System

Abstract

A system is provided.SOLUTION: A system includes a means for collecting patient data, a means for performing a discussion by an artificial intelligence model finely tuned with a plurality of expert knowledges based on the collected data, a means for generating a treatment policy from the discussion result, a means for proposing the generated treatment policy to medical staff, and a means for storing the proposal result as feedback data and performing reinforcement learning of the artificial intelligence model.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The technology of the present disclosure relates to a system. [Background technology]

[0002] Patent document 1 discloses a persona chatbot control method performed by at least one processor, the method including the steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to a description of the chatbot character, encoding the prompt, and inputting the encoded prompt into a language model to generate a chatbot utterance in response to the user utterance. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-180282 Summary of the Invention [Problem to be solved by the invention]

[0004] In cancer care, cancer boards are forums where medical staff from various medical departments gather to discuss cancer patient treatment plans. However, there are problems with the quality of medical care, such as the workload of doctors and regional differences in the capabilities of medical professionals. In particular, it is difficult for doctors to balance their daily work with preparing for and participating in cancer board meetings, and in some regions, there is a shortage of specialists, which can prevent sufficient discussions. There is a need to resolve these issues and provide optimal and equal cancer care. [Means for solving the problem]

[0005] This invention provides a system including a means for collecting patient data, a means for conducting discussions based on the collected data using an artificial intelligence model fine-tuned with multiple specialized knowledge, a means for generating a treatment plan from the discussion results, a means for proposing the generated treatment plan to medical staff, and a means for saving the proposal results as feedback data and for reinforcement learning of the artificial intelligence model. The collected patient data is obtained from an electronic medical record system, converted into an appropriate format, and stored in a database. Each artificial intelligence model fine-tuned with specialized knowledge simulates the role of a specialist and holds discussions for each role, thereby reducing the burden on doctors and providing uniform cancer care regardless of region.

[0006] "Patient data" refers to information about patients undergoing medical examination and treatment, including medical records, test results, medical history, and diagnostic information obtained from electronic medical record systems.

[0007] "Means for collection" refers to a method or device for obtaining patient data from the electronic medical record system in an appropriate format and extracting the necessary information.

[0008] "Fine-tuning" is the process of fine-tuning a pre-trained artificial intelligence model using specific expertise and datasets.

[0009] An "artificial intelligence model" is an algorithm or system that is self-learning based on data and used to solve a specific problem.

[0010] A "means for conducting discussion" is a method or device that allows multiple artificial intelligence models with specialized knowledge to exchange opinions with each other and derive the optimal treatment plan.

[0011] A "treatment plan" refers to the medical procedures and treatment plans carried out based on the patient's diagnosis, and includes surgery, drug therapy, radiation therapy, etc.

[0012] The "means for proposing" refers to a method or device for presenting the treatment plan generated by the artificial intelligence model to medical staff so that they can consider it.

[0013] "Feedback data" is information obtained based on actual treatment plans and their results, and is used to improve the artificial intelligence model.

[0014] "Reinforcement learning" is a learning algorithm that continuously trains artificial intelligence models based on feedback data to improve their accuracy and performance.

[0015] A "database" is a system that stores large amounts of data in an organized manner and enables efficient search and retrieval.

[0016] "Simulating the role of a specialist" means that the AI ​​model mimics the knowledge and judgment of a doctor in a specific specialty and provides opinions on patient treatment plans. [Brief explanation of the drawings]

[0017] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0018] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0019] First, the terms used in the following description will be explained.

[0020] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0021] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0022] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0023] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0024] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0025] [First embodiment]

[0026] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0027] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0028] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0029] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0030] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0031] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0032] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0033] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0034] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0035] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0036] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0037] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0038] This invention specifically describes a cancer board system that utilizes AI. The following explains the system's program processing in natural language, and includes specific examples.

[0039] System Configuration

[0040] The system consists of a terminal that collects and manages patient data, a server that processes the collected data and holds discussions with an artificial intelligence model with specialized knowledge, and a terminal that proposes a final treatment plan and receives feedback.

[0041] Data Collection Phase

[0042] 1. The terminal obtains patient data from the electronic medical record system. For example, it obtains the medical details (diagnosis results, test results, medical history, etc.) of Patient A, a 50-year-old male.

[0043] 2. The terminal converts the acquired patient data into an appropriate format (e.g., JSON format) and sends it to the server.

[0044] 3. The server stores the received data in the database.

[0045] AI Cancer Board Phase

[0046] 4. The server initializes multiple artificial intelligence models (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) that are fine-tuned with specialized knowledge.

[0047] 5. The server reads patient data from the database and inputs the data into each specialist AI.

[0048] 6. Each AI specialist will provide its own opinion (treatment plan) based on patient data and discuss it with other AI specialists.

[0049] 7. The server aggregates the discussion results and generates an optimal treatment plan.

[0050] Proposal Phase

[0051] 8. The server sends the proposed treatment plan to the terminal.

[0052] 9. The proposed treatment plan received by the terminal is displayed to the cancer board members (users).

[0053] 10. Users (Cancer Board members) discuss the proposed treatment plan and decide on the final treatment plan.

[0054] Feedback Phase

[0055] 11. The terminal feeds back the final treatment plan decided upon to the server.

[0056] 12. The server stores the feedback data in a database and optimizes the artificial intelligence model using reinforcement learning (RLHF).

[0057] Specific examples

[0058] For example, consider the case where Patient A is a 50-year-old man diagnosed with stage II lung cancer. This data is first obtained from the electronic medical record system by the device and sent to the server. The server receives this data and distributes it to various specialist AI models, such as the surgeon AI and internist AI. The surgeon AI determines whether surgery is appropriate, and the internist AI evaluates whether drug therapy is applicable. Based on the opinions of each specialist AI, the server generates a proposed treatment plan that states, "Chemotherapy after surgery is optimal." This proposed treatment plan is sent to the device and reviewed by members of the cancer board. If the final treatment plan is decided to be "immunotherapy in addition to chemotherapy after surgery," this information is fed back to the server and used for reinforcement learning in the AI ​​model.

[0059] In this way, the present invention can reduce the workload of doctors and reduce regional variations in the quality of medical care.

[0060] The processing flow will be explained below.

[0061] Step 1:

[0062] The terminal connects to the hospital's electronic medical record system and obtains the patient's medical data. Specifically, the patient ID is entered and an API is called to obtain the medical data (e.g., 50-year-old male, diagnosis: stage II lung cancer, medical history: high blood pressure).

[0063] Step 2:

[0064] The device formats the acquired data into an appropriate format and sends it to the server. Specifically, it converts the patient data into JSON format and posts it to the server using an HTTP request.

[0065] Step 3:

[0066] The server saves the received data in the database. Specifically, it executes an SQL query to insert the received patient data into the database.

