Images, MRI images and pseudo PET images as of January 2020 are displayed in chronological order. In the diagnostic information 102, the MRI diagnostic index is displayed in chronological order together with the SUVR value as the PET diagnostic index.

The MRI diagnostic index is an index that can be obtained based on an MRI image, and as shown in FIG. 7, for example, it can be a value related to atrophy of each part of the brain (for example, the degree of atrophy in the region of interest). However, the atrophy is not limited to this, and the volume of each site may be used. The site may be a plurality of sites or a specific site. Methods for assessing cerebral atrophy include, for example, 3D T1-weighted whole-brain MRI separated into gray matter, white matter, and cerebrospinal fluid (segmentation) and morphologically transformed into standard brain templates to determine the gray matter concentration of individual subjects. VBM (Voxel Based Morphometry), which statistically compares the absolute volume value with the normal database for each voxel, can be used.

In Alzheimer's disease, it is said that atrophy of the medial temporal region of the brain, which is not observed in normal aging, occurs at an early stage. That is, attention can be paid to medial temporal atrophy as a means for predicting the onset of Alzheimer's disease at the stage of MCI (mild cognitive impairment). As shown in FIG. 7, while grasping the stage of MCI (mild cognitive impairment) of a patient by displaying a pseudo PET image and a PET diagnostic index in a time series, the medial temporal region is displayed by displaying the MRI image and the MRI diagnostic index in a time series. The onset of Alzheimer's disease can be predicted by observing the transition of atrophy of the part.

FIG. 8 is a schematic diagram showing a third example of diagnostic information displayed on a terminal device 10 of a doctor or the like. The diagnostic information includes age-specific amyloid accumulation information 103. The age-specific amyloid accumulation information 103 displays the average value of amyloid accumulation in the cerebral cortex of many healthy subjects with normal cognitive function by age group. It should be noted that the present invention is not limited to a healthy person, and may be, for example, MCI (mild cognitive impairment). For example, healthy people in their 50s have 10% amyloid accumulation, healthy people in their 60s have 15% amyloid accumulation, healthy people in their 70s have 30% amyloid accumulation, and healthy people in their 80s have 30% amyloid accumulation. Healthy individuals have 50% amyloid accumulation. The numerical value of amyloid accumulation is an example schematically shown, and may differ from the actual numerical value, and is not limited to these numerical values. Moreover, it is not limited to each age group, and may be gender. Diagnostic information also includes patient age, amyloid accumulation in the cerebral cortex, and up-to-date pseudo-PET images. Although not shown, the accumulation of abnormal proteins such as tau protein or α-synuclein may be indicated in place of or in addition to the accumulation of amyloid.

As described above, the diagnostic information includes the PET diagnostic index (in the example of FIG. 8, amyloid accumulation) and the PET diagnostic index for each age group of the comparison subject. By comparing with the PET diagnostic index (for example, the average value of the PET diagnostic index by age group) of the comparative subject (for example, a person with normal cognitive function), the brain condition of the subject may be compared with that of the comparative subject of the same age. You can judge whether it is bad or bad.

