Medical data calculation system and medical data processing method

Abstract

Provided are a medical data calculation system and a medical data processing method for supporting a medical provider. The medical data calculation system comprises: a medical terminal device in a medical institution for receiving input of medical data including medical treatment results, diagnostic results, health checkup results, examination results, or prescription content; a shared storage for storing medical data collected from a medical terminal device group in a plurality of the medical institutions including the medical terminal device; and a plurality of support data output devices each outputting different medical support data through calculation using the medical data stored in the shared storage. The medical terminal device acquires the medical support data from any of the plurality of support data output devices, and outputs the acquired medical support data.

Description

Medical Data Calculation System and Medical Data Processing Method

[0001] The present invention relates to a medical data calculation system and a medical data processing method for assisting medical practitioners.

[0002] Medical data such as medical treatment results, diagnosis results, health check results, examination results, or prescription details for patients are preferably shared among multiple medical practitioners and researchers to contribute to the improvement of medical technology or research. However, since patients' medical data is personal information itself, basically, it can only be used by the attending physician within the medical institution diagnosing the patient for medical treatment purposes, or by researchers who have signed a cooperation contract only for research purposes.

[0003] In contrast, the standardization of electronic medical records for leveraging medical data for the cooperation of physicians and other medical practitioners has been promoted (Non-Patent Document 1, etc.). As a result, a plurality of medical terminal devices used in medical institutions include devices with abundant computing resources that can support the standardized electronic medical records.

[0004] Ministry of Health, Labour and Welfare, "Regarding the Progress of Data Health Reform and Medical DX", December 15, 2023, <URL:https: / / www.kantei.go.jp / jp / singi / kenkouiryou / data_rikatsuyou / dai11 / siryou2.pdf>

[0005] Rather than circulating the medical data itself, which is personal information, among the physicians diagnosing the patients of that medical data and researchers who have access rights to the medical data, it is preferable to provide the information obtained by calculations using the collection of those medical data as widely as possible to medical practitioners.

[0006] An object of the present invention is to provide a medical data calculation system and a medical data processing method for assisting medical practitioners.

[0007] A medical data calculation system according to one embodiment of the present disclosure includes a medical terminal device within a medical institution that accepts input of medical data including medical treatment results, diagnostic results, health checkup results, test results, or prescription details; a shared storage that stores the medical data collected from a group of medical terminal devices within a plurality of medical institutions, including the medical terminal device; and a plurality of support data output devices that output different medical support data by performing calculations using the medical data stored in the shared storage. The medical terminal device acquires the medical support data from any of the plurality of support data output devices and outputs the acquired medical support data.

[0008] The medical data calculation system and medical data processing method disclosed herein utilize a group of medical terminal devices used within a medical institution to handle electronic medical records in a manner that protects personal information, as computing resources to perform calculations for outputting various medical support data. This makes it possible to provide various medical support data obtained through calculations using these devices as computing resources to medical terminal devices that realize a standardized electronic medical record system.

[0009] According to this disclosure, it will be possible to output data that supports healthcare providers by effectively utilizing the computing resources of a group of medical terminal devices used in a way that protects personal information.

[0010] This is a schematic diagram of a medical data calculation system. This is a block diagram showing the configuration of the medical terminal device of the medical data calculation system. This is a schematic diagram of a learning model. This is an explanatory diagram of the functions of the medical support data output device. This is a block diagram showing the configuration of a data server. This is a flowchart showing an example of the output processing procedure for medical support data. This is a schematic diagram of the medical data calculation system in the second embodiment. This is a flowchart showing an example of the output processing procedure for medical support data in the second embodiment.

[0011] This disclosure will be described in detail with reference to drawings illustrating embodiments thereof.

[0012] [First Embodiment] Figure 1 is a schematic diagram of a medical data calculation system 100. The medical data calculation system 100 includes a plurality of medical terminal devices 1 installed within a medical institution, an internal network N1 to which the plurality of medical terminal devices 1 are connected, and a data server 2. The medical data calculation system 100 of the first embodiment effectively utilizes the computing resources (CPU (Central Processing Unit), MPU (Micro-Processing Unit), GPU (Graphics Processing Unit), NPU (Neural Processing Unit), etc.) and storage media owned by each of the plurality of medical terminal devices 1 to provide medical support data to the medical terminal devices 1 within the medical institution.

[0013] Internal network N1 is a closed network within the medical institution, enabling communication between medical terminal devices 1. Internal network N1 can communicate with external network N2 via communication equipment. Network N2 is a public communication network or a communication network for electronic medical records. Medical terminal devices 1 can acquire data from data server 2 via internal network N1 and network N2. Medical terminal devices 1 can also acquire data other than that from data server 2, such as weather forecast data and infectious disease outbreak reports for each municipality nationwide, which are connected to the public communication network.