[0067] Step 4:

[0068] The server initializes each specialist AI (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) in the AI ​​model, which has been fine-tuned with specialized knowledge. Specifically, it loads and configures each specialist AI model.

[0069] Step 5:

[0070] The server reads patient data from the database and inputs the data into each specialist AI. Specifically, it reads the patient data using an SQL query, converts the data into a format, and passes it to each specialist AI.

[0071] Step 6:

[0072] Each AI specialist will analyze patient data from their own perspective and offer their own opinion (treatment plan). Specific actions include the surgeon AI determining whether surgery is appropriate, and the internist AI evaluating whether drug therapy is applicable.

[0073] Step 7:

[0074] The server aggregates the results of the specialist AI discussions and generates the optimal treatment plan. Specifically, it collects the opinions of each specialist AI and applies voting and consensus-forming algorithms to determine the optimal treatment plan.

[0075] Step 8:

[0076] The server sends the generated treatment plan proposal to the terminal. Specifically, it converts the treatment plan proposal into JSON format and posts it to the terminal using an HTTP request.

[0077] Step 9:

[0078] The terminal displays the proposed treatment plan received to the cancer board members. Specifically, the received proposed treatment plan is displayed on the screen so that the cancer board members can refer to it.

[0079] Step 10:

[0080] Users (cancer board members) discuss the proposals and decide on the final treatment plan. Specifically, the cancer board members review the proposed treatment plan and reach a final consensus.

[0081] Step 11:

[0082] The terminal feeds back the final treatment plan it has decided to use to the server. Specifically, it converts the final treatment plan into JSON format and posts it to the server using an HTTP request.

[0083] Step 12:

[0084] The server stores the final treatment plan received in a database and optimizes the AI ​​model using reinforcement learning (RLHF). Specifically, it inserts the feedback data into the database, supplies new data to the AI ​​model, and executes the reinforcement learning algorithm.

[0085] Example 1

[0086] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0087] In the conventional medical system, the process of deciding on a treatment plan takes time, and the inconsistency of expertise and a lack of understanding of the information make it difficult to quickly decide on the optimal treatment plan.In addition, the management and use of collected patient data was not centralized, and collaboration between medical specialties was often ineffective.

[0088] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0089] In this invention, the server includes means for acquiring patient data from an electronic system, means for converting the acquired data into an appropriate format and sending and storing it in a high-performance database, means for holding discussions using generative AI models fine-tuned with multiple pieces of specialized knowledge based on the stored data and for each expert knowledge model to simulate opinions, means for aggregating the discussion results to generate an optimal treatment plan, means for sending the generated treatment plan to the user and displaying it in a dedicated application, and means for the user to decide on a final treatment plan, sending that feedback to the system and storing it in the database, and for reinforcement learning of the generative AI model.This enables centralized management of medical data and rapid decision-making on a treatment plan.

[0090] "Patient data" refers to information about a specific patient, such as diagnostic results, test results, treatment history, and medical history, collected by a medical institution.

[0091] "Electronic system" means a system for managing and processing data in digital form, including computer systems and electronic medical record systems installed in medical institutions.

[0092] A "database" is a high-performance information system for efficiently storing, managing, and searching collected data.

[0093] A "generative AI model" is an artificial intelligence model that has been trained using deep learning techniques to have specific expertise, such as an AI surgeon or AI internist.

[0094] "Discussion" is the process in which multiple generative AI models exchange opinions based on patient data to arrive at the optimal treatment plan.

[0095] A "treatment plan" is a specific medical procedure or treatment plan proposed based on the patient's diagnostic and test results.

[0096] "Users" are medical staff and specialists who use the system.

[0097] "Reinforcement learning" is a machine learning technique that allows artificial intelligence models to improve their performance based on feedback data.

[0098] "Feedback" is the process of returning the results of proposed treatment plans and information after their implementation to the system and using it for subsequent discussions and model updates.

[0099] The present invention provides an AI-based cancer monitoring system that comprehensively collects, processes, analyzes, and provides feedback on medical data, with the aim of quickly determining optimal treatment plans. Specific embodiments of this system are described below.

[0100] System Configuration

[0101] The system consists of a terminal that collects and manages patient data, a server that processes the collected data and holds discussions with a generative AI model with specialized knowledge, and a terminal that proposes a final treatment plan and receives feedback.

[0102] Data collection and format conversion

[0103] 1. The terminal uses a PC or tablet used in a medical institution to obtain patient data from an electronic system (electronic medical record system). For example, the medical details (diagnosis results, test results, medical history, etc.) of Patient A, a 50-year-old male, are obtained.

[0104] 2. The terminal converts the acquired patient data into an appropriate format (e.g., JSON format) and sends it to the server. This conversion is performed using a programming language such as Python. For example, the patient data is converted into the following JSON format:

[0105] json

[0106] {

[0107] "patient_id": "12345",

[0108] "age": 50,

[0109] "gender": "male",

[0110] "diagnosis": "lung cancer stage II",

[0111] "medical_history": ["hypertension", "diabetes"],

[0112] "latest_results": {

[0113] "CT_scan": "positive",

[0114] "blood_test": "normal"

[0115] }

[0116] }

[0117] 3. The server stores the received data in a high-performance database (such as MySQL or PostgreSQL). This data storage process uses SQL queries.

[0118] AI Cancer Board Discussion

[0119] 4. The server initializes multiple generative AI models (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) fine-tuned with TENSORFLOW (registered trademark) or PyTorch. Specifically, a command like model = tf.keras.models.load_model('path_to_model') is assumed.

[0120] 5. The server reads the patient data from the database and inputs it into each generative AI model. This process is also implemented using scripts such as Python.

[0121] 6. Each generative AI model provides an opinion (treatment plan) based on the patient data and discusses it with other generative AI models. For example, an internist AI might suggest that "chemotherapy is optimal," and an surgeon AI might respond that "surgery should also be used."

[0122] 7. The server aggregates the results of the discussions and generates the optimal treatment plan. For example, the treatment plan generated might be "chemotherapy after surgery is optimal."

[0123] Treatment plan suggestions and feedback

[0124] 8. The server sends the generated treatment plan to the device and displays it in a dedicated application. This process uses an HTTP POST request and a secure communication protocol (HTTPS).

[0125] 9. The device displays the received treatment plan proposal to the user (cancer board member). This is done using a web application or mobile application.

[0126] 10. Users (cancer board members) discuss the proposed treatment plan and decide on the final treatment plan. This process uses dedicated discussion tools and remote video conferencing systems (Zoom, Teams, etc.).

[0127] 11. The device sends the final treatment plan back to the server. The feedback data is also sent in JSON format.

[0128] 12. The server stores the feedback data in a database again and optimizes the generative AI model using reinforcement learning with human feedback (RLHF). This advances the AI ​​model's learning, improving the accuracy of the next discussion and treatment plan proposal.