FIG. 9 is a schematic diagram showing a fourth example of diagnostic information displayed on a terminal device 10 of a doctor or the like. FIG. 9 shows an example of evaluating the efficacy of a drug for a disease related to the brain or a disease related to amyloid β. Drugs include unapproved drugs and approved drugs. The drug may be an approved drug, an unapproved drug, a drug, an investigational drug, a test drug, a control drug, or the like. It should be noted that not only the evaluation of the medicinal effect of the drug but also the evaluation of foods and drinks other than the drug, such as health foods, foods with functional claims, special purpose foods, nutritional supplements, supplements or foods for specified health use, etc. are included. good. For each drug, regions 111 and 113 in which pseudo PET images are displayed in chronological order and regions 112 and 114 in which changes in amyloid accumulation are displayed are displayed. Changes in amyloid accumulation are shown to change over an appropriate period, for example, 2 years, 5 years, 10 years. The period can be set by doctors and researchers as appropriate. For example, suppose that a large number of subjects diagnosed with MCI are divided into two groups, one group continues to receive drug A and the other group continues to receive drug B. As shown in FIG. 9, it is assumed that the change in the average amyloid accumulation of the subjects who received the drug A increased with the period. Further, it is assumed that the rate of change in the average amyloid accumulation of the subjects who received the drug B does not change so much. In this case, it can be obtained that the drug B is effective in suppressing the progression of dementia. Adverse events (including side effects, such as nausea, vomiting, headache, bradycardia, angina, myocardial infarction, cerebrovascular attack, spasm attack, amyloid-related imaging abnormalities (ARIA, ARIA-E, ARIA-H, etc.)) , At least one such as a fall) may be displayed. As a result, not only the efficacy of the drug but also side effects and the like can be confirmed. Further, although FIG. 9 shows an example of displaying both the pseudo PET image and the change in the amyloid accumulation, only one of the pseudo PET image and the change in the amyloid accumulation may be displayed. Although not shown, changes in the accumulation of abnormal proteins such as tau protein or α-synuclein may be displayed in place of changes in amyloid accumulation or in addition to changes in amyloid accumulation.

Next, the diagnostic information displayed on the patient's terminal device 20 will be described.

FIG. 10 is a schematic diagram showing a first example of diagnostic information displayed on the patient's terminal device 20. As shown in FIG. 10, the diagnostic information includes the consultation history 201 and the cognitive impairment transition 202. In the consultation history 201, a detail icon for displaying the date of consultation, the doctor in charge, the diagnosis result, and the details of the diagnosis result is displayed. The patient can see the past consultation history in a list, and can easily recognize, for example, the transition of the diagnosis result (mild in the example of the figure). In addition, by displaying the scheduled date of the next medical examination, it is possible to prevent forgetting to receive the medical examination.

In the cognitive impairment transition 202, the transition of the disability level of the patient's cognitive dysfunction is displayed. From the past diagnosis results, the patient can judge at a glance whether the disability level has deteriorated or has remained almost unchanged. In addition, by displaying the average disability level for each age group in the cognitive impairment transition 202, the patient can determine whether it is better or worse than the average value of the same age group with reference to his / her own age.

Next, the diagnostic information displayed when the detail icon is operated will be explained.

FIG. 11 is a schematic diagram showing a second example of diagnostic information displayed on the patient's terminal device 20. In FIG. 10, when the date of consultation is selected and the detail icon is operated, the diagnostic information 203 shown in FIG. 11 is displayed. In the example of FIG. 11, a case where the detail icons of the three consultation dates are operated is shown. A pseudo PET image is displayed on the diagnosis information 203 together with the consultation date. The pseudo PET image included in the diagnostic information 203 may be a pseudo PET image output by the model unit 552, or may be an image after image processing of the pseudo PET image output by the model unit 552 so that a general patient can easily understand it. good.

FIG. 12 is a schematic diagram showing a third example of diagnostic information displayed on the patient's terminal device 20. FIG. 12 shows an example of evaluating the efficacy of a drug for a disease related to the brain or a disease related to amyloid β, tau protein or α-synuclein. As shown in FIG. 12, the diagnostic information 204 displays a patient name (may be a patient number), a drug administered to the patient, a drug administration period, and the like. In addition, pseudo PET images are displayed in chronological order over the administration period. In addition, one pseudo PET image may be displayed. As schematically shown in FIG. 12, if the amyloid accumulation or tau protein accumulation represented by the pseudo-PET image does not increase over time, the drug is a disease associated with amyloid β, tau protein or α-synuclein. It can be seen that the progress is suppressed. Adverse events (including side effects, such as nausea, vomiting, headache, bradycardia, angina, myocardial infarction, cerebrovascular attack, spasm attack, amyloid-related imaging abnormalities (ARIA, ARIA-E, ARIA-H, etc.)) , At least one such as a fall) may be displayed. As a result, not only the efficacy of the drug but also side effects and the like can be confirmed. Although not shown, changes in amyloid accumulation, changes in tau protein accumulation, or α-synuclein accumulation over the plurality of time points in addition to or in place of the pseudo PET images at multiple time points. The change in may be displayed. The amyloid accumulation state and the tau protein accumulation state α-synuclein accumulation state at a certain point in time may be displayed during the administration period.

