Medical information processing device and medical information processing program

Abstract

To acquire learning data in accordance with clinical practice.SOLUTION: A medical information processing apparatus includes an acquisition section and a generation section. The acquisition section acquires a plurality of pieces of medical examination data of a subject. The generation section acquires prediction data by performing prediction processing based on medical knowledge (knowledge) for at least one piece of medical examination data included in the plurality of pieces of medical examination data and generates learning data including the medical examination data included in the plurality of pieces of medical examination data and the prediction data.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The embodiments disclosed in this specification and the drawings relate to a medical information processing device and a medical information processing program. [Background technology]

[0002] Conventionally, machine learning has been performed using medical images collected by medical image diagnostic devices and test results obtained by various tests as learning data. Techniques for increasing the amount of learning data used in such machine learning are also known. For example, techniques for increasing the amount of learning data are known, which include adding noise to medical images, adjusting the contrast or brightness of medical images, and inverting or rotating medical images. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Special Publication No. 2016-532459 [Non-patent literature]

[0004] [Non-Patent Document 1] Christopher M. Bishop, "Pattern recognition and machine learning", (USA), 1st edition, Springer, 2006, pp. 225-290 Summary of the Invention [Problem to be solved by the invention]

[0005] One of the problems that the embodiments disclosed in this specification and the drawings aim to solve is to obtain learning data in line with clinical practice. However, the problems that the embodiments disclosed in this specification and the drawings aim to solve are not limited to the above problem. Problems corresponding to the effects of each configuration shown in the embodiments described below can also be positioned as other problems. [Means for solving the problem]

[0006] A medical information processing device according to an embodiment includes an acquisition unit and a generation unit. The acquisition unit acquires a plurality of medical data of a subject. The generation unit acquires prediction data by performing a prediction process based on medical knowledge on at least one of the plurality of medical data, and generates learning data including the medical data included in the plurality of medical data and the prediction data. [Brief description of the drawings]

[0007] [Figure 1] FIG. 1 is a block diagram showing an example of the configuration of a medical image processing apparatus according to the first embodiment. [Diagram 2] FIG. 2 is a diagram for explaining the generation of a medical data set by the generation function according to this embodiment. [Figure 3A] FIG. 3A is a diagram for explaining an example of a trained model according to the first embodiment. [Figure 3B] FIG. 3B is a diagram for explaining an example of learning data according to the first embodiment. [Figure 4] FIG. 4 is a diagram for explaining a process by the generating function according to the first embodiment. [Diagram 5] FIG. 5 is a flowchart showing a processing procedure by the medical image processing apparatus according to the first embodiment. [Figure 6] FIG. 6 is a diagram for explaining a process by the generating function according to the second embodiment. [Figure 7] FIG. 7 is a flowchart showing a processing procedure by the medical image processing apparatus according to the second embodiment. [Figure 8] FIG. 8 is a diagram for explaining a process by the generating function according to the third embodiment. [Figure 9] FIG. 9 is a diagram for explaining an example of a simulation according to another embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0008] Hereinafter, the medical information processing device and the medical information processing program according to the present application will be described in detail with reference to the drawings. Note that the medical information processing device and the medical information processing program according to the present application are not limited to the following embodiments. In addition, the embodiments can be combined with other embodiments or conventional techniques as long as there is no contradiction in the processing content.

[0009] (First embodiment) FIG. 1 is a block diagram showing an example of the configuration of a medical information processing device 4 according to the first embodiment. For example, the medical information processing device 4 according to the first embodiment is included in a medical information processing system 100 that is communicatively connected to a medical image diagnostic device 1, a terminal device 2, and a medical information storage device 3 via a network 200 as shown in FIG. 1. Here, each device included in the medical information processing system 100 is in a state in which it can directly or indirectly communicate with each other, for example, via an in-hospital LAN (Local Area Network) installed in a hospital. Note that the medical information processing system 100 shown in FIG. 1 may be communicatively connected to devices other than those shown in the figure.

[0010] For example, the medical information processing system 100 includes various systems such as a Hospital Information System (HIS), a Radiology Information System (RIS), a diagnostic report system, a medical image processing system (PACS: Picture Archiving and Communication System), and a Laboratory Information System (LIS: Laboratory Information System).

[0011] The medical image diagnostic device 1 captures an image of a subject and collects medical images. Then, the medical image diagnostic device 1 transmits the collected medical images to a terminal device 2, a medical information storage device 3, and a medical information processing device 4. For example, the medical image diagnostic device 1 is an X-ray diagnostic device, an X-ray CT (Computed Tomography) device, an MRI (Magnetic Resonance Imaging) device, an ultrasonic diagnostic device, a SPECT (Single Photon Emission Computed Tomography) device, a PET (Positron Emission computed Tomography) device, or the like.