[0129] Specific examples

[0130] For example, if Patient A is a 50-year-old man diagnosed with stage II lung cancer, the data is retrieved from the electronic system by the terminal, converted to JSON format, and sent to the server. The server distributes the data to each generative AI model, with the surgeon AI determining whether surgery is appropriate and the internist AI evaluating whether drug therapy is applicable. As a result of the discussion, the server generates a treatment plan proposal stating that "chemotherapy after surgery is optimal" and sends it to the terminal. The cancer board members consider this and ultimately decide on a treatment plan of "immunotherapy in addition to chemotherapy after surgery," and this information is fed back to the server. The AI ​​model uses this feedback to perform reinforcement learning, improving the accuracy of future proposals.

[0131] Prompt Sentence Examples

[0132] "Patient A is a 50-year-old man who has been diagnosed with stage II lung cancer. His medical information includes his latest test results and medical history. Each AI specialist should propose a treatment plan based on this information."

[0133] In this way, the present invention can reduce the workload of doctors and reduce regional variations in the quality of medical care.

[0134] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0135] Step 1:

[0136] The terminal retrieves patient data from an electronic system (electronic medical record system). The input is the patient ID and treatment request, and the output is the raw patient data. Specifically, the terminal sends a GET / patients / {patient_id} request to the electronic medical record system via the API to retrieve the patient's medical information (for example, the diagnosis results, test results, medical history, etc. of 50-year-old male Patient A).

[0137] Step 2:

[0138] The terminal converts the acquired patient data into an appropriate format (for example, JSON format). The input is raw patient data, and the output is JSON formatted patient data. Specifically, a script running on the terminal parses the raw data and converts it into JSON format as follows:

[0139] json

[0140] {

[0141] "patient_id": "12345",

[0142] "age": 50,

[0143] "gender": "male",

[0144] "diagnosis": "lung cancer stage II",

[0145] "medical_history": ["hypertension", "diabetes"],

[0146] "latest_results": {

[0147] "CT_scan": "positive",

[0148] "blood_test": "normal"

[0149] }

[0150] }

[0151] Step 3:

[0152] The terminal sends the converted data to the server. The input is JSON-formatted patient data, and the output is a data transfer completion notification to the server. Specifically, the terminal uses an HTTP POST request to send data to POST / api / patient_data.

[0153] Step 4:

[0154] The server stores the received data in a database (for example, MySQL or PostgreSQL). The input is patient data in JSON format, and the output is a notification that the data has been saved. Specifically, a backend program running on the server inserts the received data into the database. For example, it executes the SQL query INSERT INTO patient_data (data) VALUES ('{...}').

[0155] Step 5:

[0156] The server initializes multiple generative AI models. The input is the file path of each AI model, and the output is an instance of the initialized AI model. Specifically, the server loads the AI ​​model using the TensorFlow or PyTorch API as follows: model = tf.keras.models.load_model('path_to_model').

[0157] Step 6:

[0158] The server reads patient data from the database and inputs it into each generative AI model. The input is the patient ID, and the output is the result of applying the data to the AI ​​model. Specifically, the server executes the SQL query SELECT FROM patient_data WHERE patient_id = '12345' and inputs the obtained data into each generative AI model.

[0159] Step 7:

[0160] Each generative AI model provides its own opinion (treatment plan) based on patient data and discusses it with other generative AI models. The input is the patient data and opinions from other AI models, and the output is the results of the discussion. Specifically, the system simulates the process in which each generative AI model performs processing, integrates the results, and exchanges opinions.

[0161] Step 8:

[0162] The server aggregates the discussion results and generates the optimal treatment plan. The input is the discussion results, and the output is the optimal treatment plan. Specifically, the server analyzes the output of the AI ​​model and generates the treatment plan that "chemotherapy after surgery is optimal."

[0163] Step 9:

[0164] The server sends the proposed treatment plan to the terminal. The input is the optimal treatment plan, and the output is a notification of data transfer completion to the terminal. Specifically, the server sends the treatment plan data to the terminal using an HTTP POST request.

[0165] Step 10:

[0166] The proposed treatment plan received by the device is displayed to the user. The input is the proposed treatment plan, and the output is the display of the treatment plan to the user. Specifically, the web application or mobile application on the device displays the received data in a GUI.

[0167] Step 11:

[0168] Users (cancer board members) discuss the proposed treatment plans and decide on the final treatment plan. The input is the proposed treatment plan, and the output is the final treatment plan. In concrete terms, users use discussion tools or remote video conferencing systems to hold discussions, and then enter and save their final decision in a form within the system.

[0169] Step 12:

[0170] The terminal feeds back the final treatment plan it has decided to the server. The input is the final treatment plan, and the output is a notification to the server that feedback data has been sent. Specifically, the terminal sends the feedback data to the server using an HTTP POST request.

[0171] Step 13:

[0172] The server stores the feedback data in a database again and optimizes the generative AI model using reinforcement learning. The input is the feedback data, and the output is the updated result of the AI ​​model based on reinforcement learning. Specifically, the server retrains the AI ​​model based on the feedback data and executes code such as model.fit(feedback_data).

[0173] (Application example 1)

[0174] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0175] In conventional cancer board systems, the process of generating collective opinions and confirming treatment plans takes a lot of time and effort, making it difficult for medical professionals to respond quickly.In addition, it is difficult for specialists in remote locations to confirm treatment plans in real time, making it difficult to provide efficient medical care.

[0176] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0177] In this invention, the server includes a means for collecting patient data, a means for conducting discussions based on the collected data using an artificial intelligence model fine-tuned with multiple specialized knowledge, and a means for generating a treatment plan from the discussion results. This allows the proposed treatment plan to be displayed on a mobile communication terminal where it can be checked by a specialist in real time, enabling quick and efficient confirmation of the treatment plan for each patient and appropriate treatment.

[0178] "Patient data" refers to information about individual patients collected by a medical facility, including diagnostic results, test results, medical history, etc.

[0179] "Means of collection" refers to the mechanism for obtaining patient data from an electronic medical record system, etc., converting it into an appropriate format, and entering it into the system.

[0180] An "artificial intelligence model" refers to a machine learning model that has specific expertise and is fine-tuned based on that expertise; in the medical field, it simulates the judgment of a medical specialist.

[0181] "Means for conducting discussions" refers to the process in which multiple artificial intelligence models work together to analyze patient data, automatically exchange opinions from the perspective of each specialist, and discuss the optimal treatment plan.

[0182] "Means for generating a treatment plan" refers to the process of aggregating the results of discussions by the artificial intelligence model and creating a proposal for the most effective treatment method for the patient.

[0183] "Means for proposing to medical staff" refers to a mechanism for displaying the generated treatment plan to medical staff in an appropriate format and encouraging their confirmation and discussion.