[0014] Each medical terminal device 1 is a device that can log in to a standard electronic medical record system using authentication based on a Healthcare Public Key Infrastructure (HPKI) electronic certificate for the healthcare sector of medical service providers, including physicians. Medical terminal devices 1 are diverse and can be a server computer, a desktop personal computer, a laptop personal computer, a tablet device, or a smartphone.

[0015] The medical data calculation system 100 shown in Figure 1 stores medical data received as input from multiple medical terminal devices 1 within the medical institution in a storage medium, and uses the collected computing resources to perform calculations using a language model, which will be more computationally intensive, as described later. The multiple medical terminal devices 1 may be nodes in a private blockchain within the medical institution, and the calculations may be executed based on smart contracts. It may also be configured so that one of the multiple medical terminal devices 1 acts as the master. The medical data calculation system 100 of the medical institution appropriately selects various trained models and configuration combinations, including language models, from the data server 2 based on the functions that can be provided by those models and combinations, and outputs medical support data obtained by analyzing the medical data acquired at the medical institution.

[0016] Data Server 2 provides various pre-trained models and settings, each designed to output medical support data to further improve the accuracy of medical examinations when given medical data, categorized by type of medical support data. The pre-trained models and settings for each type may share a common language model. While the pre-trained models are fundamentally trained using widely collected medical data, they are designed to be optimized by training with medical data obtainable by each medical institution. The pre-trained models may be categorized by functional hierarchy, including detailed models such as language models specialized in natural language input and output, vectorization modules that vectorize features of input text, models specialized in image recognition, and learning models specialized in image generation, as well as a leadership model that performs comprehensive inference using the outputs from various models. Each medical institution can select a combination of pre-trained models and settings from Data Server 2.

[0017] Data server 2 receives feedback from a portion of the medical support data output from the learning models and combinations of learning models learned at each medical institution, enabling the medical support data obtained at one medical institution to be shared with other medical institutions.

[0018] The detailed configuration and processing procedures of such a medical data calculation system 100 will be described below. Figure 2 is a block diagram showing the configuration of the medical terminal device 1 of the medical data calculation system 100. As described above, the medical terminal device 1 of a single medical institution can adopt multiple configurations. The medical terminal device 1 includes a processing unit 10, a storage unit 11, a communication unit 12, a display unit 13, and an operation unit 14.

[0019] The processing unit 10 includes one or more processors such as CPUs, MPUs, GPUs, and NPUs. The processing unit 10 also includes memory, which is a temporary storage medium such as SRAM (Static Random Access Memory) and DRAM (Dynamic Random Access Memory). The processing unit 10 may be configured as a single hardware (SoC: System On a Chip) integrating the processor, memory, storage unit 11, and communication unit 12.

[0020] The storage unit 11 is a relatively large-capacity non-temporary storage medium such as a hard disk or SSD (Solid State Drive). The storage unit 11 stores the program (program product) necessary for the processing unit 10 to execute processing, and setting data for reference. The program product includes a medical data calculation program P1, datasets of multiple learning models M11, M12, M13, ..., and setting data.

[0021] The medical data calculation program P1 included in the program product is a program that, when read into memory and executed by the processing unit 10, causes a general-purpose computer to function as the medical terminal device 1 of this disclosure, which performs various processes described later.

[0022] The multiple learning models M11, M12, M13, ... are models that output medical support data to make medical examinations more appropriate when medical data is input. Of the multiple learning models M11, M12, M13, ..., learning model M11 is a language model that accepts queries in natural language and outputs answers in natural language. Of the other learning models M12, M13, ..., learning model M12 is an image recognition model that has been trained to output image diagnostic results from medical images when medical images included in medical data are given. Of the learning models M12, M13, ..., learning model M13 is a model that outputs a prediction of the number of infected people as an infectious disease epidemic prediction for each infectious disease by being given patient numbers at medical institutions, drug prescription data at medical institutions, and weekly data announced by each local government in a time series. The learning models M12, M13, ... include models that have been trained to output nutritional guidance data that suggests appropriate meals when diagnostic results included in medical data and patient attributes are given. The learning models M12, M13, ... may include models trained to output precautions for taking medication, taking into account drug interactions with other prescribed drugs linked to the identification data, given diagnostic results, prescription results, and patient identification data included in medical data. The learning models M12, M13, ... may also include models trained to output medical fee items, given diagnostic results, test results, etc., included in medical data.

[0023] The configuration data stored in the memory unit 11 includes table-based data that can be referenced without using other learning models M12, M13, ... of the learning model M11 that controls input and output, such as electronic medical records of medical institutions and medical fee tables.

[0024] The learning models M11, M12, M13, ... and the configuration data stored in the memory unit 11 are downloaded by the processing unit 10 from the data server 2 via the communication unit 12 and stored in the memory unit 11. The medical data calculation program P1 stored in the memory unit 11 may also be downloaded by the processing unit 10 from an external download server or from the data server 2 via the communication unit 12 and stored in the memory unit 11. These program products stored in the memory unit 11 may also be program products stored in a computer-readable non-temporary storage medium 9, which the processing unit 10 reads and stores in the memory unit 11. The non-temporary storage medium 9 may store the medical data calculation program P9, learning models M91, 92, 93, ... and configuration data.

