Estimator learning device, estimator learning method, and estimator learning program

Abstract

Provided is an estimating device comprising a processor for constructing an estimator for estimating the rate of deterioration of a subject's kidney disease by machine learning in which a learning magnetic resonance (MR) image capturing a kidney-including area of the subject is acquired, a kidney image extracting the kidney portion included in the learning MR image is generated, the kidney image is converted by non-rigid conversion into a standard kidney image, a learning feature amount based on the kidney image obtained by the conversion is calculated, and a data set comprising the calculated learning feature amount and the rate of deterioration of the subject's kidney disease is used as training data.

Description

[Technical Field]

[0001] The present invention relates to an estimator learning device, an estimator learning method, and an estimator learning program. [Background technology]

[0002] The estimated glomerular filtration rate (eGFR) is an index used to examine kidney function. eGFR is calculated based on age, sex, and serum creatinine level. Serum creatinine level can be detected through a blood test. The change in eGFR over time (eGFR slope) is also used as an indicator of the rate of deterioration of kidney disease. If the change in eGFR over time is small, kidney function is considered stable. If the eGFR decreases over time, kidney disease is considered to be worsening. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Publication No. 2019-204484 [Patent Document 2] Special Publication No. 2017-502439 Summary of the Invention [Problem to be solved by the invention]

[0004] The eGFR slope can be calculated, for example, from the rate of change (the amount of change in eGFR per hour) between the eGFR at a certain point in time and the eGFR several months to several years later. However, since the eGFR slope must be calculated at least twice at intervals, it is difficult to determine the eGFR slope (the rate of deterioration of kidney disease) at a certain point in time. Renal MR (Magnetic Resonance) imaging is also available as a clinical test for noninvasively detecting kidney conditions. However, it is difficult to estimate the rate of deterioration of kidney disease using renal MR imaging.

[0005] An object of the present invention is to provide a technique that can estimate the rate of progression of renal disease based on renal MR images. [Means for solving the problem]

[0006] In order to solve the above problems, the following measures are adopted. That is, the first aspect is: A training MR (Magnetic Resonance) image of a region of a subject including the kidney is acquired, a kidney image is generated by extracting the kidney portion included in the training MR image, the kidney image is converted into a standard kidney image by non-rigid transformation, and training features are calculated based on the converted kidney image, a processor that constructs an estimator that estimates the worsening rate of the renal disease of the subject by machine learning using, as training data, a dataset including the calculated learning features and the worsening rate of the renal disease of the subject; The estimation device is provided with:

[0007] The disclosed aspects may be realized by a program being executed by an information processing device. That is, the disclosed configuration may be specified as a program for causing an information processing device to execute the processes executed by each means in the above aspects, or a computer-readable recording medium on which the program is recorded. The disclosed configuration may also be specified as a method by which an information processing device executes the processes executed by each means. The disclosed configuration may also be specified as a system including an information processing device that performs the processes executed by each means. [Effects of the Invention]

[0008] According to the present invention, it is possible to provide a technique that can estimate the rate of progression of renal disease based on renal MR images. [Brief explanation of the drawings]

[0009] [Figure 1] FIG. 1 is a diagram illustrating an example of the configuration of an estimation device according to an embodiment. [Figure 2]FIG. 2 is a diagram showing an example of a T2* map image. [Figure 3] FIG. 3 is a diagram showing an example of an operation flow when an estimator is constructed in the estimation device. [Figure 4] FIG. 4 is an MR image of a region including the kidneys of a subject. [Figure 5] FIG. 5 is an image in which the area corresponding to the kidney is extracted from the MR image in FIG. [Figure 6] FIG. 6 is a diagram illustrating an example of non-rigid transformation processing of a kidney. [Figure 7] FIG. 7 is a diagram showing an example of a histogram generated by the feature amount calculation process. [Figure 8] FIG. 8 is a diagram showing an example of an operational flow when the estimation device estimates the rate of deterioration of a renal disease. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, embodiments will be described with reference to the drawings. The configurations of the embodiments are examples, and the configuration of the invention is not limited to the specific configurations of the disclosed embodiments. When implementing the invention, specific configurations according to the embodiments may be appropriately adopted.