[0184] "Feedback data" refers to the information that is input into the system regarding the treatment plan finalized by medical staff and stored for later use in reinforcement learning of the artificial intelligence model.

[0185] "Mobile communication terminals" include portable devices such as smartphones and tablets that allow specialists to check treatment plans in real time wherever they are.

[0186] The present invention relates to a system for efficiently processing patient data and proposing optimal treatment plans to specialists. The system includes a means for collecting patient data, a means for discussing the collected data with an artificial intelligence model fine-tuned by multiple specialized knowledge, and a means for generating optimal treatment plans.

[0187] System Configuration

[0188] This system consists of the following main elements:

[0189] Terminal: This terminal retrieves patient data from the electronic medical record system, converts it into an appropriate format, and sends it to the server.

[0190] Server: Discussions are held using AI models fine-tuned with multiple areas of expertise (e.g., AI surgeons, AI internists, AI radiologists, AI pathologists, etc.) to generate optimal treatment plans.

[0191] Mobile communication terminal: A device (such as a smartphone or tablet) that allows specialists to check the generated treatment plan in real time.

[0192] Data Collection Phase

[0193] 1. The terminal retrieves patient data from the electronic medical record system. For example, information on the diagnosis, test results, and medical history of a 50-year-old male patient A.

[0194] 2. The terminal converts the acquired patient data into an appropriate format (e.g., JSON format) and sends it to the server.

[0195] 3. The server stores the received data in the database.

[0196] AI Cancer Board Phase

[0197] 4. The server initializes an artificial intelligence model that is fine-tuned with multiple areas of expertise.

[0198] 5. The server reads patient data from the database and inputs the data into each specialist AI.

[0199] 6. Each AI specialist will propose a treatment plan based on patient data and discuss it with other AI specialists.

[0200] 7. The server aggregates the discussion results and generates an optimal treatment plan.

[0201] Proposal Phase

[0202] 8. The server sends the proposed treatment plan to the mobile communication terminal.

[0203] 9. The proposed treatment plan received by the mobile communication terminal is displayed to the specialist in real time, who then confirms and considers it.

[0204] Feedback Phase

[0205] 10. The final treatment plan will be decided based on the treatment plan reviewed by the specialist.

[0206] 11. The terminal feeds back the final treatment plan to the server.

[0207] 12. The server stores the feedback data in a database and optimizes the artificial intelligence model using reinforcement learning (RLHF).

[0208] Specific examples

[0209] For example, consider the case where Patient A is a 50-year-old man diagnosed with stage II lung cancer. This data is first obtained from the electronic medical record system by the device and sent to the server. The server receives this data and distributes it to various specialist AI models, such as the surgeon AI and internist AI. The surgeon AI determines whether surgery is appropriate, and the internist AI evaluates whether drug therapy is applicable. Based on the opinions of each specialist AI, the server generates a proposed treatment plan stating that "chemotherapy after surgery is optimal." This proposed treatment plan is sent to the mobile communication device and reviewed in real time by specialists.

[0210] Prompt Sentence Examples

[0211] "Please suggest a treatment plan for Patient A."

[0212] This system enables medical staff to quickly and accurately confirm treatment plans and take appropriate action, improving efficiency in medical settings and providing optimal care to patients.

[0213] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0214] Step 1:

[0215] The terminal retrieves patient data from the electronic medical record system.

[0216] Input: Patient basic information, diagnosis results, test results, medical history, etc.

[0217] Output: Retrieved patient data

[0218] Specific operation: Access the electronic medical record system, search for the specified patient's data, extract the necessary information, and convert this data into an appropriate format (e.g., JSON format).

[0219] Step 2:

[0220] The terminal transmits the acquired patient data to the server.

[0221] Input: Patient data in JSON format

[0222] Output: Patient data sent to the server

[0223] Specific operation: The terminal uses a network connection to send the converted patient data to the server, using a data communication protocol such as an HTTP POST request.

[0224] Step 3:

[0225] The server stores the received data in a database.

[0226] Input: Patient data sent to the server

[0227] Output: Patient data stored in a database

[0228] Specific operation: The server parses the received JSON format data and stores the contents in a database appropriately. It uses a database management system (e.g., MySQL or PostgreSQL).

[0229] Step 4:

[0230] The server initializes an artificial intelligence model that is fine-tuned with multiple areas of expertise.

[0231] Input: Patient data stored in the database

[0232] Output: Initialized artificial intelligence model

[0233] Specific operation: The server loads models such as surgeon AI, internist AI, radiologist AI, and pathologist AI, and prepares patient data to be input into these models.

[0234] Step 5:

[0235] The server reads patient data from the database and inputs the data into each specialist AI.

[0236] Input: Patient data read from database

[0237] Output: Patient data input to each specialist AI

[0238] Specific operation: The server extracts specific patient data from the database and inputs this data into each specialist AI. The input data is used as a prompt for each specialist AI.

[0239] Step 6:

[0240] Each AI specialist will propose a treatment plan based on patient data and hold discussions with other AI specialists.

[0241] Input: Patient data entered into the specialist AI

[0242] Output: Treatment options after discussion

[0243] Specific operation: Each AI specialist generates a treatment plan based on the prompts and exchanges opinions with other AI specialists. Through this process, discussions are held and candidate treatment plans are generated for consensus.

[0244] Step 7:

[0245] The server aggregates the discussion results and generates an optimal treatment plan.

[0246] Input: Treatment plan candidates

[0247] Output: Optimal treatment plan

[0248] Specific operation: The server selects the most suitable treatment plan from multiple treatment plan candidates using a statistical or algorithmic method and generates it as the optimal treatment plan.

[0249] Step 8:

[0250] The server transmits the generated proposed treatment plan to the mobile communication terminal.

[0251] Input: Optimal treatment plan

[0252] Output: Treatment plan sent to mobile communication terminal

[0253] Specific operation: The server sends the optimal treatment plan to the mobile communication device. HTTP POST and WebSocket are used as communication protocols.

[0254] Step 9:

[0255] The proposed treatment plan received by the mobile communication terminal is displayed in real time to the specialist, who then checks and considers it.

[0256] Input: Treatment plan sent to mobile communication terminal

[0257] Output: Confirmation and review results by specialists

[0258] Specific operation: The mobile communication terminal displays the received treatment plan on the screen in real time, and a specialist doctor reviews it and considers its appropriateness.

[0259] Step 10:

[0260] The final treatment plan will be decided based on the treatment options reviewed by the specialist.

[0261] Input: Confirmation and review results by specialists

[0262] Output: Final treatment plan

[0263] Specific operation: The specialists discuss the treatment plans displayed on the screen and select the most appropriate treatment plan.

[0264] Step 11:

[0265] The terminal feeds back the final treatment plan to the server.