[0025] The communication unit 12 enables communication via the internal network N1 or network N2. The communication unit 12 may be a wired network card or a Wi-Fi wireless communication device. The communication unit 12 may also be a wireless communication device connected to a carrier network. The processing unit 10 can send and receive data with other medical terminal devices 1 or data servers 2 via the communication unit 12.

[0026] The display unit 13 is a display such as a liquid crystal display or an organic EL (Electro-Luminescence) display. The display unit 13 may also be a touch panel-integrated display. The processing unit 10 displays the UI (User Interface) of a standard electronic medical record system on the display unit 13. The processing unit 10 can also display the UI as a portal client, which is displayed based on the medical data calculation program P1, on the display unit 13.

[0027] The operation unit 14 is a user interface that can input and output to the processing unit 10. The operation unit 14 is, for example, a touch panel built into the display unit 13. The operation unit 14 may include a pointing device such as a mouse and a keyboard, or it may also include physical buttons, switches, and physical dials. The operation unit 14 may be configured to accept voice commands using a microphone and a voice recognition processing unit.

[0028] Figure 3 is an overview diagram of the learning model M11. As described above, the learning model M11 uses a language model and is a leadership model that appropriately selects learning models M12, M13, ... and settings in response to natural language queries, and integrates the outputs from the selected learning models M12, M13, .... The learning model M11 includes a language model. The language model is not large-scale and is a small language model that can be used with the limited computing resources of each medical institution.

[0029] The learning model M11 includes an Embedding module that performs tokenization and vectorization operations on input queries (prompts) before providing them to the language model. The Embedding module outputs the features of the given natural language text.

[0030] The learning model M11 includes an image recognition model that outputs vector data indicating the features of an image (such as what is depicted) when an image is input along with a query. The learning model M11 is configured to concatenate the vector data output from the image recognition model with the query (prompt) as vector data before inputting it to the language model. The learning model M11 may use settings corresponding to the input query, or setting data stored in the shared storage which is a collection of memory units 11, as knowledge.

[0031] The learning model M11 outputs text that is output from the language model. The learning model M11 may also be used to ensure that the output is appropriate before the user inputs a query, so that this is reflected in the response to subsequent queries.

[0032] Some or all of the language model, image recognition model, and Embedding module constituting the learning model M11 may be provided by an external server connected via network N. The learning model M11 may employ a multimodal model and assume the input of image data. The learning model M11 can employ models of other architectures.

[0033] The processing unit 10, using the medical data calculation program P1, performs calculations using the learning model M11 by aggregating available computing resources from the processing units 10 of other medical terminal devices 1 within the medical institution, and functions as a medical support data output device through this aggregation. The processing unit 10 can launch the portal client UI using the medical data calculation program P1. The portal client UI may be in the form of a web browser or a client program. The UI may include input screens for text and numerical values ​​corresponding to inquiries, or it may include a selection screen that accepts the selection of a medical support data output device.

[0034] Figure 4 is an explanatory diagram of the functions of the medical support data output devices 101, 102, .... The processing unit 10, in cooperation with other processing units 10, performs the functions of different medical support data output devices 101, 102, ... in response to queries, based on the medical data calculation program P1 and the learning model M11. Based on the medical data calculation program P1, the processing unit 10 provides queries input from operators using the medical terminal device 1 to a master process based on the learning model M11, which is a language model executed with aggregated computing resources, or to the processing unit 10 of the master medical terminal device 1. If the medical data calculation system 100 employs blockchain, the master process is the smart contract that is called first. The master process or the master processing unit 10 internally switches which medical support data output device 101, 102, ... function to perform in response to queries. Based on the medical data calculation program P1, the processing unit 10 may accept an active selection from the operator regarding which medical support data output device 101, 102, ... function to perform. The master process or master processing unit 10 may internally select which medical support data output device 101, 102, ... to perform in response to input to the electronic medical record system, based on the medical data calculation program P1.

[0035] As shown in Figure 4, the processing unit 10 of the medical terminal device 1 functions as a medical support data output device 101 that outputs infectious disease epidemic predictions, or as a medical support data output device 102 that performs medical image diagnosis. The processing unit 10 also functions as a medical support data output device 103 that outputs data to support nutritional guidance, or as a medical support data output device 104 that outputs data for instructing patients on taking prescribed medications. The processing unit 10 may also function as a medical support data output device that supports prescription data, rehabilitation guidance, or the calculation of medical fees.