[0011] [Embodiment] (Configuration example) FIG. 1 is a diagram illustrating an example of the configuration of an estimation device according to this embodiment. The estimation device 100 illustrated in FIG. 1 has the configuration of a typical computer. The estimation device 100 according to this embodiment includes a processor 101, a memory 102, a storage unit 103, an input unit 104, an output unit 105, and a communication control unit 106. These components are connected to one another via a bus. The memory 102 and the storage unit 103 are computer-readable recording media. The hardware configuration of the estimation device 100 is not limited to the example illustrated in FIG. 1, and components may be omitted, replaced, or added as appropriate.

[0012] The estimation device 100 of this embodiment acquires clinical information of a subject (such as gender, age, urinary protein, and serum creatinine level), MR images of a region of the subject including the kidneys, and a renal disease deterioration rate (eGFR slope) calculated after the MR images of the subject. The estimation device 100 performs predetermined processing on the MR images to construct an estimator that estimates the renal disease deterioration rate from the clinical information and MR images of the subject. The estimation device 100 also uses the constructed estimator to estimate the renal disease deterioration rate (eGFR slope) from the clinical information and MR images of the subject. MR images are images of a body region captured using a nuclear magnetic resonance apparatus. MR images of the kidneys reflect the renal oxygen status, renal fibrosis, and renal perfusion. The renal oxygen status, renal fibrosis, and renal perfusion are related to renal disease.

[0013] The estimating device 100 can be realized using a dedicated or general-purpose computer such as a workstation (WS), a personal computer (PC), a smartphone, a tablet terminal, or an electronic device equipped with a computer. The estimating device 100 can be realized using a computer (server device) that provides services over a network. The estimating device 100 can be realized by a computer that executes parallel processing using a message passing interface (MPI), which is a large-scale parallelization of CPUs or GPUs. The estimating device 100 can realize functions that meet a predetermined purpose by having a processor 101 load a program stored in a recording medium into a working area of ​​a memory 102 and execute the program, and by controlling each component, etc. through the execution of the program.

[0014] The processor 101 is, for example, a CPU (Central Processing Unit), etc. The processor 101 loads and executes programs stored in the memory 102 to perform image acquisition processing, kidney region extraction processing, non-rigid transformation processing, feature amount calculation processing, estimator construction processing, estimation processing, etc. The processor 101 can also acquire data used in each processing from other devices via the storage unit 103 or the communication control unit 106.

[0015] The image acquisition process is a process of acquiring MR images (MRI) of a region of the subject, including the kidneys, from another information processing device or the like. In addition, in the image acquisition process, information about the subject, such as the subject's clinical information, is acquired along with the MR images. Examples of the subject's clinical information include gender, age, mean blood pressure, urinary protein level, uric acid level, serum creatinine level, eGFR, and whether or not the subject has a history of diabetes. Other information may also be used as the subject's clinical information. Furthermore, it is assumed that the eGFR slope (the rate of deterioration of renal disease) of the subject corresponding to the MR images used in constructing the estimator has been calculated after the MR images were taken (e.g., several months to several years later). In other words, it is assumed that the eGFR slope has been calculated based on the time change in eGFR after the MR images of the subject were taken. For example, it is assumed that the eGFR slope (the rate of deterioration of renal disease) of the subject has been calculated based on the eGFR at the time the MR images were taken and the eGFR one year after the MR images were taken. Examples of MR images that can be used include a T2* map image, an R2* map image, and an ADC (Apparent Diffusion Coefficient) map image. The MR images used here are not limited to these. The T2* map image used here is created based on 12 T2*-weighted images taken at different echo times (TE). To suppress the influence of fat components that penetrate into the renal parenchyma from the renal hilum, all TE times are set to be in-phase. The T2* map image is not limited to those described here. The R2* map image is an image created based on the R2* value (R2*=1 / T2*), which is the reciprocal of the T2* value used in the T2* map image.

[0016] Figure 2 shows examples of T2* map images. The image on the left of Figure 2 is an example of a T2* map image including a normal kidney. The image on the right of Figure 2 is an example of a T2* map image including a kidney with chronic kidney disease and reduced renal function.