[0266] Input: Final treatment plan

[0267] Output: Feedback sent to the server

[0268] Specific operation: The terminal inputs the final treatment plan and sends it to the server as feedback.

[0269] Step 12:

[0270] The server stores the feedback data in a database and optimizes the artificial intelligence model using reinforcement learning (RLHF).

[0271] Input: Feedback of final treatment plan

[0272] Output: An artificial intelligence model optimized by reinforcement learning

[0273] Specific operation: The server stores the feedback data in a database and uses this data to perform reinforcement learning on the artificial intelligence model, improving the accuracy and effectiveness of the model.

[0274] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0275] This invention specifically describes an AI-based cancer board system that combines an emotion engine that recognizes the user's emotions. The following explains the system's program processing in natural language, with specific examples.

[0276] System Configuration

[0277] The system consists of a terminal that collects and manages patient data, a server that processes the collected data and holds discussions with an artificial intelligence model with specialized knowledge, a terminal that proposes a final treatment plan and receives feedback, and an emotion engine that recognizes the user's emotions.

[0278] Data Collection Phase

[0279] 1. The terminal obtains patient data from the electronic medical record system. For example, it obtains the medical details (diagnosis results, test results, medical history, etc.) of Patient A, a 50-year-old male.

[0280] 2. The terminal converts the acquired patient data into an appropriate format (e.g., JSON format) and sends it to the server.

[0281] 3. The server stores the received data in the database.

[0282] AI Cancer Board Phase

[0283] 4. The server initializes multiple artificial intelligence models (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) that are fine-tuned with specialized knowledge.

[0284] 5. The server reads patient data from the database and inputs the data into each specialist AI.

[0285] 6. Each AI specialist will provide its own opinion (treatment plan) based on patient data and discuss it with other AI specialists.

[0286] 7. The server aggregates the discussion results and generates an optimal treatment plan.

[0287] Proposal Phase

[0288] 8. The server sends the proposed treatment plan to the terminal.

[0289] 9. The proposed treatment plan received by the terminal is displayed to the cancer board members (users).

[0290] 10. Users (Cancer Board members) discuss the proposed treatment plan, and the emotion engine analyzes users' emotions in real time, taking their emotional state into account when deciding on the final treatment plan.

[0291] Feedback Phase

[0292] 11. The terminal feeds back the final treatment plan decided upon to the server.

[0293] 12. The server stores the feedback data in a database and optimizes the AI ​​model using reinforcement learning (RLHF).

[0294] Emotional Engine Phase

[0295] 13. The device uses an emotion engine to collect and analyze the emotions of Cancer Board members (users) in real time. Specifically, the device evaluates the user's emotions based on speech content, facial expressions, heart rate, and other biometric data.

[0296] 14. The device sends the analysis results to the server, which are used to generate the final treatment plan and feedback for the discussion.

[0297] Specific examples

[0298] For example, consider the case where Patient A is a 50-year-old man diagnosed with stage II lung cancer. This data is first obtained from the electronic medical record system by the device and sent to the server. The server receives this data and distributes it to various specialist AI models, such as the surgeon AI and internist AI. The surgeon AI determines whether surgery is appropriate, and the internist AI evaluates whether drug therapy is applicable. Based on the opinions of each specialist AI, the server generates a proposed treatment plan that states, "Chemotherapy after surgery is optimal." This proposed treatment plan is sent to the device and reviewed by members of the cancer board. During the review, the emotion engine monitors the emotions of medical staff in real time and issues a warning if excessive stress is detected. If the final treatment plan is decided to be "immunotherapy in addition to chemotherapy after surgery," this information is fed back to the server and used for reinforcement learning in the AI ​​model.

[0299] In this way, the present invention not only reduces the workload of doctors and reduces regional variations in the quality of medical care, but also takes into account the emotions and stress of medical staff, providing a more humane and safe medical environment.

[0300] The processing flow will be explained below.

[0301] Step 1:

[0302] The terminal connects to the hospital's electronic medical record system and obtains the patient's medical data. Specifically, the patient ID is entered and an API is called to obtain the medical data (e.g., 50-year-old male, diagnosis: stage II lung cancer, medical history: high blood pressure).

[0303] Step 2:

[0304] The terminal formats the acquired data into an appropriate format (for example, JSON format) and sends it to the server. Specifically, the patient data is converted into JSON format and POSTed to the server using an HTTP request.

[0305] Step 3:

[0306] The server saves the received data in the database. Specifically, it executes an SQL query to insert the received patient data into the database.

[0307] Step 4:

[0308] The server initializes each specialist AI (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) in the AI ​​model, which has been fine-tuned with specialized knowledge. Specifically, it loads each specialist AI model into memory and performs the necessary settings (parameters and model weights).

[0309] Step 5:

[0310] The server reads patient data from the database and inputs it into each specialist AI. Specifically, it reads the patient data using an SQL query, converts it into a different format, and passes the input data to each specialist AI model.

[0311] Step 6:

[0312] Each AI specialist analyzes patient data from their own perspective and offers their own opinion (treatment plan). Specifically, the AI ​​surgeon determines whether surgery is appropriate, while the AI ​​internist evaluates whether drug therapy is applicable. Each AI model generates its own analysis results.

[0313] Step 7:

[0314] The server aggregates the results of the specialist AI discussions and generates the optimal treatment plan. Specifically, it integrates the analysis results and opinions of each specialist AI and determines the optimal treatment plan using a consensus-forming algorithm and weighting coefficients.

[0315] Step 8:

[0316] The server sends the proposed treatment plan to the terminal. Specifically, it converts the proposed treatment plan into JSON format and posts it to the terminal using an HTTP request.

[0317] Step 9:

[0318] The proposed treatment plan received by the device is displayed to the Cancer Board members. Specifically, the proposed treatment plan is displayed on the screen so that the user can confirm the proposal.

[0319] Step 10:

[0320] Users (Cancer Board members) discuss the proposals, and the emotion engine analyzes the user's emotions in real time. Specifically, sensors collect biometric data such as speech content, facial expressions, and heart rate, and an algorithm is run to evaluate the user's emotional state. A final treatment plan is then decided.

[0321] Step 11:

[0322] The terminal feeds back the final treatment plan it has decided to use to the server. Specifically, it converts the final treatment plan into JSON format and posts it to the server using an HTTP request.

[0323] Step 12:

[0324] The server stores the final treatment plan received in a database and optimizes the AI ​​model using reinforcement learning (RLHF). Specifically, it inserts the feedback data into the database, supplies new data to the AI ​​model, and executes the reinforcement learning algorithm.

[0325] Step 13:

[0326] The device uses an emotion engine to collect and analyze the emotions of Cancer Board members (users) in real time. Specifically, it evaluates the user's emotions based on speech content, facial expressions, heart rate, and other biometric data.