[0036] The medical support data output device 101 analyzes diagnostic results and prescription data registered via the electronic medical record system from medical terminal devices 1 within a medical institution, which are sequentially stored in a shared storage that aggregates parts of the storage units 11 of multiple medical terminal devices 1. For each infectious disease, it outputs an infectious disease outbreak prediction. The infectious disease outbreak prediction is medical support data that helps to make medical examinations more appropriate.

[0037] The medical support data output device 101 outputs results using a learning model M13 that learns to output a prediction of the number of infected people as an infectious disease epidemic prediction for each infectious disease, by providing the medical institution with diagnosis results and drug prescription data at the medical institution in a time series. The medical support data output device 101 sequentially creates training data with medical data obtained within the medical institution (diagnosis results of infectious diseases, number of patients diagnosed with infectious diseases, number of patients without infectious diseases, number of prescriptions for specific drugs within the medical institution) as explanatory variables, and the number of infected people in the region including the location of the medical institution, which is included in the weekly data announced by each local government, as the dependent variable. The medical support data output device 101 uses the created training data to train and optimize the learning model M13 for the medical institution. The medical support data output device 101 outputs a prediction of the number of infected people in the region including the location of the medical institution by providing the medical data used as explanatory variables to the sequentially trained learning model M13. The processing unit 10 of the medical terminal device 1 provides medical data of explanatory variables to the medical support data output device 101 once a week, obtains weekly epidemic forecasts, and stores them in the storage unit 11. The processing unit 10 then feeds back the stored weekly epidemic forecasts to the data server 2.

[0038] The medical support data output device 102 performs medical image diagnosis functions. When medical images taken by X-ray or the like are provided, the medical support data output device 102 analyzes the medical images and outputs data indicating the lesions in the organs shown in the medical images. Specifically, when medical images are provided, the medical support data output device 102 sequentially creates training data with medical images of patients obtained at the medical institution as explanatory variables and the interpretation results and diagnostic results of those medical images as dependent variables, so that it outputs the disease name and probability related to the organ. The interpretation results, which are the dependent variables in the training data, may be image data indicating the range in which the lesion is shown in the medical image, or they may be text corresponding to the explanation of the medical image. The medical support data output device 102 uses the created training data to train and optimize the learning model M12 for the medical institution. The medical support data output device 102 obtains the interpretation results and diagnostic results (disease name and probability) of the medical images by providing medical images of patients at the medical institution to the sequentially trained learning model M12. The medical support data output device 102 uses a learning model M11 that includes a language model to output the image interpretation results and diagnostic results as natural language text. The medical support data output device 102 may have separate functions depending on the type of image, the organ being diagnosed, or the purpose of image acquisition (suspected disease name or injury name), such as chest X-ray, gastric X-ray, fundus photograph, or catheter image.

[0039] The medical support data output device 103 analyzes the diagnosis or test results and outputs data to support nutritional guidance according to the analysis results. When the medical support data output device 103 is given the results of the medical interview and diagnosis at the medical institution, it uses the language model learning model M11 to output a treatment plan based on the relationship between the medical interview results, diagnosis results, or test results included in the setting data, the emphasis on diet and exercise in the treatment plan, and the guidance items within them. The medical support data output device 103 may also use learning models M12, M13, ... which have been trained on learning data in which the medical interview results or diagnosis results are the explanatory variable and the treatment plan previously set by medical professionals including doctors is the dependent variable. The medical support data output device 103 may also automatically create and output document data of the treatment plan using the learning model M11.

[0040] The medical support data output device 104 outputs data to support medication guidance to patients who have been prescribed medication. When the medical support data output device 104 is given prescription data from a doctor at the medical institution and patient identification data (e.g., My Number), it acquires the patient's attributes and prescription history, including prescriptions from other medical institutions, and outputs precautions for taking the medication by referring to drug data provided by public institutions. The medical support data output device 104 reads the patient's attributes, patient identification data, diagnostic history, and patient interview results, which are associated with the patient's registration number stored in a shared storage that aggregates parts of the storage units 11 of multiple medical terminal devices 1. Based on the patient's identification data, the medical support data output device 104 acquires the prescription history from other medical institutions via the electronic medical record communication network. The medical support data output device 104 analyzes the acquired patient attributes, diagnostic content, and prescription data, extracts information corresponding to that data from drug data acquired from public institutions, and determines whether there are any suspicions regarding drug interactions, etc. The medical support data output device 104 outputs patient attributes, diagnostic results, and medication precautions based on prescription data, using a language model, the learning model M11. The medical support data output device 104 may also use a learning model that has been trained to output medication precautions when patient attributes, diagnostic results, and prescription data are provided.

[0041] The medical support data output devices 101, 102, ... may function as devices that output prescription data indicating whether or not a drug should be taken with other prescription drugs linked to identification data, based on drug data stored as settings or drug data provided by external services. The medical support data output devices 101, 102, ... may function as devices that analyze the history of diagnostic results stored in association with the patient's unique identification data and output rehabilitation plan and guidance data. The medical support data output devices 101, 102, ... may analyze the contents of the patient's electronic medical record and perform a function to support the calculation of medical fees using a remuneration table formulated by a public institution or a remuneration table specified by a medical institution.