Next, the operation of the information processing apparatus 50 will be described.

FIG. 13 is a flowchart showing an example of the procedure for generating brain diagnostic information. Hereinafter, for convenience, the subject of processing will be described as the control unit 51. The control unit 51 determines whether or not it is the first medical examination (S11), and if it is the first medical examination (YES in S11), determines whether or not the subject has a PET image (S12). When there is a PET image (YES in S12), the control unit 51 acquires the PET image of the subject (S13). When it is not the first examination (NO in S11) and when there is no PET image of the subject (NO in S12), the control unit 51 performs the process of step S14 described later.

The control unit 51 acquires an MRI image of the subject (S14), and inputs the acquired MRI image and PET image (if there is a PET image) into the trained model (S15). The control unit 51 calculates the PET diagnostic index based on the pseudo PET image output by the trained model (S16), and calculates the MRI diagnostic index based on the acquired MRI image (S17).

The control unit 51 determines the presence or absence of another MRI image at the time of imaging (S18), and if there is another MRI image (YES in S18), the process after step S14 is repeated. When there is no other MRI image (NO in S18), the control unit 51 generates diagnostic information (S19) and ends the process.

FIG. 14 is a flowchart showing an example of the procedure of the trained model generation process. The control unit 51 reads out the model (S31) and sets the initial values of the parameters of the neural network (S32). The control unit 51 sets the count value n = 1 (S33) and acquires an MRI image and a PET image at the time of imaging tun (S34).

The control unit 51 inputs the PET image at the imaging time t1 and the MRI image at the imaging time tun into the model (model unit 562), and outputs the pseudo PET image at the imaging time tun and the PET at the imaging time tun output by the model. The model parameters are adjusted so that the value of the loss function based on the image is minimized (S35).

The control unit 51 determines whether or not the value of the loss function is within the permissible range (S36), and if it is not within the permissible range (NO in S36), the process after step S34 is repeated. When the value of the loss function is within the permissible range (YES in S36), the control unit 51 determines whether or not all the processing at the time of imaging is completed (S37).

When all the processing at the time of imaging is not completed (NO in S37), the control unit 51 adds 1 to the count value (S38), and repeats the processing in steps S34 and subsequent steps. When all the processing at the time of imaging is completed (YES in S37), the control unit 51 stores the generated trained model in the model unit 552 (S39), and ends the processing.

The information processing device 50 can also be realized by using a computer equipped with a CPU (processor), RAM, and the like. A computer program (which can be recorded on a recording medium) that defines the processing procedure as shown in FIGS. 13 and 14 is read by a recording medium reading unit provided in the computer, the read computer program is loaded into the RAM, and the computer program is loaded. The information processing device 50 can be realized on a computer by executing the program on a CPU (processor).

In the above-described embodiment, MCI (mild cognitive impairment) and AD (Alzheimer's disease) have been mainly described, but the diseases targeted by this embodiment are not limited to these, and for example, Lewy body dementias. It can also be applied to dementia (DLB), multiple sclerosis (MS), Parkinson's disease, frontal lobe temporal lobe degeneration (FTLD), and other neurodegenerative diseases. For example, brain atrophy in Lewy body disease is associated with the accumulation of amyloid β or tau protein, similar to Alzheimer's disease. Also, Alzheimer's disease (amyloid plaque), dementia (Lewy bodies), synucleinopathy (protein aggregates in neurons, nerve fibers, glial cells), multiple system atrophy, Parkinson's disease (Lewy bodies), motor function. Disorders and the like are associated with the accumulation of α-synuclein.

In the computer program of the present embodiment, when one or a plurality of MRI images taken at one or a plurality of time points regarding the subject's brain are acquired by the computer and the MRI image regarding the brain is input, the pseudo PET for the brain is described. The acquired one or a plurality of MRI images are input to the trained model that outputs an image, and a process of outputting a pseudo PET image at the one or a plurality of time points regarding the subject's brain is executed.