[0017] The kidney region extraction process extracts the area corresponding to the kidney from the MR images acquired in the image acquisition process. The kidney region extraction process uses an estimator that estimates the kidney area from MR images constructed in advance using deep learning or the like. In the kidney region extraction process, if there are multiple MR images for one subject, the MR image with the largest extracted kidney may be used in subsequent processing. The MR image with the largest kidney is likely to include the area near the center of the kidney, which is thought to contribute to improving the performance of the estimator when constructing it.

[0018] The non-rigid transformation process is a process for transforming the shape of the kidney extracted in the kidney region extraction process into a standard kidney shape. A well-known non-rigid transformation process is used here. The standard kidney shape is, for example, the shape of a normal kidney. Transforming the kidney into the standard kidney shape improves the performance of the estimator during construction.

[0019] The feature calculation process extracts kidney features from an image of the kidney that has been transformed into a standard kidney shape by non-rigid transformation. For example, the kidney is divided into 12 layers from the cortex to the medulla, and the pixel values ​​of the pixels in each layer are tallied to obtain the feature. Examples of the pixel value aggregation include calculating the frequency of pixel values ​​for each layer and the average pixel value for each layer.

[0020] The estimator construction process is a process of constructing an estimator that estimates the rate of deterioration of a subject's renal disease from the subject's clinical information and kidney features, using a dataset containing the subject's clinical information, kidney features determined from the subject's MR images, and the rate of deterioration of the subject's renal disease as training data.

[0021] The estimation process uses the estimator constructed in the estimator construction process to estimate the rate of deterioration of the subject's renal disease from the subject's clinical information and kidney features obtained from the subject's MR images.

[0022] The memory 102 is configured by, for example, a random access memory (RAM), a RAM, and a read only memory (ROM). The memory 102 is also called a main storage device.

[0023] The storage unit 103 is, for example, an erasable programmable read only memory (EPROM) or a hard disk drive (HDD). The storage unit 103 may also include removable media, i.e., portable recording media. The removable media is, for example, a universal serial bus (USB) memory or a disc recording medium such as a compact disc (CD) or a digital versatile disc (DVD). The storage unit 103 is also called a secondary storage device.

[0024] The storage unit 103 stores various programs, various data, and various tables used in the estimation device 100 in a readable and writable recording medium. The storage unit 103 stores an operating system (OS), various application programs, various tables, etc. The information stored in the storage unit 103 may be stored in the memory 102. Furthermore, the information stored in the memory 102 may be stored in the storage unit 103.

[0025] The storage unit 103 has installed therein programs for executing image acquisition processing, kidney region extraction processing, non-rigid transformation processing, feature amount calculation processing, estimator construction processing, estimation processing, etc. The storage unit 103 also stores various data used in each process, etc. in the estimation device 100. The storage unit 103 stores MR images of the subject, clinical information, eGFR slope, etc.

[0026] The operating system is software that mediates between software and hardware, manages memory space, manages files, and manages processes and tasks. The operating system also includes a communication interface. The communication interface is a program that exchanges data with other external devices connected via the communication control unit 106. Examples of external devices include other information processing devices and external storage devices.

[0027] The input unit 104 includes a keyboard, a pointing device, a wireless remote control, a touch panel, etc. The input unit 104 may also include an input device for video or images such as a camera, and an input device for audio such as a microphone.

[0028] The output unit 105 includes a display device such as an LCD (Liquid Crystal Display), an EL (Electroluminescence) panel, a CRT (Cathode Ray Tube) display, a PDP (Plasma Display Panel), a printer, etc. The output unit 105 may also include an audio output device such as a speaker.

[0029] The communication control unit 106 connects to other devices and controls communication between the estimating device 100 and the other devices. The communication control unit 106 is, for example, a LAN (Local Area Network) interface board, a wireless communication circuit for wireless communication, or a communication circuit for wired communication. The LAN interface board and the wireless communication circuit are connected to a network such as the Internet.

[0030] The steps of writing a program include processes that are executed chronologically in the order described, as well as processes that are not necessarily executed chronologically but are executed in parallel or individually. Some of the steps of writing a program may be omitted.