[0042] Figure 5 is a block diagram showing the configuration of data server 2. Data server 2 uses a server computer. In the following description, data server 2 will be described as being composed of a single server computer, but it may also be configured so that multiple server computers perform distributed processing and function as a single data server 2. Data server 2 may also function as a cloud server while ensuring security.

[0043] The data server 2 comprises a processing unit 20, a storage unit 21, and a communication unit 22. The processing unit 20 includes one or more processors such as CPUs, MPUs, and GPUs. The processing unit 20 includes memory, which is a temporary storage medium such as SRAM or DRAM. The processing unit 20 may be configured as a single hardware unit integrating the processor, memory, storage unit 21, and communication unit 22.

[0044] The storage unit 21 is a relatively large-capacity non-volatile storage medium such as a hard disk or SSD. In the storage unit 21, a database 210 for storing data provided from each medical institution is constructed. The database 210 includes learning models M21, M22, …, and may be obtainable from the medical terminal device 1. The learning model M21 is a language model similar to the learning model M11, and the learning model M22 is an image recognition model similar to the learning model M12. The learning models M21, M22, … are a group of models that are the basis of the learning models M11, M12, … when optimized in the medical terminal device 1. The storage unit 21 stores, in association with the identification data of the medical institution, the identification data of the region where the medical institution is located, the number of patients, and the prediction of the spread of infectious diseases.

[0045] The communication unit 22 realizes communication via the network N2. The communication unit 22 may be a network card for wired use or a wireless communication device for WiFi. The communication unit 22 may also be a wireless communication device that connects to a carrier network. The processing unit 20 can transmit and receive data with the medical terminal device 1 via the communication unit 22.

[0046] In the medical data calculation system 100 configured as described above, each time the medical terminal device 1 inputs an interrogation result, a diagnosis result, or an examination result based on the electronic medical record function, it stores them as medical data in the shared storage. The medical terminal device 1 collaborates with other medical terminal devices 1 within the medical institution to perform a function by any of a plurality of functions (medical support data output devices 101 to 104, …), and outputs support data based on a schedule in response to an operator's input. A specific example of the output process of the medical support data will be described below.

[0047] FIG. 6 is a flowchart showing an example of the output process procedure of the medical support data. The master process or the processing unit 10 of the master medical terminal device 1 (hereinafter simply referred to as the master) issues a processing ID (step S101).

[0048] The master calls one of the medical support data output devices 101, 102, … based on an input from an operator or a schedule (step S102). In step S102, the master calls the UI of the portal client based on the medical data calculation program P1. The processing unit 10 receives an inquiry based on the operation unit 14 through the UI, and selects one of the medical support data output devices 101, 102, … by means of the learning model M11 which is a language model. In step S102, when the master starts processing based on a schedule, it may select and activate a specific medical support data output device 101, 102, … without calling the UI of the portal client.

[0049] The master acquires necessary medical data from the shared storage for the called medical support data output devices 101, 102, … (step S103). The master provides the acquired medical data to the medical support data output devices 101, 102, … (step S104), and acquires medical support data from the medical support data output devices 101, 102, … which perform calculations using the aggregated computing resources (step S105). The master stores the acquired medical support data in the shared storage in association with the processing ID (step S106).

[0050] The processing unit 10 of each medical terminal device 1 calculates the processing load for each of steps S104 and S105 (step S107). In step S107, the processing unit 10 calculates the processing amount from the load rate in the processing unit 10 due to the processing in steps S104 and S105, the clock of the processing unit 10, and the capacity of the built-in RAM. The processing unit 10 stores the calculated processing load in the shared storage in association with the processing ID issued in step S101 and the identification data of its own device (unique identification data such as the operator's unique identification data (HPKI) or MAC address) (step S108).

[0051] The processing unit 10 of each medical terminal device 1 displays the acquired medical support data on the display unit 13 (step S109), and ends the processing.

[0052] The processing procedure shown in Figure 6 will be explained in detail with a specific example. In the first example, the processing unit 10 of the medical terminal device 1, based on the medical data calculation program P1, calls the medical support data output device 101, which outputs infectious disease outbreak predictions, once a week, for example, at a predetermined time on Friday, Saturday, or Sunday (S102). The processing unit 10 uses the functions of the medical support data output device 101 to acquire time-series data of infectious disease diagnosis results, the number of prescriptions by type of drug, and the number of patients, which have been performed at the medical institution up to that point, input from the medical terminal device 1, and stored in the shared storage (S103).