[0012] The terminal device 2 is a device operated by a doctor or a medical technician working in a hospital. For example, the terminal device 2 is realized by a personal computer (PC), a tablet PC, a personal digital assistant (PDA), a mobile phone, or the like. The terminal device 2 displays medical images received from the medical image diagnostic device 1 or the medical information storage device 3 on its own display, and accepts various operations via its own input interface. For example, the terminal device 2 accepts input operations such as various test results and diagnostic information, and transmits them to the medical information storage device 3 or the medical information processing device 4.

[0013] The medical information storage device 3 stores various medical information in the medical information processing system 100. Specifically, the medical information storage device 3 stores medical information including medical images received from the medical image diagnostic device 1, various test results and diagnostic information received via the terminal device 2, and various information received from other devices connected to the medical information processing system 100. Here, the medical information storage device 3 stores each piece of information included in the medical information by linking it to a patient ID or the like. For example, the medical information storage device 3 is realized by a computer device such as a DB (Database) server, and stores the medical information in a semiconductor memory element such as a RAM (Random Access Memory) or a flash memory, or a storage circuit such as a hard disk or an optical disk. Note that, although one medical information storage device 3 is illustrated in FIG. 1, the embodiment is not limited to this, and a plurality of DB servers may cooperate to function as the medical information storage device 3.

[0014] The medical information processing device 4 acquires various information from the medical image diagnostic device 1, the terminal device 2, and the medical information storage device 3, and performs various information processing using the acquired information. For example, the medical information processing device 4 is realized by computer equipment such as a server, a workstation, a personal computer, or a tablet terminal.

[0015] As shown in FIG. 1, the medical information processing device 4 has a communication interface 41, a memory circuitry 42, an input interface 43, a display 44, and a processing circuitry 45.

[0016] The communication interface 41 is connected to the processing circuitry 45, and controls communication between each device in the medical information processing system 100. Specifically, the communication interface 41 receives various information from each device, and outputs the received information to the processing circuitry 45. For example, the communication interface 41 is realized by a network card, a network adapter, a NIC (Network Interface Controller), or the like.

[0017] The memory circuitry 42 is connected to the processing circuitry 45 and stores various data. Specifically, the memory circuitry 42 stores various information received from the medical image diagnostic apparatus 1, the terminal device 2, and the medical information storage device 3, information input via the input interface 43, processing results of the medical information processing device 4, and the like. For example, the memory circuitry 42 stores medical data 421, medical knowledge 422, and a trained model 423, as shown in FIG. 1.

[0018] The medical data 421 includes medical images, analysis information of the medical images, examination results, medical information, etc. received from the medical image diagnostic apparatus 1, the terminal device 2, and the medical information storage device 3. The medical data 421 is acquired from the medical image diagnostic apparatus 1, the terminal device 2, and the medical information storage device 3 via the communication interface 41, and stored in the memory circuitry 42. The medical data 421 is also acquired by processing of the processing circuitry 45, and stored in the memory circuitry 42.

[0019] The medical knowledge 422 includes information such as findings based on the results of clinical research, various guidelines, and simulations based on them. For example, the medical knowledge 422 includes mutual relationships between multiple medical data, causal relationships between diseases, and simulations for medical images constructed based on them. The medical knowledge 422 is stored in the memory circuitry 42 based on an input operation via the input interface 43.

[0020] The trained model 423 is generated by machine learning using the medical data 421 as training data. Specifically, the trained model 423 is generated by machine learning using a plurality of medical data as training data and stored in the memory circuitry 42. Note that the trained model 423 may be generated by the processing circuitry 45 and stored in the memory circuitry 42, or may be generated by an external device disposed outside the medical information processing system 100 and stored in the memory circuitry 42.

[0021] The memory circuitry 42 also stores various programs for implementing various functions by being read and executed by the processing circuitry 45. For example, the memory circuitry 42 is implemented by a semiconductor memory element such as a random access memory (RAM) or a flash memory, a hard disk, an optical disk, or the like.

[0022] The input interface 43 is connected to the processing circuit 45 and accepts various instructions and information input operations from an operator. Specifically, the input interface 43 converts the input operation accepted from the operator into an electric signal and outputs it to the processing circuit 45. For example, the input interface 43 is realized by a trackball, a switch button, a mouse, a keyboard, a touch pad that performs an input operation by touching the operation surface, a touch screen in which a display screen and a touch pad are integrated, a non-contact input circuit using an optical sensor, and a voice input circuit. Note that in this specification, the input interface 43 is not limited to only those that have physical operation parts such as a mouse and a keyboard. For example, an electric signal processing circuit that receives an electric signal corresponding to an input operation from an external input device provided separately from the device and outputs the electric signal to a control circuit is also included as an example of the input interface 43.

[0023] The display 44 is connected to the processing circuit 45 and displays various information and images. Specifically, the display 44 converts the information and image data sent from the processing circuit 45 into an electric signal for display and outputs it. For example, the display 44 is realized by a liquid crystal monitor, a CRT (Cathode Ray Tube) monitor, a touch panel, etc.