[0031] In this embodiment, the series of processes executed by the processor 101 can be executed by either hardware or software. The hardware components are hardware circuits, such as a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a gate array, a combination of logic gates, an analog circuit, etc.

[0032] (Example of operation) <Estimator Construction> 3 is a diagram showing an example of an operational flow for constructing an estimator in the estimation device 100. Using the subject's clinical information, feature values ​​based on the subject's MR images, and the subject's renal disease deterioration rate as training data, the estimation device 100 constructs an estimator that estimates the renal disease deterioration rate from the clinical information and feature values.

[0033] In S101, the processor 101 of the estimation device 100 acquires MR images of the subject, etc., which are the basis of training data used to construct the estimator. The processor 101 acquires MR images of the subject, which are stored in the storage unit 103. The MR images are images of a region of the subject that includes the kidneys. The processor 101 also acquires information about the subject, such as clinical information about the subject, from the storage unit 103, along with the MR images of the subject. The processor 101 also acquires the renal disease deterioration rate (eGFR slope) of the subject, which is stored in the storage unit 103. It is assumed that the renal disease deterioration rate of the subject is calculated after the MR images are taken (e.g., several months to several years later). It is assumed that the renal disease deterioration rate of the subject is stored in the storage unit 103 in advance. The processor 101 may acquire these data from another device via the communication control unit 106, a network, etc. The MR images, clinical information, and renal disease deterioration rate acquired here are the MR images, clinical information, and renal disease deterioration rate used for learning.

[0034] In S102, the processor 101 extracts a portion corresponding to the kidney from the MR image acquired in S101. The processor 101 extracts the portion corresponding to the kidney from the MR image using an estimator that has been previously constructed by deep learning or the like to estimate the kidney portion from the MR image. The estimator is previously stored in the storage unit 103. The processor 101 stores an image of the portion (region) corresponding to the kidney (kidney image) from the extracted MR image in the storage unit 103. The processor 101 may also store in the storage unit 103 an MR image and information on the portion (region) corresponding to the kidney included in the image in association with each other. When there are multiple MR images for one subject, the processor 101 extracts a portion corresponding to the kidney from each MR image. Furthermore, the processor 101 uses the MR image with the largest extracted portion corresponding to the kidney in subsequent processing. The extraction of the portion corresponding to the kidney may be performed by other methods. For example, the extraction of the portion corresponding to the kidney may be performed by the user specifying the region of the kidney via the input unit 104.

[0035] Figure 4 is an MR image of the area including the kidneys of a subject. The MR image in Figure 4 is an image of a person's abdomen viewed from the front. The kidneys are visible on both sides of the image.

[0036] Figure 5 is an image in which the parts corresponding to the kidneys have been extracted from the MR image in Figure 4. In Figure 5, the parts corresponding to the kidneys (right kidney and left kidney) have been extracted by an estimator that estimates the kidney parts from the MR image in Figure 4.

[0037] In S103, the processor 101 converts the image of the portion corresponding to the kidney extracted in S102 into a standard kidney shape by non-rigid transformation. A well-known method is used as the non-rigid transformation. Here, the non-rigid transformation is a process of converting the kidney shape of the image of the portion corresponding to the kidney extracted in S102 into a standard kidney shape. The standard kidney shape is, for example, the shape of a normal kidney. The kidney shape may be deformed or atrophied due to the influence of renal disease, etc. By converting to the standard kidney shape by non-rigid transformation, the kidney shape is made uniform and the influence of deformation, atrophy, etc. is reduced. Furthermore, by making the kidney shape uniform by non-rigid transformation, the performance of the estimator is improved. The processor 101 stores the kidney image converted by non-rigid transformation in the memory unit 103.