The information processing apparatus of the present embodiment has an acquisition unit that acquires one or a plurality of MRI images captured at one or a plurality of time points regarding the subject's brain, and when an MRI image regarding the brain is input, the pseudo brain. The trained model that outputs the PET image is provided with an output unit that inputs the acquired one or a plurality of MRI images and outputs a pseudo PET image at the one or a plurality of time points regarding the subject's brain.

The information processing method of the present embodiment acquires one or a plurality of MRI images captured at one or a plurality of time points regarding the subject's brain, and when the MRI image regarding the brain is input, the pseudo PET image regarding the brain is obtained. The acquired one or more MRI images are input to the trained model to be output, and the pseudo PET images at the one or more time points regarding the subject's brain are output.

The trained model generation method of the present embodiment acquires the MRI image and the PET image captured at the first time point regarding the brain, and obtains the MRI image and the PET image regarding the brain taken at the second time point after the first time point. Using the PET images acquired and the MRI and PET images captured at the first and second time points, respectively, the MRI and PET images of the brain captured at the first time point, and the above-mentioned When an MRI image of the brain captured at the second time point is input, a trained model that outputs a pseudo PET image of the brain at the second time point is generated.

When the image output device of the present embodiment inputs an MRI image and a PET image taken at the first time point regarding the brain and an MRI image about the brain taken at the second time point after the first time point, the image output device has the same. A pseudo PET image relating to the brain at the second time point is output.

The computer program acquires one or more MRI images of the subject's brain taken at one or more time points. The MRI image is not only an MRI image obtained by an MRI apparatus (for example, a T1 weighted image, a T2 weighted image, a diffusion weighted image, a flare image, a diffusion tensor image), but also a process obtained by performing a predetermined operation on the MRI signal. It also includes images (eg, QSM images) and the like. The plurality of time points can be, for example, a time point for performing an MRI examination of a subject (patient).

When the trained model inputs an MRI image at one or more time points regarding the subject's brain, it outputs a pseudo PET image at each of the one or more time points regarding the subject's brain. The "pseudo PET image" is a word used to distinguish it from an actual PET image, and is substantially equivalent to a PET image. Similar to the PET image, the pseudo PET image includes, for example, distribution information that visualizes the distribution of amyloid β (for example, the SUVR value for each voxel). SUVR (Standardized Uptake Value Ratio) is the sum of SUVs (Standardized Uptake Value: amyloid β protein accumulation) of four sites of cerebellar gray matter (prefrontal cortex, anterior and posterior cingulate cortex, parietal lobe, and lateral temporal lobe). , Can be determined by dividing by the SUV of the cerebellum. The trained model can be configured with a neural network, and for example, GAN (Generative Adversarial Network), U-net, or a combination thereof can be used.

The pre-learning model (also simply referred to as "model") uses an MRI image of the brain as input data for learning, and a PET image at the same time as when the MRI image of the brain was captured as teacher data. Input an MRI image of the brain into the model and adjust the model parameters (eg, weights, biases, filter parameters, etc.) so that the pseudo PET image output by the model is closer to the actual PET image as teacher data. Allows you to generate a trained model.

Since the pseudo PET image at one or a plurality of time points is output, it is possible to provide brain diagnostic information equivalent to the PET image without performing a PET examination.

In the computer program of the present embodiment, the computer acquires a PET image of the subject's brain captured at the first time point among the plurality of time points, and the MRI image and PET image taken at the first time point of the brain. A learning model that outputs a pseudo PET image of the brain at the second time point when an image and an MRI image of the brain imaged at the second time point after the first time point of the plurality of time points are input. Is input with a PET image of the subject's brain captured at the first time point, and outputs a pseudo PET image of the subject's brain at the second time point.

The computer program acquires PET images of the subject's brain captured at the first of multiple time points. The first time point may be the first of multiple time points (ie, the time point of the PET examination received at the same time as the first MRI examination), or the same time point as the second or third MRI examination after the first time. good.