[0327] Step 14:

[0328] The device sends the analysis results to the server, which uses them to generate the final treatment plan and feedback for the discussion. Specifically, the emotion evaluation results generated by the emotion engine are converted into JSON format and posted to the server using an HTTP request.

[0329] Example 2

[0330] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0331] Conventional medical systems focus on collecting patient data and proposing treatment plans, but it is difficult to consider the emotions and stress levels of medical staff, which can affect the quality and safety of medical care. Furthermore, there are insufficient methods for utilizing collected feedback data and optimizing artificial intelligence models using reinforcement learning. A system that can solve these problems is needed.

[0332] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0333] In this invention, the server includes means for collecting patient data, means for holding discussions based on the collected data using an AI model fine-tuned with multiple specialized knowledge, means for generating a treatment plan from the discussion results, means for proposing the generated treatment plan to medical staff, means for saving the proposal results as feedback data and for reinforcement learning the AI ​​model, means for recognizing user emotions and analyzing emotional data, and means for utilizing the emotional data for feedback and adjustment of the treatment plan. This makes it possible to continuously propose optimal treatment plans while taking into account the emotions and stress of medical staff.

[0334] "Patient data" refers to all information about patients collected in the healthcare system, including diagnostic results, test results, medical history, and treatment history.

[0335] An "electronic medical record system" is a medical information system for managing, storing, and sharing patient data in digital format.

[0336] An "artificial intelligence model" is a machine learning algorithm that has specific expertise and uses that expertise to analyze data and make decisions.

[0337] "Discussion" is a process in which multiple artificial intelligence models exchange their opinions and judgments to determine the optimal treatment plan.

[0338] A "treatment policy" is a plan that indicates the most appropriate methods and procedures for medical examination and treatment according to the patient's condition.

[0339] "Medical staff" refers to professional workers who treat and care for patients in medical settings, such as doctors, nurses, and medical technicians.

[0340] "Feedback data" is data collected based on proposed treatment plans and implementation results, and is information used to improve artificial intelligence models.

[0341] "Reinforcement learning" is a machine learning technique that allows artificial intelligence models to learn through experience and improve their performance.

[0342] An "emotion engine" is a software or hardware mechanism that recognizes a user's emotions and analyzes them in real time.

[0343] "Emotion data" is data that indicates the emotional state of a user, collected from the content of the user's speech, facial expressions, heart rate, and the like.

[0344] This invention is an AI cancer board system that collects and analyzes patient data, and proposes optimal treatment plans through discussions using artificial intelligence models with multiple specialized knowledge. This system comprehensively supports everything from collecting patient data to proposing treatment plans, providing feedback, and analyzing the emotions of medical staff.

[0345] System Configuration

[0346] The system consists of the following components:

[0347] 1. Terminal for collecting and managing patient data

[0348] 2. A server that processes the collected data and holds discussions with AI models with specialized knowledge

[0349] 3. A device that proposes the final treatment plan and receives feedback

[0350] 4. Emotion engine that recognizes user emotions

[0351] Data Collection Phase

[0352] 1. The terminal retrieves patient data from the electronic medical record system. The terminal accesses the electronic medical record system and retrieves data such as diagnosis results, test results, and medical history of Patient A, a 50-year-old male.

[0353] 2. The device converts the acquired patient data into an appropriate format, such as JSON, and sends it to a server via the Internet.

[0354] 3. The server saves the received data in the database and logs the message "Patient A's data has been saved."

[0355] AI Cancer Board Phase

[0356] 1. The server initializes multiple AI models, including surgeon AI, internist AI, radiologist AI, and pathologist AI, and fine-tunes each model with specialized knowledge.

[0357] 2. The server reads patient data from the database and inputs the data into each specialist AI.

[0358] 3. Each AI specialist will provide its own opinion (treatment plan) based on the patient data, discuss it with other AI specialists, and decide on the final recommendation.

[0359] Proposal Phase

[0360] 1. The server sends the generated treatment plan proposal to the terminal and notifies the terminal that "chemotherapy after surgery is optimal."

[0361] 2. The terminal displays this proposed treatment plan to the cancer board members and uses it as material for discussion.

[0362] Feedback Phase

[0363] 1. Users (Cancer Board members) will discuss the proposed treatment plan and decide on the final treatment plan.

[0364] 2. The terminal feeds back the final treatment plan to the server.

[0365] 3. The server stores this feedback data in a database and uses reinforcement learning (RLHF) techniques to optimize the AI ​​model.

[0366] Emotional Engine Phase

[0367] 1. The device uses an emotion engine to collect and analyze the user's emotions in real time. Specifically, it evaluates the user's emotions based on speech content, facial expressions, heart rate, and other biometric data.

[0368] 2. The device sends the analysis results to the server, which is used to generate the final treatment plan and provide feedback for the discussion.

[0369] Specific examples

[0370] For example, if Patient A, a 50-year-old male, is diagnosed with stage II lung cancer, the following process will occur:

[0371] 1. The terminal retrieves this data from the electronic medical record system and sends it to the server via the Internet.

[0372] 2. The server distributes this data to each specialist AI model, such as surgeon AI and internist AI.

[0373] 3. The AI ​​surgeon judges that "surgery is appropriate," and the AI ​​physician gives the opinion that "drug therapy is applicable." Based on these opinions, the final treatment plan is generated: "Chemotherapy after surgery is optimal."

[0374] 4. The device displays this proposed treatment plan to cancer board members and encourages discussion.

[0375] 5. During the discussion, the emotion engine monitors the emotions of medical staff in real time and may warn them that "excessive stress has been detected."

[0376] 6. If the final treatment plan is decided to be "immunotherapy in addition to chemotherapy after surgery," this information is again fed back to the server and used for reinforcement learning of the AI ​​model.

[0377] Examples of prompt statements

[0378] Enter the following prompt into the generative AI model:

[0379] Patient A is a 50-year-old man diagnosed with stage II lung cancer. The following data is provided: diagnosis, test results, medical history, etc. Each AI specialist must propose a treatment plan based on this data and decide on the final treatment plan.

[0380] In this way, the present invention can reduce the workload of doctors, reduce regional variations in the quality of medical care, take into consideration the emotions and stress of medical staff, and provide a more humane and safe medical environment.

[0381] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0382] Step 1:

[0383] The terminal retrieves patient data from the electronic medical record system.

[0384] Input: User-specified patient ID

[0385] Processing: The terminal accesses the electronic medical record system and retrieves data related to the specified patient ID, such as diagnosis results, test results, medical history, etc. For example, Patient A is a 50-year-old male who has been diagnosed with stage II lung cancer.

[0386] Output: Patient data (diagnosis results, test results, medical history, etc.)

[0387] Step 2:

[0388] The terminal converts the acquired patient data into an appropriate format (e.g., JSON format).