[0024] The processing circuitry 45 controls the operation of the medical information processing device 4 in response to an input operation received from an operator via the input interface 43. For example, the processing circuitry 45 is realized by a processor. As shown in FIG. 1, the processing circuitry 45 executes a control function 451, a generation function 452, and a learning function 453. Here, the control function 451 is an example of an acquisition unit. Also, the generation function 452 is an example of a generation unit. Also, the learning function 453 is an example of a learning unit.

[0025] The control function 451 controls various processes in the medical information processing device 4. The generation function 452 generates learning data based on the medical data 421 and the medical knowledge 422. The learning function 453 generates a trained model 423 using the learning data generated by the generation function 452, and stores the trained model 423 in the memory circuitry 42. The processing of each of these functions will be described in detail later.

[0026] The configuration of the medical information processing device 4 according to the first embodiment has been described above. With such a configuration, the medical information processing device 4 makes it possible to obtain learning data that is clinically appropriate. As described above, in collecting learning data used in machine learning, a technique is known in which the learning data is increased by processing acquired medical images. However, when data is generated using such a method, there is a possibility that data that is not appropriate in relation to other data used in machine learning is generated.

[0027] In other words, a combination of multiple data used in machine learning may be generated that is not clinically valid, and may end up being used in machine learning as a valid combination. Therefore, in the present application, training data having combinations that are in line with clinical practice is obtained by generating training data based on clinical knowledge.

[0028] The medical image processing device 4 according to this embodiment will be described in detail below.

[0029] The control function 451 controls the execution of processes according to various requests input via the input interface 43. For example, the control function 451 controls the transmission and reception of medical images, examination results, etc. via the communication interface 41, the storage of various information in the memory circuitry 42, the display of information (for example, medical images, processing results by each function, etc.) on the display 44, etc.

[0030] For example, the control function 451 acquires medical data from the medical image diagnostic device 1, the terminal device 2, and the medical information storage device 3, and stores the data in the memory circuitry 42. In addition, for example, the control function 451 controls the display 44 to display a GUI for executing processing on the medical data, and the processing results of each function.

[0031] The generating function 452 obtains prediction data by performing a prediction process based on medical knowledge on the medical data, and generates learning data including the medical data and the prediction data. Specifically, the generating function 452 reads out a plurality of pieces of medical data from the medical data 421 stored in the storage circuitry 42, and obtains new medical data by performing a prediction process based on medical knowledge on at least one of the read out plurality of pieces of medical data. Then, the generating function 452 generates learning data from at least one of the read out plurality of pieces of medical data and the obtained new medical data. Hereinafter, the learning data generated by the generating function 452 is also referred to as a medical data set.

[0032] As described above, the generating function 452 performs a prediction process based on medical knowledge on the medical data to obtain new medical data and generate learning data. Here, in the present application, various data can be obtained as the new medical data.

[0033] Fig. 2 is a diagram for explaining the generation of a medical dataset by the generation function 452 according to this embodiment. Here, the medical dataset shown in Fig. 2 indicates a set of multiple medical data used as learning data. As shown in Fig. 2, the generation function 452 applies the medical dataset to a simulation model based on clinical knowledge to generate a clinically appropriate medical dataset ("medical dataset (clinically appropriate)") or a clinically inappropriate medical dataset ("medical dataset (clinically inappropriate)").

[0034] For example, as shown in FIG. 2, the generation function 452 generates a clinically appropriate medical dataset by applying a "medical dataset (partially missing)" to the simulation model. Here, the "medical dataset (partially missing)" refers to a medical dataset in which some of the multiple medical data used as learning data is missing. That is, the generation function 452 generates learning data by acquiring the missing medical data from the multiple medical data used as learning data through a simulation based on medical knowledge. Here, since the medical data acquired by the generation function 452 has been acquired through a simulation based on medical knowledge, it becomes clinically appropriate data in combination with other medical data.

[0035] Also, for example, the generating function 452 generates a clinically appropriate medical dataset by applying a "medical dataset (partially modified)" to the simulation model as shown in FIG. 2. Here, the "medical dataset (partially modified)" is a medical dataset that includes all (without deficiencies) of the multiple medical data used as learning data, and indicates a medical dataset that is the target of data modification based on medical knowledge. That is, the generating function 452 generates learning data by obtaining a new combination from a combination of multiple medical data already existing as learning data through a simulation based on medical knowledge. Here, since the medical data obtained by the generating function 452 is obtained through a simulation based on medical knowledge, the combination becomes clinically appropriate data.