When the trained model inputs an MRI image and a PET image taken at the first time point regarding the brain and an MRI image about the brain taken at the second time point after the first time point, the trained model relates to the brain at the second time point. Output a pseudo PET image. For example, when the MRI image and PET image obtained by the MRI examination and the PET examination performed at the first examination and the MRI image obtained by the second and subsequent MRI examinations are input to the trained model, the trained model is obtained. It is possible to output a pseudo PET image related to the brain at the time of the second and subsequent MRI examinations.

The pre-learning model uses MRI and PET images captured at the first time point on the brain and MRI images taken at the second time point after the first time point as training input data, and is used as a teacher. As the data, the actual PET image obtained by the PET examination at the second time point is used. Input the MRI image and PET image at the first time point and the MRI image at the second time point to the model, and the parameters of the model so that the pseudo PET image output by the model approaches the actual PET image as the teacher data. Trained models can be generated by adjusting (eg, weights, biases, filter parameters, etc.). Compared to simply inputting the MRI image at the second time point into the model, the information of the MRI image and PET image at the first time point before the second time point is also used, so it reflects the change in the state of the brain. It is expected that the model can be machine-learned and the accuracy of the pseudo PET image output by the model can be improved.

The computer program of the present embodiment causes the computer to execute a process of displaying the pseudo PET images at the plurality of time points in chronological order.

The computer program displays pseudo PET images at multiple time points in chronological order. This makes it possible to provide brain diagnostic information while comparing pseudo PET images at a plurality of time points.

The computer program of the present embodiment causes the computer to execute a process of generating diagnostic information based on the pseudo PET images at the plurality of time points.

The computer program generates diagnostic information based on pseudo PET images at multiple time points. Since pseudo PET images at a plurality of time points can be obtained, it is possible to obtain diagnostic information based on not only the state of the brain at a specific time point but also changes in the state of the brain over a plurality of time points. This makes it possible to provide brain diagnostic information equivalent to a PET image without performing a PET examination.

In the computer program of the present embodiment, the diagnostic information includes a PET diagnostic index specified based on the pseudo PET images at the plurality of time points.

The diagnostic information includes a PET diagnostic index specified based on pseudo PET images at a plurality of time points. The PET diagnostic index is an index that can be obtained based on a pseudo PET image, and may be, for example, the degree of accumulation (SUV) representing the accumulation of amyloid β or tau protein, or SUVR. The degree of accumulation of amyloid β or tau protein may be the degree of accumulation of each part of the brain, the total degree of accumulation of a plurality of parts, or the degree of accumulation in the entire brain. Since it is based on pseudo PET images at a plurality of time points, it is possible to generate diagnostic information considering the transition of accumulation of amyloid β or tau protein.

In the computer program of the present embodiment, the diagnostic information includes a brain disease diagnosis result for the subject and the PET diagnostic index for each pseudo PET image.

The diagnostic information includes the brain disease diagnosis result and the PET diagnostic index for the subject for each pseudo PET image. Brain disease diagnosis results include, for example, NC (healthy elderly), MCI (mild cognitive impairment) and the like. By including the brain disease diagnosis result and the transition of the PET diagnostic index, for example, the possibility of developing AD can be estimated.

In the computer program of the present embodiment, the diagnostic information includes the PET diagnostic index and the PET diagnostic index of the comparison target person.

The diagnostic information includes the PET diagnostic index and the PET diagnostic index of the comparison target. By comparing with the PET diagnostic index (for example, the average value of the PET diagnostic index by age group) of the comparative subject (for example, a person with normal cognitive function), the brain condition of the subject may be compared with that of the comparative subject of the same age. You can judge whether it is bad or bad.

The terminal device of this embodiment is a pseudo-PET image of a subject's brain administered with a particular drug over one or more time points, amyloid accumulation over the one or more time points, and the one or more time points. Display at least one of the tau protein accumulations over.

The terminal device may be used by a doctor or the like, or may be used by a subject (patient). One pseudo-PET image, or pseudo-PET image, is displayed in chronological order over the administration period of a specific drug, or changes in amyloid accumulation state or amyloid accumulation, or changes in tau protein accumulation state or tau protein accumulation are displayed. Therefore, it is possible to confirm the presence or absence of the effect of the drug.