[0389] Input: Retrieved patient data

[0390] Processing: The terminal converts the patient data into JSON format, for example, { "name": "Patient A", "age": 50, "diagnosis": "Lung cancer stage II"}.

[0391] Output: Transformed patient data (JSON format)

[0392] Step 3:

[0393] The terminal transmits the converted data to the server.

[0394] Input: Transformed patient data (JSON format)

[0395] Processing: The device sends data to the server via the Internet.

[0396] Output: Data sent to the server

[0397] Step 4:

[0398] The server stores the received data in a database.

[0399] Input: Submitted patient data (JSON format)

[0400] Processing: The server connects to the database and saves the patient data. It leaves a log stating "Patient A's data has been saved."

[0401] Output: Patient data stored in a database

[0402] Step 5:

[0403] The server initializes an artificial intelligence model that is fine-tuned with expert knowledge.

[0404] Input: Fine-tuned AI model information

[0405] Processing: The server initializes multiple AI models, such as surgeon AI, internist AI, radiologist AI, and pathologist AI, and leaves a log stating, "Each AI model has been initialized."

[0406] Output: Initialized AI model

[0407] Step 6:

[0408] The server reads patient data from the database and inputs the data into each specialist AI.

[0409] Input: Patient data stored in the database

[0410] Processing: The server reads the patient data and inputs it into each specialist AI.

[0411] Output: Patient data input into each AI model

[0412] Step 7:

[0413] Each specialist AI will give their own opinion (treatment plan) based on patient data.

[0414] Input: Patient data entered into each specialist AI

[0415] Processing: The AI ​​surgeon determines that "surgery is appropriate," and the AI ​​internist gives the opinion that "drug therapy is applicable."

[0416] Output: Treatment plan issued by each AI

[0417] Step 8:

[0418] Each AI specialist will hold discussions with other AI specialists.

[0419] Input: Opinion of each specialist AI (treatment plan)

[0420] Processing: Each AI exchanges opinions within the server and decides on the final recommendation: "Chemotherapy after surgery is optimal."

[0421] Output: Optimal treatment plan

[0422] Step 9:

[0423] The proposed treatment plan generated by the server is sent to the terminal.

[0424] Input: Optimal treatment plan

[0425] Processing: The server sends the treatment plan to the terminal.

[0426] Output: Treatment plan sent to the terminal

[0427] Step 10:

[0428] The proposed treatment plan received by the device is displayed to the cancer board members (users).

[0429] Input: Treatment plan sent from the server

[0430] Processing: The device displays the treatment plan and notifies the user, "We suggest chemotherapy after surgery."

[0431] Output: Display of proposed treatment plan

[0432] Step 11:

[0433] Users will discuss the proposed treatment plan.

[0434] Input: proposed treatment plan

[0435] Process: Users hold a discussion based on the treatment plan. The discussion takes the form of "Do you have any additional opinions regarding this treatment plan?"

[0436] Output: Discussion transcript

[0437] Step 12:

[0438] The emotion engine analyzes user emotions in real time.

[0439] Input: User's speech, facial expressions, heart rate, and other biometric data

[0440] Processing: The emotion engine analyzes this data and assesses the user's stress level and emotional state.

[0441] Output: Parsed emotion data

[0442] Step 13:

[0443] The terminal feeds back the final treatment plan it has decided to use to the server.

[0444] Input: Cancer Board's final treatment plan

[0445] Processing: The terminal sends the final treatment plan to the server via the Internet.

[0446] Output: Final treatment plan sent to the server

[0447] Step 14:

[0448] The server stores the feedback data in a database and optimizes the AI ​​model through reinforcement learning.

[0449] Input: Feedback of final treatment plan data

[0450] Processing: The server stores this in a database and runs reinforcement learning (RLHF) to optimize the AI ​​model.

[0451] Output: Optimized AI model

[0452] (Application example 2)

[0453] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0454] The objective of this invention is to provide a technology that utilizes an emotion engine built into a conventional system for analyzing and discussing patient data to evaluate the emotional state of medical staff and passengers in real time, determine optimal treatment plans, and adjust autonomous driving, thereby improving the quality of medical care and the safety and comfort of passengers.

[0455] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0456] In this invention, the server includes means for collecting patient data, means for holding discussions based on the collected data using an AI model fine-tuned with multiple specialized knowledge, means for generating a treatment plan from the discussion results, means for proposing the generated treatment plan to medical staff, means for saving the proposal results as feedback data and for reinforcement learning the AI ​​model, means for acquiring the collected data using an in-vehicle terminal and evaluating the emotional state of passengers using an emotion engine, and means for adjusting the autonomous driving engine based on the evaluated emotional state. This makes it possible to determine a treatment plan that takes emotional states into consideration and achieve safe and comfortable autonomous driving while reducing the workload of medical staff.

[0457] "Patient data" refers to data that includes medical information such as a patient's diagnosis results, test results, medical history, and so on.

[0458] "Means of collection" refers to the means of obtaining and storing patient data from electronic medical record systems, etc.

[0459] A "fine-tuned AI model" is an AI model that has been specialised to have specialized knowledge, with parameters tuned for a specific purpose or role.

[0460] A "means for conducting discussions" is a means for multiple artificial intelligence models to exchange opinions with each other and execute a process to arrive at the optimal decision.

[0461] "Means for generating a treatment plan" refers to a means for determining a treatment plan and method based on the results of the discussion.

[0462] The "means for making suggestions to medical staff" is a means for presenting the generated treatment plan to medical staff.

[0463] "Means for reinforcement learning" refers to means for performing a learning process to improve the performance of an artificial intelligence model based on feedback data.

[0464] The "Emotion Engine" is an artificial intelligence model that recognizes and evaluates the emotional state of passengers and users in real time.

[0465] An "autonomous driving engine" is an artificial intelligence model and its execution system for autonomously driving and controlling a vehicle.

[0466] An "in-vehicle terminal" is a hardware device installed in an autonomous vehicle for acquiring and processing data.

[0467] "Collected data" refers to biometric data such as heart rate and facial expression data acquired by the in-vehicle terminal, as well as other associated data.

[0468] The "means for evaluating" is the means by which the emotion engine analyzes the emotional state of the passenger based on the collected data.

[0469] The "adjusting means" is a means for changing the autonomous vehicle's driving strategy based on the assessed emotional state.

[0470] MODE FOR CARRYING OUT THE INVENTION

[0471] System Configuration

[0472] The system of this invention combines passenger emotion recognition in an autonomous vehicle with autonomous driving adjustment. This system consists of a terminal that collects and manages patient data, a server that discusses the collected data with multiple AI models with specialized knowledge, an in-vehicle terminal that proposes a final driving policy and receives feedback, and an emotion engine that recognizes the user's emotions.

[0473] Data Collection Phase

[0474] The server collects biometric data such as facial expressions and heart rate of passengers through cameras and heart rate sensors on the in-vehicle terminals. This data is converted into an appropriate format (e.g., JSON format) and sent to the server.