[0036] Also, for example, the generation function 452 generates a clinically inappropriate clinical data set by applying a "clinical data set" to a simulation model with partially modified knowledge, as shown in FIG. 2. Here, the "clinical data set" is a clinical data set that includes all (without deficiencies) of the multiple medical data used as learning data. Also, partial modification of the clinical knowledge modifies part of the simulation model so that it does not follow physical laws or physiological laws. In other words, the generation function 452 generates clinically inappropriate learning data by obtaining new combinations from combinations of multiple medical data already available as learning data through a simulation that is not based on clinical knowledge.

[0037] As described above, the generation function 452 can generate learning data in line with clinical practice by simulation based on medical knowledge. Therefore, the learning function 453 can generate a trained model with high clinical accuracy by performing machine learning using the learning data generated by the generation function 452.

[0038] In the first embodiment, an example of generating learning data by acquiring missing medical data through a simulation based on medical knowledge will be described below. An example of generating a clinically appropriate medical data set by modifying a medical data set and an example of generating a clinically inappropriate medical data set by modifying a part of the medical knowledge will be described in the second and third embodiments, respectively.

[0039] In the following embodiment, a case where learning data for generating a trained model that outputs a value of FFR (Fractional Flow Reserve) in response to input of medical data will be described as an example. That is, the trained model generated by machine learning using the learning data generated by the generation function 452 outputs an FFR value in response to input of medical data, as shown in FIG. 3A. This makes it possible to estimate the value of FFR only from medical data without measuring FFR with a cardiac catheter or CT imaging of the coronary artery. Note that the example described in this embodiment is merely an example, and various other trained models can be generated. Also, FIG. 3A is a diagram for explaining an example of a trained model according to the first embodiment.

[0040] When generating such a trained model, medical data and FFR values ​​are used as training data. FIG. 3B is a diagram for explaining an example of training data according to the first embodiment. For example, in this embodiment, as shown in FIG. 3B, "chief complaint (chest pain, etc.)", "blood test (cholesterol, etc.)", "electrocardiogram (ST depression, etc.)", "echocardiogram (coronary artery blood flow waveform, etc.)", and "FFR value" are used as training data. That is, by learning the relationship between "chief complaint (chest pain, etc.)", "blood test (cholesterol, etc.)", "electrocardiogram (ST depression, etc.)", and "echocardiogram (coronary artery blood flow waveform, etc.)" and "FFR value", a trained model functioned to output an FFR value in response to input of medical data is generated.

[0041] When there is a lack of data in the learning data shown in Fig. 3B, the generation function 452 acquires predicted data (missing data) other than the data types included in the medical data by performing a prediction process (simulation) in accordance with the medical knowledge. For example, the "chief complaint", "blood test", "electrocardiogram", and "echocardiogram" in Fig. 3B are medical data collected for many subjects. On the other hand, the FFR value measured by a pressure wire during a cardiac catheterization is rarely collected from subjects with mild stenosis, and is data that is easily insufficient.

[0042] In this way, when the FFR value is missing in the learning data, the generating function 452 acquires the FFR value by a simulation based on the medical knowledge. Fig. 4 is a diagram for explaining the processing by the generating function 452 according to the first embodiment. For example, as shown in Fig. 4, the generating function 452 acquires the FFR value by a blood flow analysis simulation using CT image data collected by a coronary artery CT.

[0043] In this case, the generating function 452 acquires coronary angiography CT image data of the subject included in the medical data 421. Then, the generating function 452 acquires an FFR value in the coronary artery by performing a fluid analysis on the coronary angiography CT image data.

[0044] When performing fluid analysis based on coronary angiography CT image data to obtain an FFR value, the generation function 452 reads out coronary angiography CT image data of multiple time phases collected over time from the medical data 421, and extracts time-series vascular shape data by performing image processing on the read out coronary angiography CT image data of multiple time phases.

[0045] Here, the generating function 452 sets a target region for calculating an index value in a blood vessel region included in the coronary angiography CT image data. Specifically, the generating function 452 sets the target region in the blood vessel region by an instruction from an operator via the input interface 43 or by image processing. Then, the generating function 452 extracts, from the coronary angiography CT image data, for example, the center line of the blood vessel (coordinate information of the center line), the cross-sectional area of ​​the blood vessel and the lumen in a cross section perpendicular to the center line, the distance from the center line to the inner wall and the distance from the center line to the outer wall in the cylindrical direction in a cross section perpendicular to the center line, etc.

[0046] Furthermore, the generating function 452 sets analysis conditions for the fluid analysis. Specifically, the generating function 452 sets the physical property values ​​of blood, the conditions for the iterative calculation, the initial values ​​of the analysis, and the like, as the analysis conditions. For example, the generating function 452 sets the viscosity and density of blood as the physical property values ​​of blood. Furthermore, the generating function 452 sets the maximum number of iterations in the iterative calculation, the relaxation coefficient, the allowable value of the residual, and the like, as the conditions for the iterative calculation. Furthermore, the generating function 452 sets the initial values ​​of the blood flow rate, pressure, fluid resistance, and pressure boundary, and the like, as the initial values ​​of the analysis. Note that the various values ​​used by the generating function 452 may be built into the system in advance, or may be defined interactively by the operator.