[0038] FIG. 6 is a diagram showing an example of non-rigid transformation processing of a kidney. The example in FIG. 6 shows an example in which the kidneys of a first subject and a second subject are transformed into a standard kidney. Points A1, B1, and C1 in the kidney of the first subject in FIG. 6 correspond to points A0, B0, and C0 in the standard kidney, respectively. Points A2, B2, and C2 in the kidney of the second subject in FIG. 6 correspond to points A0, B0, and C0 in the standard kidney, respectively. Through the non-rigid transformation processing, each pixel (each point) in the image of the kidney before transformation is associated with one of the pixels (points) in the image of the standard kidney. The pixel values ​​(color, shading) of corresponding pixels (each point) are maintained before and after transformation. For example, points A0, A1, and A2 correspond to each other. That is, for example, the pixel value of the kidney of the first subject corresponding to the pixel value of point A0 in the standard kidney is the pixel value of point A1 in the image of the kidney of the first subject.

[0039] In S104, the processor 101 calculates kidney features from the image of the kidney with the standard shape converted in S103. The processor 101, for example, divides the kidney into multiple layers (e.g., 12 layers) from the cortex to the medulla, and calculates the feature by aggregating pixel values ​​of pixels included in each layer. For example, the multiple layers from the cortex to the medulla may be layers that have been previously divided into multiple layers by a specialist or the like for a normal kidney. By dividing the kidney into multiple layers from the cortex to the medulla, feature values ​​corresponding to each internal part of the kidney can be calculated. The feature for each layer may be, for example, a histogram representing the frequency of occurrence of pixel values ​​in each layer or an average value of pixel values ​​in each layer. Although pixel values ​​are aggregated for each layer here, pixel values ​​of pixels included in the kidney image may be aggregated without dividing the kidney into multiple layers. That is, the processor 101 may use a histogram representing the frequency of occurrence of pixel values ​​of all pixels included in the kidney image as the feature, or the average value of pixel values ​​of all pixels as the feature. The processor 101 may use pixel values ​​of one or more specific pixels (positions) in the kidney image as features. The specific kidney position is, for example, a position that is medically considered to have a high correlation with the rate of deterioration of renal disease. The position can be easily specified by aligning the shape of the kidney to a standard shape. The processor 101 stores the calculated feature in the memory unit 103. The feature is stored in association with the clinical information of the subject and the rate of deterioration of renal disease. Other quantities based on the kidney image may also be used as features. Multiple quantities may be used as features from the above-mentioned histogram of pixel values ​​for each layer, the average pixel value for each layer, the histogram of pixel values ​​of the kidney image, the average pixel value of the kidney image, and pixel values ​​of one or more specific pixels (positions) in the kidney image. The feature calculated here is a quantity indicating the characteristics of the kidney contained in the MR image. The feature calculated here is a feature for learning.

[0040] FIG. 7 is a diagram showing an example of a histogram generated by the feature calculation process. The horizontal axis of the histogram in FIG. 7 indicates pixel values, and the vertical axis indicates the frequency of occurrence (number of occurrences) of each pixel value. The histogram shows the frequency of occurrence (number of occurrences) of pixel values ​​of pixels included in the image of each layer. Here, the vertical axis indicates the frequency of occurrence (number of occurrences) of pixel values, but it may also indicate the rate of occurrence of pixel values. Here, pixel values ​​are assumed to take values ​​from 0 to 255.

[0041] The processor 101 performs the processes from S101 to S104 using MR images of various subjects, calculates the feature amount for each MR image, and stores it in the storage unit 103.

[0042] In S105, the processor 101 uses a deep learning model of machine learning to construct an estimator that estimates the renal disease worsening rate (eGFR slope) from the feature values ​​and clinical information of multiple subjects, including the feature values ​​calculated in S104, the clinical information of the subjects acquired in S101, and the renal disease worsening rate (eGFR slope) of the subjects, as training data. Any deep learning model may be used here. The processor 101 stores the constructed estimator in the memory unit 103. To construct the estimator, techniques using deep learning using a neural network, regression SVM, regression random forest, multiple regression analysis, look-up table, or other learning space techniques may be used. Methods other than machine learning may also be used to construct the estimator. Using more training data allows for the construction of an estimator with higher performance. Here, the training data dataset does not necessarily include all or part of the clinical information. That is, an estimator may be constructed using multiple datasets including the feature values ​​calculated in S104 and the rate of deterioration of renal disease as training data. If the dataset does not include some or all of the clinical information, it is not necessary to acquire the clinical information that is not included in S101.