The terminal device of the present embodiment displays adverse events related to the drug.

Adverse events include side effects. For example, including at least one of nausea, vomiting, headache, bradycardia, angina, myocardial infarction, cerebrovascular attack, seizure attack, amyloid-related imaging abnormalities (ARIA, ARIA-E, ARIA-H, etc.), falls, etc. .. As a result, not only the efficacy of the drug but also side effects and the like can be confirmed.

1 Communication network 10, 20 Terminal equipment 50 Information processing equipment 51 Control unit 52 Communication unit 53 Storage unit 54 Diagnostic information generation unit 55 Estimating unit 551 Input data generation unit 552 Model unit 56 Learning processing unit 561 Learning data generation unit 562 Model unit 563 Parameter determination unit 57 Interface unit 61 Image DB
71 PET device 72 MRI device

Claims (13)

1. On the computer
   Obtain one or more MRI images of the subject's brain taken at one or more time points.
   When an MRI image about the brain is input, the acquired one or more MRI images are input to the trained model that outputs a pseudo PET image about the brain, and the acquired MRI image is input to the subject's brain at the one or more time points. Output a pseudo PET image,
   A computer program that executes processing.
2. On the computer
   A PET image of the subject's brain captured at the first of the plurality of time points was acquired.
   When the MRI image and PET image taken at the first time point regarding the brain and the MRI image about the brain taken at the second time point after the first time point among the plurality of time points are input, the second time point is mentioned. The PET image of the subject's brain captured at the first time point is input to the learning model that outputs the pseudo PET image of the brain in the second time point, and the pseudo PET image of the subject's brain at the second time point is obtained. Output,
   The computer program according to claim 1, wherein the processing is executed.
3. On the computer
   The pseudo PET images at the plurality of time points are displayed in chronological order.
   The computer program according to claim 1 or 2, wherein the process is executed.
4. On the computer
   Generate diagnostic information based on the pseudo PET images at the plurality of time points.
   The computer program according to any one of claims 1 to 3, wherein the processing is executed.
5. The diagnostic information is
   Includes PET diagnostic indicators identified based on the pseudo-PET images at the plurality of time points.
   The computer program according to claim 4.
6. The diagnostic information is
   Each pseudo PET image includes a brain disease diagnosis result for the subject and the PET diagnostic index.
   The computer program according to claim 5.
7. The diagnostic information is
   The PET diagnostic index and the PET diagnostic index of the comparison subject are included.
   The computer program according to claim 5 or 6.
8. An acquisition unit that acquires one or more MRI images taken at one or more time points with respect to the subject's brain.
   When an MRI image about the brain is input, the acquired one or more MRI images are input to the trained model that outputs a pseudo PET image about the brain, and the acquired MRI image is input to the subject's brain at the one or more time points. An information processing device including an output unit that outputs a pseudo PET image.
9. Pseudo-PET images of a subject's brain treated with a particular drug over one or more time points, amyloid accumulation over one or more time points, and at least one of the tau protein accumulations over one or more time points. indicate,
   Terminal device.
10. Display adverse events related to the drug,
    The terminal device according to claim 9.
11. Obtain one or more MRI images of the subject's brain taken at one or more time points.
    When an MRI image about the brain is input, the acquired one or more MRI images are input to the trained model that outputs a pseudo PET image about the brain, and the acquired MRI image is input to the subject's brain at the one or more time points. Output a pseudo PET image,
    Information processing method.
12. MRI and PET images taken at the first time point regarding the brain are acquired and
    MRI images and PET images related to the brain captured at the second time point after the first time point were acquired.
    Using the acquired MRI and PET images captured at the first and second time points, respectively, the MRI and PET images of the brain captured at the first time point and the second time point are captured. When an MRI image relating to the brain is input, a trained model that outputs a pseudo PET image relating to the brain at the second time point is generated.
    Trained model generation method.
13. When the MRI image and PET image taken at the first time point regarding the brain and the MRI image about the brain taken at the second time point after the first time point are input, the pseudo image about the brain at the second time point is input. Output PET image,
    Image output device.
    

Images