[0047] Then, the generating function 452 calculates an index value related to blood flow in a blood vessel by fluid analysis using the coronary angiography CT image data. Specifically, the generating function 452 executes fluid analysis using blood vessel shape data and analysis conditions, and calculates an index value related to blood flow in a target region of the blood vessel. For example, the generating function 452 calculates index values ​​such as pressure, blood flow rate, blood flow velocity, vector, and shear stress for each predetermined position of the blood vessel based on blood vessel shape data such as the contour of the lumen and outer wall of the blood vessel, the cross-sectional area and the core line of the blood vessel, and setting conditions such as the physical properties of blood, the conditions of the iterative calculation, and the initial values ​​of the analysis. Then, the generating function 452 can obtain an FFR value for each position based on the pressure for each position of the coronary artery. Note that applications related to fluid analysis are included in the medical knowledge 422 and stored in the memory circuitry 42.

[0048] The generating function 452 can also obtain other data from the results of the above-mentioned fluid analysis. For example, for a subject for which echocardiography has not been performed and data such as coronary artery blood flow waveforms have not been obtained, the generating function 452 can generate learning data by estimating data such as coronary artery blood flow waveforms using the results of the fluid analysis.

[0049] When the learning data is generated as described above, the learning function 453 generates a trained model by machine learning using the learning data generated by the generation function 452. Here, in the present embodiment, a case will be described in which the processing circuitry 45 of the medical information processing device 4 executes the learning function 453 to generate the trained model 423, but the embodiment is not limited thereto. For example, a case may also be where another information processing device executes the process described below to generate the trained model 423, and the generated trained model 423 is stored in the storage circuitry 42.

[0050] For example, the learning function 453 performs machine learning by inputting "chief complaint (chest pain, etc.)", "blood test (cholesterol, etc.)", "electrocardiogram (ST depression, etc.)", "echocardiogram (coronary artery blood flow waveform, etc.)", and "FFR value" as learning data into a machine learning engine, and generates a trained model that is functionalized to output an FFR value according to the input of medical data. Here, as the machine learning engine, for example, a neural network described in the well-known non-patent document "Pattern recognition and machine learning" by Christopher M. Bishop, 1st edition, Springer, 2006, pp. 225-290, can be applied.

[0051] In addition, the machine learning engine may use various algorithms such as the neural network described above, for example, deep learning, logistic regression analysis, nonlinear discriminant analysis, support vector machine (SVM), random forest, and naive Bayes.

[0052] As a result of such machine learning, the learning function 453 generates a trained model 423 that outputs an FFR value for the inputs of "chief complaint (chest pain, etc.)", "blood test (cholesterol, etc.)", "electrocardiogram (ST depression, etc.)", and "echocardiogram (coronary artery blood flow waveform, etc.)" input to the trained algorithm. Then, the learning function 453 stores the generated trained model 423 in the memory circuitry 42. At this time, if a previously created trained model 423 has already been stored in the memory circuitry 42, the learning function 453 can also replace the stored trained model 423 with the newly created trained model 423.

[0053] Next, the procedure of processing by the medical information processing device 4 will be described. Fig. 5 is a flowchart showing the procedure of processing by the medical information processing device 4 according to the first embodiment. Here, steps S101 and S102 in Fig. 5 are steps realized by the processing circuitry 45 calling a program corresponding to the generating function 452 from the storage circuitry 42 and executing it. Also, steps S103 and S104 in Fig. 5 are steps realized by the processing circuitry 45 calling a program corresponding to the learning function 453 from the storage circuitry 42 and executing it.

[0054] 5, in the medical information processing device 4, the processing circuitry 45 first acquires a plurality of medical data of the subject, and determines whether the acquired medical data set is insufficient (step S101). If the medical data is insufficient (step S101, YES), the processing circuitry 45 acquires the insufficient data by simulation based on medical knowledge (step S102). If the medical data is not insufficient, the processing circuitry 45 omits the process of step S102 (step S101, NO).

[0055] Then, the processing circuitry 45 generates a trained model by machine learning using the medical data set as training data (step S103), and stores the generated trained model in the memory circuitry 42 (step S104).

[0056] As described above, according to the first embodiment, the control function 451 acquires multiple medical data of the subject. The generation function 452 acquires prediction data by performing prediction processing based on medical knowledge on at least one medical data included in the multiple medical data, and generates learning data including the medical data included in the multiple medical data and the prediction data. Therefore, the medical information processing device 4 according to the first embodiment can generate medical data based on medical knowledge as learning data, and makes it possible to acquire learning data in line with clinical practice.

[0057] According to the first embodiment, the generation function 452 obtains prediction data other than the data type included in the medical data by prediction processing according to the medical knowledge. Therefore, the medical information processing device 4 according to the first embodiment makes it possible to obtain clinically appropriate medical data and use it as learning data even when there is a shortage in the multiple medical data used as learning data.