[0053] The processing unit 10, using the functions of the medical support data output device 101, provides the time-series data obtained in step S103—the diagnosis result of the infectious disease, the number of prescriptions for each type of drug, and the number of patients—to the trained learning model M13 for predicting the epidemic of an infectious disease (S104). The processing unit 10, using the functions of the medical support data output device 101, outputs a prediction from the learning model M13 for the number of infected people for that infectious disease in the region including the location of the medical institution in the following week (S105).

[0054] The processing unit 10 executes the processes of steps S104 and S105 for a specific group of infectious diseases and displays a screen on the display unit 13 that includes a list of predicted numbers of infected people for each of the specific group of infectious diseases (S109).

[0055] In the first example, the processing unit 10 sends the predicted number of infected people for each infectious disease to the data server 2. The data server 2 collects the predictions for each infectious disease that are sent on a schedule from medical institutions to which the medical data calculation system 100 is applied. The processing unit 20 of the data server 2 performs statistical processing (calculating the mean, median, etc.) on the collected results for each region, including the location of the medical institution, and feeds them back to the medical terminal device 1 of each medical institution. In step S108, the processing unit 10 may display the predicted number of infected people after the statistical processing by the data server 2.

[0056] In the first example, as an option for the standard electronic medical record system, processing based on the medical data calculation program P1 can be performed on multiple medical terminal devices 1 located in each medical institution, making it possible to obtain infectious disease outbreak predictions on each medical terminal device 1.

[0057] In the second example, the processing unit 10 of the medical terminal device 1 accepts input of captured medical images via the UI of a portal client displayed as an option (plugin) for the standard electronic medical record system, based on the medical data calculation program P1. The UI of the portal client accepts medical image data via drag and drop, or accepts input of the location of the medical image data on the electronic medical record system (directory / folder / patient's unique identification data and image number, etc.) (S103). The processing unit 10 may also accept selection of the type of medical image or the suspected disease name (purpose of image acquisition).

[0058] The processing unit 10 calls the medical support data output device 102 for medical image diagnosis (S102) and provides medical image data (S104). In step S102, the processing unit 10 calls the medical support data output device 102 according to the type of medical image or the purpose of shooting. The processing unit 10 uses the functions of the medical support data output device 102 to output the disease name and probability diagnosed from the medical image using the learning model M12. The processing unit 10 acquires the output disease name and probability using the functions of the medical support data output device 102 (S105). The processing unit 10 displays the acquired disease name and probability on the display unit 13 (S109).

[0059] In the second example, as an option for the standard electronic medical record system, processing based on the medical data calculation program P1 can be performed on multiple medical terminal devices 1 located in each medical institution, thereby enabling each medical terminal device 1 to display information that supports the interpretation of medical images.

[0060] In the third example, the processing unit 10 of the medical terminal device 1 receives the medical interview results and diagnosis results based on the medical data calculation program P1, using the UI of a portal client displayed as an option (plugin) for the standard electronic medical record system. The processing unit 10 may also receive the medical interview results and diagnosis results as images taken from documents written on paper, or it may provide medical interview items and options in the UI and accept selections.

[0061] The processing unit 10 calls the medical support data output device 103 which outputs support data for nutritional guidance (S102). The processing unit 10 reads data relating to the interview results, diagnosis results or test results stored in the shared storage, the emphasis on diet and exercise in the treatment plan, and the guidance items within them (S103). The processing unit 10 provides the interview results and diagnosis results received by the portal client UI read in step S102 (S104).

[0062] The processing unit 10 uses the aggregated computing resources to execute the processing of the learning model M11 in the medical support data output device 103, and outputs a treatment plan including key items and guidance items for diet and exercise, using the relationship data read in step S102 as knowledge. The processing unit 10 uses the learning model, which is trained by the medical support data output device 103 to output a treatment plan including key items and guidance items for diet and exercise, to output data indicating the contents of the treatment plan. The processing unit 10 acquires the treatment plan data output from the medical support data output device 103 (S105), generates natural language treatment advice using the language model learning model M11, and displays it on the display unit 13 (S109).

[0063] In the third example, as an option for the standard electronic medical record system, processing based on the medical data calculation program P1 can be performed on multiple medical terminal devices 1 located in each medical institution, allowing each medical terminal device 1 to display a screen based on information about the contents of the treatment plan.

[0064] The master processing unit 10 calculates the processing load of each medical terminal device 1 within the medical institution for outputting the medical support data, based on the medical data calculation program P1, as the contribution of computing resources to that medical support data. The processing unit 10 may also calculate a reward for each medical terminal device 1 according to the calculated contribution. Even within a medical institution, the contribution may be calculated for each department, and the reward may be distributed from the medical institution's profits to the departments that own each medical terminal device 1. For example, in the case of medical support data that can be provided externally via the data server 2, such as infectious disease outbreak predictions, the reward may be calculated so that a larger reward is distributed to the medical institution that provides more computing resources and outputs the outbreak prediction.