[0043] Estimating the rate of progression of kidney disease 8 is a diagram showing an example of an operational flow when the estimation device estimates the worsening rate of renal disease. The estimation device 100 estimates the worsening rate of renal disease of the subject using clinical information of the subject to be estimated, feature values ​​based on MR images of the subject, and the estimator constructed according to the operational flow in FIG.

[0044] In S201, the processor 101 acquires MR images and clinical information of a subject to be estimated from the storage unit 103 or another device via the communication control unit 106, a network, etc. If some or all of the clinical information is not included in the dataset used as training data when the estimator is constructed in the operational flow of Figure 3, the processor 101 does not need to acquire the clinical information that is not included.

[0045] The processing from S202 to S204 is the same as the processing from S102 to S104 in Fig. 3. Through the processing from S202 to S204, the processor 101 calculates feature amounts based on the MR image of the subject to be estimated, and stores the calculated feature amounts in the storage unit 103. The feature amounts calculated by the processor 101 here are the same type of feature amounts as the feature amounts (feature amounts included in the training data) used when constructing the estimator in the operational flow of Fig. 3. That is, for example, if the feature amounts used when constructing the estimator are histograms of pixel values ​​for each layer of an image of the kidney, the feature amounts calculated here are histograms of pixel values ​​for each layer of an image of the kidney of the subject to be estimated.

[0046] In S205, the processor 101 estimates the renal disease worsening rate (eGFR slope) based on the clinical information acquired in S201 and the feature amount calculated in S204, using the estimator for estimating the renal disease worsening rate constructed in the operational flow of Fig. 3. The processor 101 stores the estimated renal disease worsening rate in the storage unit 103 in association with information identifying the subject to be estimated. This allows the estimation device 100 to estimate the renal disease worsening rate (eGFR slope) using MR images of the subject to be estimated, etc. Furthermore, if some or all of the clinical information has not been acquired in S201, the processor 101 estimates the renal disease worsening rate (eGFR slope) without using the clinical information that has not been acquired.

[0047] This allows the estimation device 100 to estimate the worsening rate of renal disease using an MR image of a region including the kidney of the subject and an estimator that estimates the worsening rate of renal disease.

[0048] Here, the kidney is the subject, but the present invention can be similarly applied to other organs.

[0049] (Performance of estimators in estimating the rate of progression of kidney disease) Here, we investigated the effectiveness of estimators using the following training data for 164 actual subjects: (1) clinical information (urinary protein content), (2) mean pixel values ​​of kidney images, (3) histograms of pixel values ​​(frequency of occurrence) for each layer of kidney images, (4) (2) and (3), and (5) (1), (2), and (3). Each estimator was constructed based on the above. The correlation coefficients between the eGFR slope estimated by each estimator and the actual eGFR slope were calculated. The correlation coefficients were: (1) 0.56, (2) 0.61, (3) 0.79, (4) 0.85, and (5) 0.87. The higher the correlation coefficient, the better the estimator's performance. Generally, a correlation coefficient of 0.70 or higher is considered useful. From these results, the performance of the estimator is improved when features based on kidney images (MR images) are used as training data, compared to when non-image clinical information (urinary protein content) is used as training data. It is also useful to use histograms of pixel values ​​for each layer of kidney images ((3), (4), (5)). Furthermore, the performance of the estimator is further improved when clinical information and the average pixel values ​​of kidney images are added to the training data in addition to the histograms of pixel values ​​for each layer of kidney images.

[0050] (Actions and Effects of the Embodiments) The estimation device 100 acquires MR images including a subject's kidneys, the subject's clinical information, and the subject's eGFR slope (the rate of deterioration of renal disease). The estimation device 100 extracts kidney images from the acquired MR images. The estimation device 100 performs non-rigid transformation to convert the kidney shape of the extracted kidney image into a standard shape and then calculates feature quantities based on the kidney image. For example, the estimation device 100 divides the kidney into multiple layers from the cortex to the medulla and calculates the occurrence frequency of pixel values ​​in each layer as feature quantities. The estimation device 100 uses multiple datasets including at least the MR image-based feature quantities and the eGFR slope as training data to construct an estimator that estimates the eGFR slope from at least the MR image-based feature quantities. The estimation device 100 can construct an estimator with higher performance by using the MR image-based feature quantities. Furthermore, non-rigid transformation of the subject's kidney shape into a standard shape results in a uniform kidney shape, thereby improving the performance of the estimator. Furthermore, the estimation device 100 can calculate feature amounts corresponding to the internal parts of the kidney by dividing the kidney image into multiple layers (e.g., 12 layers) from the cortex to the medulla. The performance of the estimator is improved by using feature amounts corresponding to the internal parts of the kidney.