[0058] According to the first embodiment, the learning function 453 generates a trained model by machine learning using learning data. Therefore, the medical information processing device 4 according to the first embodiment makes it possible to obtain a trained model with high clinical accuracy.

[0059] Second embodiment In the above-mentioned first embodiment, a case where missing medical data is obtained by a simulation based on medical knowledge to generate learning data is described. In the second embodiment, a case where a clinically appropriate medical data set is generated by modifying a medical data set is described. In the following, differences from the above-mentioned embodiments will be mainly described, and detailed description of commonalities with the contents already described will be omitted. In addition, each embodiment described below may be implemented individually or in appropriate combination.

[0060] The generating function 452 according to the second embodiment obtains prediction data obtained by modifying data included in medical data through prediction processing according to medical knowledge. Specifically, the generating function 452 generates new learning data by modifying complete learning data according to medical knowledge.

[0061] Fig. 6 is a diagram for explaining processing by the generating function 452 according to the second embodiment. Fig. 6 shows processing on the learning data shown in Fig. 3B. That is, processing on a medical data set including "chief complaint", "blood test", "electrocardiogram", "echocardiogram" and "FFR value" is shown. For example, as shown in Fig. 6, the generating function 452 generates a new medical data set by partially modifying CT image data collected by a coronary artery CT.

[0062] As an example, the generating function 452 changes part of the structure of the coronary artery, myocardium, and aorta contained in the CT image data. The generating function 452 then performs fluid analysis using the changed data to obtain data such as FFR values ​​and coronary artery blood flow waveforms, thereby generating a new medical data set. When changing the structure, for example, changes based on age or changes along with the progression of stenosis are used. As an example, the generating function 452 changes the structure of the coronary artery based on changes in blood vessel shape and changes in the flexibility of the blood vessel wall that occur with age. The generating function 452 also changes the structure of the coronary artery along with the progression of stenosis.

[0063] Information on age-related changes in blood vessel shape, changes in the flexibility of blood vessel walls, and the progression of stenosis is included in the medical knowledge 422 and stored in the memory circuitry 42. That is, the generation function 452 reads out this information from the memory circuitry 42 and changes the structure based on the read out information.

[0064] Furthermore, for example, the generating function 452 can generate a new medical data set by changing parameters of the analysis conditions used in the fluid analysis. As an example, the generating function 452 changes physiological parameters such as blood density and viscosity included in the analysis conditions of the fluid analysis. The generating function 452 then performs fluid analysis with the changed parameters and obtains data such as FFR values ​​and coronary artery blood flow waveforms to generate a new medical data set. When changing the blood density and viscosity, the generating function 452 changes the blood density and viscosity within a physiologically possible range.

[0065] As described above, the generating function 452 according to the second embodiment generates a new medical data set by modifying an existing medical data set based on medical knowledge. Here, the generating function 452 also modifies other medical data related to the modified content based on medical knowledge. For example, when changing blood density or blood viscosity included in the parameters of the fluid analysis, the generating function 452 modifies information such as cholesterol included in the medical data "blood test" based on medical knowledge. In other words, the generating function 452 modifies the entire medical data set by performing a simulation according to the modification.

[0066] Next, the procedure of processing by the medical information processing device 4 will be described. FIG. 7 is a flowchart showing the procedure of processing by the medical information processing device 4 according to the second embodiment. Here, FIG. 7 shows a case where processing according to this embodiment is executed in addition to processing for acquiring missing data by simulation. Steps S201 to S204 in FIG. 7 are steps realized by the processing circuitry 45 calling a program corresponding to the generating function 452 from the storage circuitry 42 and executing it. Also, steps S205 and S206 in FIG. 7 are steps realized by the processing circuitry 45 calling a program corresponding to the learning function 453 from the storage circuitry 42 and executing it.

[0067] As shown in Fig. 7, in the medical information processing device 4, the processing circuitry 45 first acquires a plurality of medical data of the subject, and determines whether the acquired medical data set is insufficient (step S201). If the medical data is insufficient (step S201, YES), the processing circuitry 45 acquires the insufficient data by simulation based on medical knowledge (step S202). If the medical data is not insufficient, the processing circuitry 45 omits the process of step S202 (step S201, NO).

[0068] Then, the processing circuitry 45 modifies a part of the medical data set (step S203), and modifies the medical data set by a simulation according to the modification (step S204).

[0069] Thereafter, the processing circuitry 45 generates a trained model by machine learning using the modified clinical data set as training data (step S205), and stores the generated trained model in the memory circuitry 42 (step S206).

[0070] As described above, according to the second embodiment, the generating function 452 obtains prediction data by changing data included in multiple medical data through prediction processing according to medical knowledge. Therefore, the medical information processing device 4 according to the second embodiment makes it possible to easily increase the amount of learning data using clinically relevant medical data.