[0065] Thus, as an option for the standard electronic medical record system, by enabling processing based on the medical data calculation program P1 on multiple medical terminal devices 1 present in each medical institution, each medical terminal device 1 can acquire medical support data from the aforementioned medical support data output devices 101, 102, .... The heavy processing based on the language models of the medical support data output devices 101, 102, ... can be efficiently performed by utilizing the combined computing resources of multiple medical terminal devices 1, allowing for efficient use of medical terminal devices 1, which may possess abundant computing resources.

[0066] The data (parameters) of the analytical learning models M12, M13, ..., which are retrained and optimized using medical data stored in the shared storage of each medical institution, are fed back to the data server 2, and the data server 2 may perform federated learning. The data server 2 statistically processes the learning models M12 trained at each medical institution and creates a redistributable, integrated learning model M12. The same applies to the other learning models M13, ...

[0067] [Second Embodiment] In the second embodiment, calculations can be performed across different medical institutions. In the second embodiment, any medical terminal device 1 connected to a standard electronic medical record system can receive medical support data in common. In addition, in the second embodiment, each of the multiple medical terminal devices 1 has a blockchain address so that it can function as a node in a blockchain system. In the second embodiment, the medical support data output devices 101, 102, ... are implemented as smart contracts deployed on the blockchain. Each medical terminal device 1 obtains the results of processing procedures based on these smart contracts as an option for the standard electronic medical record system.

[0068] Figure 7 is a schematic diagram of the medical data calculation system 200 in the second embodiment. The medical data calculation system 200 in the second embodiment includes a group of medical terminal devices 1 installed in each of multiple medical institutions, an internal network N1 connecting the groups of medical terminal devices 1 within each medical institution, and a network N2 that allows each medical terminal device 1 to connect across medical institutions. As described above, the medical terminal devices 1 can be connected to the blockchain system 5. The collection of storage media of each node in the blockchain system 5 is used as shared storage 50. Medical data entered by the medical terminal devices 1 of each medical institution is stored in the shared storage 50. Access to medical data can also be restricted by requiring the purchase of NFTs (Non-Fungible Tokens) issued for the medical data, thereby protecting medical data, which is a collection of personal information.

[0069] In the second embodiment as well, the medical terminal device 1 is a collection of devices that communicate with each other via the network N2 and authenticated based on HPKI, and they cooperate to perform the functions of the medical support data output devices 101, 102, 103, ...

[0070] By adopting the blockchain system 5, medical data can be stored in a tamper-proof manner, access to this data can be restricted as personal information, and calculations using this medical data can be collaboratively performed by multiple medical terminal devices 1. Furthermore, since micropayments can be made as rewards for contributions to each calculation using tokens that can be issued on the blockchain system 5, and by adopting the blockchain system 5, it is also compatible with federated learning of learning models M11, M12, M13, ... used in each smart contract. In the second embodiment, the data server 2 is not required. Data sharing across different medical institutions and different regions is performed between multiple medical terminal devices 1.

[0071] In the second embodiment, the medical support data output devices 101, 102, ... are deployed as smart contracts on the blockchain system 5, as described above. In the blockchain system 5, medical data from different medical institutions is also shared across the entire medical data calculation system 200.

[0072] Figure 8 is a flowchart showing an example of the output processing procedure for medical support data in the second embodiment. The processing unit 10 of the medical terminal device 1 broadcasts a request transaction to the blockchain system 5 where the medical support data output devices 101, 102, ... are deployed, based on input from the operator or a schedule (step S201). In step S201, the processing unit 10 may broadcast a request for prediction of the number of infected people with an infectious disease, specifying the infectious disease and region, or a request for image diagnosis of medical images, specifying the medical image data and the purpose of taking the captured image (suspected disease name). The processing unit 10 may also broadcast a request for nutritional guidance, specifying the patient's unique identification data, the results of the medical interview, and identification data that identifies the diagnosis results (medical institution ID and consultation ID). The processing unit 10 naturally includes the blockchain address of the medical terminal device 1 in the request transaction.

[0073] Blockchain system 5 issues an ID to identify the transaction for the request transaction (step S501) and captures the transaction (step S502). Based on the captured transaction, blockchain system 5 invokes a process based on the learning model M11, which is a language model corresponding to the request, and further invokes medical support data output devices 101, 102, ... (smart contracts) (step S503). In step S503, if a request for prediction of the number of infected people is broadcast, blockchain system 5 invokes the contract of medical support data output device 101, and if a request for blockchain image diagnosis is broadcast, it invokes the contract of medical support data output device 102. Blockchain system 5 also invokes the contracts of medical support data output devices 103, ... as appropriate for other requests.

[0074] The blockchain system 5 uses the combined computing resources of each processing unit 10 of the medical terminal device 1 to read the necessary medical data from the shared storage 50 at the called medical support data output devices 101, 102, ... (step S504). The blockchain system 5 provides the acquired medical data to the medical support data output devices 101, 102, ... (step S505), causing the medical support data output devices 101, 102, ... to output medical support data (step S506). The medical support data may be stored in the shared storage 50 in association with the issued ID.