[0051] The estimation device 100 acquires MR images including the kidneys of a subject to be estimated, and calculates feature quantities based on the MR images. The estimation device 100 estimates the worsening rate of renal disease in the subject to be estimated using the calculated feature quantities and the constructed estimator. The estimation device 100 can more accurately estimate the worsening rate of renal disease by using the feature quantities based on the MR images including the kidneys.

[0052] <Computer-readable recording medium> A program that causes a computer or other machine or device (hereinafter referred to as a computer, etc.) to realize any of the above functions can be recorded on a computer-readable recording medium. Then, by having the computer, etc. read and execute the program from this recording medium, the function can be provided.

[0053] Here, a computer-readable recording medium refers to a recording medium that stores information such as data and programs electrically, magnetically, optically, mechanically, or chemically and can be read by a computer, etc. Such a recording medium may be provided with elements that constitute a computer, such as a CPU and memory, and the CPU may be made to execute the program.

[0054] Among such recording media, those that can be removed from a computer or the like include, for example, flexible disks, magneto-optical disks, CD-ROMs, CD-R / Ws, DVDs, DATs, 8mm tapes, memory cards, and the like.

[0055] Furthermore, examples of recording media fixed to computers include hard disks and ROMs.

[0056] (others) Although the embodiments of the present invention have been described above, these are merely examples, and the present invention is not limited to these. Various modifications based on the knowledge of those skilled in the art, such as combinations of each configuration, are possible as long as they do not deviate from the spirit of the claims. [Explanation of symbols]

[0057] 100 Estimator 101 processors 102 memory 103 Storage section 104 Input section 105 Output section 106 Communication control unit

Claims

1. acquiring training MR (Magnetic Resonance) images of a region of a subject including the kidneys, generating kidney images by extracting kidney portions included in the training MR images, converting the kidney images into standard kidney images by non-rigid transformation, and calculating training features based on the converted kidney images; a processor that constructs an estimator that estimates the worsening rate of the renal disease of the subject by machine learning using, as training data, a dataset including the calculated learning features and the worsening rate of the renal disease of the subject; An estimation device comprising:

2. The processor calculates the learning features as an appearance frequency of pixel values ​​of pixels included in each layer obtained by dividing the converted kidney image into a plurality of layers. The estimation device according to claim 1 .

3. The processor: acquiring an MR image of a region including a kidney of a subject to be estimated, generating an image of the kidney by extracting a portion of the kidney included in the MR image, converting the image of the kidney into a standard image of the kidney by non-rigid transformation, and calculating feature amounts based on the converted image of the kidney; using the estimator to estimate a worsening rate of a renal disease of the subject to be estimated for the feature amount; The estimation device according to claim 1 or 2.

4. The computer acquiring training MR images of a region of a subject including the kidneys, generating kidney images by extracting kidney portions included in the training MR images, converting the kidney images into standard kidney images by non-rigid transformation, and calculating training features based on the converted kidney images; An estimator is constructed to estimate the rate of deterioration of the subject's renal disease by machine learning using a dataset including the calculated learning features and the rate of deterioration of the subject's renal disease as training data. An estimation method to do this.

5. The computer acquiring training MR images of a region of a subject including the kidneys, generating kidney images by extracting kidney portions included in the training MR images, converting the kidney images into standard kidney images by non-rigid transformation, and calculating training features based on the converted kidney images; An estimator is constructed to estimate the rate of deterioration of the subject's renal disease by machine learning using a dataset including the calculated learning features and the rate of deterioration of the subject's renal disease as training data. A guessing program to do that.

Images