[0071] (Third embodiment) In the third embodiment, a case where a clinically inappropriate clinical data set is generated by modifying a part of the clinical knowledge will be described. In the following, differences from the above-mentioned embodiments will be mainly described, and detailed descriptions of commonalities with the contents already described will be omitted. In addition, each embodiment described below may be implemented individually or in appropriate combination.

[0072] The generating function 452 according to the third embodiment obtains prediction data other than the data type included in the plurality of medical data by a prediction process contrary to the medical knowledge. The generating function 452 according to the third embodiment also obtains prediction data obtained by changing the data included in the plurality of medical data by a prediction process contrary to the medical knowledge. Specifically, the generating function 452 performs a simulation contrary to the medical knowledge by making modifications to a simulation based on the medical knowledge that do not comply with physical laws or physiological laws. The generating function 452 then generates a medical dataset including the medical data obtained by the simulation contrary to the medical knowledge as clinically inappropriate learning data.

[0073] Fig. 8 is a diagram for explaining processing by the generating function 452 according to the third embodiment. Fig. 8 shows processing for the learning data shown in Fig. 3B. That is, an example is shown in which "chief complaint", "blood test", "electrocardiogram", "echocardiogram" and "FFR value" are used as learning data.

[0074] For example, the generating function 452 generates a new clinical data set that is not clinically relevant by making partial modifications to the blood flow analysis simulation executed on the CT image data collected by the coronary artery CT as shown in Fig. 8. As an example, the generating function 452 makes modifications to a part of the mathematical model of the blood flow analysis simulation that do not follow the laws of physics and physiology. As a result, the FFR value obtained from the modified blood flow analysis simulation is not clinically relevant and is erroneous data.

[0075] Here, the above-mentioned clinically inappropriate medical data set may be generated from a medical data set in which some medical data is missing, or may be generated from a medical data set in which all medical data is complete. That is, the generating function 452 generates clinically inappropriate learning data by modifying the simulation model before executing the simulation in the process described in the first and second embodiments.

[0076] The learning function 453 can increase the robustness of the trained model by using the clinically invalid training data generated as described above for machine learning.

[0077] As described above, according to the third embodiment, the generation function 452 obtains prediction data other than the data type included in the multiple medical data by prediction processing contrary to the medical knowledge. In addition, the generation function 452 obtains prediction data in which data included in the multiple medical data is changed by prediction processing contrary to the medical knowledge. Therefore, the medical information processing device 4 according to the third embodiment can intentionally generate clinically inappropriate learning data, and can improve the robustness of the trained model by performing machine learning using the learning data.

[0078] (Other embodiments) Although the first to third embodiments have been described above, the present invention may be embodied in various different forms other than the above-mentioned first to third embodiments.

[0079] In the above-described embodiment, a blood flow analysis simulation is used as a simulation based on clinical knowledge. However, the embodiment is not limited to this, and any simulation based on clinical knowledge may be used.

[0080] For example, the generating function 452 can acquire medical data by using an onset model that models the onset of a disease. Fig. 9 is a diagram for explaining an example of a simulation according to another embodiment. Here, Fig. 9 shows the relationship between each factor leading to the onset of CHD (Coronary Heart Disease).

[0081] For example, the generating function 452 can acquire missing medical data by a simulation using the CHD onset model shown in Fig. 9. As an example, when cholesterol values ​​are missing, the generating function 452 generates learning data including cholesterol values ​​by estimating LDL cholesterol values ​​from the subject's BMI (Body Mass Index) and the level of CHD.

[0082] In addition, in each of the above-mentioned embodiments, examples have been described in which the acquisition unit, generation unit, and learning unit in this specification are respectively realized by the control function 451, generation function 452, and learning function 453 of the processing circuit 45, but the embodiments are not limited to this. For example, the acquisition unit, generation unit, and learning unit in this specification may be realized by the control function 451, generation function 452, and learning function 453 described in the embodiments, or may be realized by hardware alone or a combination of hardware and software.

[0083] As described above, when the processing circuit 45 is realized by a processor, each processing function possessed by the processing circuit 45 is stored in the storage circuit 42 in the form of a program executable by a computer. Then, the processing circuit 45 realizes the function corresponding to each program by reading each program from the storage circuit 42 and executing it. In other words, the processing circuit 45 in a state in which each program is read has each function shown in the processing circuit 45 of FIG. 1. Note that, in FIG. 1, it has been described that each processing function is realized by a single processor, but it is also possible to configure a processing circuit by combining multiple independent processors and realize the function by each processor executing a program. Also, the processing function possessed by the processing circuit 45 may be realized by being appropriately distributed or integrated in a single or multiple processing circuits. Also, in the example shown in FIG. 1, it has been described that a single storage circuit 42 stores a program corresponding to each processing function, but it is also possible to configure a configuration in which multiple storage circuits are distributed and arranged, and the processing circuit reads out the corresponding program from each storage circuit.