[0075] During the processing in steps S505 and S506, the processing unit 10 of each medical terminal device 1 calculates the processing load that contributed to the processing in steps S505 and S506 (step S202). In step S202, the processing unit 10 calculates the processing amount from the load ratio of the processing unit 10 due to the processing in steps S505 and S506, the clock of the processing unit 10, and the capacity of the built-in RAM. The processing unit 10 stores the calculated processing load in shared storage, associating it with the processing ID issued in step S101 and the identification data of its own device (operator's unique identification data (HPKI) or unique identification data such as MAC address) (step S203).

[0076] The medical terminal device 1 that made the request retrieves the medical support data output in response to the request from the blockchain system 5 (step S204). The processing unit 10 of each medical terminal device 1 displays the retrieved medical support data on the display unit 13 (step S205) and terminates the process.

[0077] In the processing procedure shown in Figure 8, as explained in the first embodiment, the medical data calculation system 200 can predict infectious diseases or output image diagnostic results in response to requests from the affiliated medical terminal device 1.

[0078] As a result, in the second embodiment, large-scale calculations can be performed using a large collection of medical terminal devices 1 spanning multiple medical institutions. The medical data calculation system 200 of the second embodiment is configured to allow for coordinated processing while separating the medical terminal devices 1 by medical institution. Therefore, even if a medical terminal device 1 operating at a medical institution is closed, if it performs calculations as a node in the blockchain system 5, it can receive a distribution of rewards in proportion to its contribution to the medical support data from those calculations. The medical data calculation system 200 of the second embodiment may also distribute as rewards, in proportion to the contribution of computational resources to the medical support data, the fees paid by each medical institution for the optional use of the standard electronic medical record system, or the sales profits for researchers and developers of NFTs corresponding to medical data.

[0079] In the second embodiment, the medical support data output devices 101, 102, ... are implemented as smart contracts deployed on the blockchain. However, the embodiment is not limited to this, and each of the medical support data output devices 101, 102, ... may be configured to execute processing while the medical terminal device 1 is connected to different medical institutions via communication.

[0080] The embodiments disclosed above are illustrative in all respects and not restrictive. The scope of the invention is indicated by the claims, and all modifications within the meaning and scope of the claims are equivalent.

[0081] 100 Medical data calculation system 1 Medical terminal device 10 Processing unit 101, 102, 103, 104 Medical support data output device 11 Storage unit 12 Communication unit P1 Medical data calculation program M11, M12, M13 Learning model 2 Data server 20 Processing unit 21 Storage unit

Claims

1. A medical data calculation system comprising: a medical terminal device within a medical institution that accepts input of medical data including medical examination results, diagnosis results, health checkup results, test results, or prescription details; a shared storage that stores the medical data collected from a group of medical terminal devices within multiple medical institutions, including the medical terminal device; and a plurality of support data output devices that output different medical support data by performing calculations using the medical data stored in the shared storage, wherein the medical terminal device acquires the medical support data from any of the plurality of support data output devices and outputs the acquired medical support data.

2. The medical data calculation system according to claim 1, wherein the plurality of support data output devices aggregate the computing resources of the plurality of medical terminal devices within the medical institution to perform calculations.

3. The medical data calculation system according to claim 2, wherein each of the plurality of support data output devices uses a learning model learned from medical data stored in the shared storage, and performs calculations by the learning model using the aggregated computing resources.

4. The medical data calculation system according to any one of claims 1 to 3, wherein each of the medical terminal devices within the plurality of medical institutions receives a distribution of rewards in proportion to the contribution of computing resources to the output medical support data.

5. The medical data calculation system according to any one of claims 1 to 4, wherein the plurality of support data output devices are divided by region, by disease name, or by support purpose based on the location of the medical institution of the medical terminal device from which the medical data was collected.

6. The medical data calculation system according to any one of claims 1 to 5, wherein each of the plurality of support data output devices analyzes the collected medical data and outputs medical support data, image analysis data, health guidance support data, prescription support data, nutrition guidance support data, or rehabilitation guidance support data.

7. The medical data calculation system according to claim 6, wherein one of the plurality of support data output devices analyzes the collected medical data, the date and time data on which the medical data was input, and the location of the medical institution of the medical terminal device on which the medical data was collected, and outputs a regional epidemic prediction of infectious diseases.

8. A medical data processing method comprising the following steps: a computer installed in a medical institution, upon successful authentication for data input within the medical institution, accepts input of medical data including treatment results, diagnosis results, health check results, test results, or prescription details; transmits the medical data to a shared storage that stores the medical data collected from multiple medical terminal devices within the medical institution, including the computer; obtains the medical support data from one of the multiple support data output devices that output different medical support data through calculations using the medical data stored in the shared storage; and outputs the obtained medical support data.

Images