[0084] The term "processor" used in the above description means a circuit such as a central processing unit (CPU), a graphics processing unit (GPU), an application specific integrated circuit (ASIC), a programmable logic device (e.g., a simple programmable logic device (SPLD), a complex programmable logic device (CPLD), and a field programmable gate array (FPGA)). The processor realizes its function by reading and executing a program stored in the memory circuit 42. Instead of storing a program in the memory circuit 42, the processor may be configured to directly incorporate the program into the circuit. In this case, the processor realizes its function by reading and executing the program incorporated in the circuit. The processor of this embodiment is not limited to being configured as a single circuit, and may be configured as a single processor by combining multiple independent circuits to realize its function.

[0085] Here, the program (medical information processing program) executed by the processor is provided by being pre-installed in a ROM (Read Only Memory) or a storage circuit. The program may be provided by being recorded in a computer-readable storage medium such as a CD (Compact Disk)-ROM, a FD (Flexible Disk), a CD-R (Recordable), or a DVD (Digital Versatile Disk) in a format that can be installed in these devices or in a format that can be executed. The program may also be provided or distributed by being stored on a computer connected to a network such as the Internet and downloaded via the network. For example, the program is composed of modules including each of the above-mentioned functional units. As actual hardware, a CPU reads out and executes the program from a storage medium such as a ROM, and each module is loaded onto a main storage device and generated on the main storage device.

[0086] According to at least one of the embodiments described above, learning data in line with clinical practice can be obtained.

[0087] Although some embodiments have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, substitutions, modifications, and combinations of embodiments can be made without departing from the spirit of the invention. These embodiments and their modifications are included in the scope of the invention and its equivalents as described in the claims, as well as in the scope and spirit of the invention. [Explanation of symbols]

[0088] 4 Medical information processing equipment 45 Processing circuit 451 Control Functions 452 Generation function 453 Learning Function

Claims

1. An acquisition unit that acquires a plurality of medical data of a subject; a generation unit that acquires prediction data by performing a prediction process based on medical knowledge on first medical data in first learning data including the plurality of medical data, and generates second learning data including second medical data other than the first medical data in the first learning data and the prediction data; A medical information processing device comprising:

2. The medical information processing device according to claim 1 , wherein the generation unit obtains prediction data other than the data type included in the plurality of medical data by performing prediction processing in accordance with the medical knowledge.

3. The medical information processing device according to claim 1 , wherein the generating unit obtains prediction data other than the data type included in the plurality of medical data by a prediction process contrary to the medical knowledge.

4. The medical information processing device according to claim 1 , wherein the generation unit obtains prediction data by changing data included in the plurality of medical data through a prediction process that contradicts the medical knowledge.

5. The medical information processing device according to any one of claims 1 to 4, further comprising a learning unit that generates a trained model by machine learning using the first training data and the second training data.

6. An acquisition unit that acquires a plurality of medical data of a subject; A generation unit that performs a prediction process based on medical knowledge on at least one medical data included in the plurality of medical data to obtain prediction data, and generates learning data including the medical data included in the plurality of medical data and the prediction data; Equipped with The generation unit obtains prediction data other than a data type included in the plurality of medical data by a prediction process that contradicts the medical knowledge.

7. An acquisition unit that acquires a plurality of medical data of a subject; A generation unit that performs a prediction process based on medical knowledge on at least one medical data included in the plurality of medical data to obtain prediction data, and generates learning data including the medical data included in the plurality of medical data and the prediction data; Equipped with The generation unit obtains prediction data by changing data included in the plurality of medical data through prediction processing that contradicts the medical knowledge.

8. An acquisition function for acquiring a plurality of medical data of a subject; a generation function for acquiring prediction data by performing a prediction process based on medical knowledge on first medical data in first learning data including the plurality of medical data, and generating second learning data including second medical data other than the first medical data in the first learning data and the prediction data; A medical information processing program to realize this on a computer.

9. An acquisition function for acquiring a plurality of medical data of a subject; A generation function of acquiring prediction data by performing a prediction process based on medical knowledge on at least one of the medical data included in the plurality of medical data, and generating learning data including the medical data included in the plurality of medical data and the prediction data; This is realized on a computer. The generation function is a medical information processing program that obtains prediction data other than a data type included in the multiple medical data by performing prediction processing that contradicts the medical knowledge.

10. An acquisition function for acquiring a plurality of medical data of a subject; A generation function of acquiring prediction data by performing a prediction process based on medical knowledge on at least one of the medical data included in the plurality of medical data, and generating learning data including the medical data included in the plurality of medical data and the prediction data; This is realized on a computer. The generation function is a medical information processing program that obtains prediction data by changing data contained in the multiple medical data through prediction processing that contradicts the medical knowledge.

Images