[0475] Emotion Recognition Phase

[0476] The server stores the received biometric data in a database. It then initializes the emotion engine and inputs the collected data. The emotion engine evaluates the passenger's emotional state in real time and passes the results back to the server.

[0477] Autonomous driving phase

[0478] Based on the evaluation results of the emotion engine, the server will instruct the autonomous driving engine to adjust its driving policy. For example, if it determines that a passenger is feeling stressed, the autonomous driving engine can slow down the vehicle and play relaxing music in the car, ensuring a safe and comfortable riding experience for passengers.

[0479] Feedback Phase

[0480] The final driving plan is fed back from the on-board terminal to the server and stored in a database. The server uses the feedback data to reinforce learning of the artificial intelligence model and improve the system's performance.

[0481] Hardware and Software Used

[0482] Hardware: In-vehicle cameras, heart rate sensors, in-vehicle terminals

[0483] Software: EmotionRecognizer (emotion engine), OpenCV (image processing library), Requests (HTTP request library)

[0484] Specific examples

[0485] For example, when Passenger A boards an autonomous vehicle, the in-vehicle terminal uses a camera and heart rate sensor to capture facial expression and heart rate data. This data is sent to the server, where the emotion engine on the server analyzes Passenger A's emotional state. If Passenger A is determined to be stressed, the server will instruct the autonomous driving engine to slow down the vehicle and play relaxing music. This process allows Passenger A to enjoy a safer and more comfortable riding experience.

[0486] Prompt Sentence Examples

[0487] "Build a system that can identify passenger emotions in real time based on in-vehicle camera footage and heart rate data, and adjust the autonomous driving speed and route accordingly. For example, if a passenger feels stressed, the system can slow down the vehicle for safety and play relaxing music."

[0488] As described above, the system of the present invention enables operational management of autonomous vehicles that takes into account the emotional state of passengers, thereby improving comfort and safety.

[0489] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0490] Step 1:

[0491] The device uses a camera and heart rate sensor to capture passengers' facial expressions and heart rate data in real time.

[0492] Input: Passenger facial video, heart rate data

[0493] Data processing and calculation: The video data from the camera is divided into frames and facial features are extracted. The heart rate sensor periodically measures the heart rate and outputs it as digital data.

[0494] Output: Collected facial expression data and heart rate data

[0495] Step 2:

[0496] The facial expression data and heart rate data acquired by the device are converted into JSON format and sent to the server.

[0497] Input: facial expression data, heart rate data

[0498] Data processing and calculation: Serialization process is performed to convert data into JSON format.

[0499] Output: JSON formatted data packet

[0500] Step 3:

[0501] The server saves the received JSON format data in the database.

[0502] Input: JSON data packet

[0503] Data processing and calculation: Performs write operations on the database.

[0504] Output: Biometric data stored in a database

[0505] Step 4:

[0506] The server initializes the emotion engine and inputs biometric data obtained from the database to analyze the emotional state.

[0507] Input: Biometric data stored in a database

[0508] Data processing and calculation: The emotion engine (EmotionRecognizer) evaluates the emotional state based on the input data and generates emotion tags.

[0509] Output: Evaluated emotional state (emotion tag)

[0510] Step 5:

[0511] The server receives the evaluation results of the emotion engine and issues instructions to the autonomous driving engine to adjust the driving policy.

[0512] Input: Evaluated emotional state (emotion tag)

[0513] Data processing and calculation: Executes an algorithm that generates operational adjustment instructions corresponding to the emotional state.

[0514] Output: Operation adjustment instruction

[0515] Step 6:

[0516] The self-driving engine receives instructions from the server and adjusts speed and changes the in-car environmental settings.

[0517] Input: Operation adjustment instructions

[0518] Specific actions: Slowing down the vehicle, playing relaxing music, etc.

[0519] Output: Coordinated operating conditions

[0520] Step 7:

[0521] The final operation plan and its results are fed back to the server by the in-vehicle terminal.

[0522] Input: Adjusted operating conditions

[0523] Data processing and calculation: The operation result data is converted back into JSON format and sent to the server.

[0524] Output: Feedback data sent to the server

[0525] Step 8:

[0526] The server stores the feedback data in a database and performs reinforcement learning.

[0527] Input: Feedback data

[0528] Data processing and calculation: Improve the performance of AI models by running reinforcement learning algorithms based on feedback data.

[0529] Output: AI model optimized by reinforcement learning

[0530] Through the above processing steps, the system can analyze passengers' emotional states and optimize the operation of autonomous vehicles in real time.

[0531] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0532] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (registered trademark) (Internet search engine).<URL: https: / / openai.com / blog / chatgpt> ), Gemini (registered trademark) (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0533] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0534] [Second embodiment]

[0535] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0536] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0537] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0538] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0539] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0540] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0541] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0542] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0543] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0544] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0545] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0546] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0547] This invention specifically describes a cancer board system that utilizes AI. The following explains the system's program processing in natural language, and includes specific examples.

[0548] System Configuration

[0549] The system consists of a terminal that collects and manages patient data, a server that processes the collected data and holds discussions with an artificial intelligence model with specialized knowledge, and a terminal that proposes a final treatment plan and receives feedback.

[0550] Data Collection Phase

[0551] 1. The terminal obtains patient data from the electronic medical record system. For example, it obtains the medical details (diagnosis results, test results, medical history, etc.) of Patient A, a 50-year-old male.

[0552] 2. The terminal converts the acquired patient data into an appropriate format (e.g., JSON format) and sends it to the server.

[0553] 3. The server stores the received data in the database.

[0554] AI Cancer Board Phase

[0555] 4. The server initializes multiple artificial intelligence models (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) that are fine-tuned with specialized knowledge.

[0556] 5. The server reads patient data from the database and inputs the data into each specialist AI.

[0557] 6. Each AI specialist will provide its own opinion (treatment plan) based on patient data and discuss it with other AI specialists.

[0558] 7. The server aggregates the discussion results and generates an optimal treatment plan.

[0559] Proposal Phase

[0560] 8. The server sends the proposed treatment plan to the terminal.

[0561] 9. The proposed treatment plan received by the terminal is displayed to the cancer board members (users).

[0562] 10. Users (Cancer Board members) discuss the proposed treatment plan and decide on the final treatment plan.

[0563] Feedback Phase

[0564] 11. The terminal feeds back the final treatment plan decided upon to the server.

[0565] 12. The server stores the feedback data in a database and optimizes the artificial intelligence model using reinforcement learning (RLHF).

[0566] Specific examples

[0567] For example, consider the case where Patient A is a 50-year-old man diagnosed with stage II lung cancer. This data is first obtained from the electronic medical record system by the device and sent to the server. The server receives this data and distributes it to various specialist AI models, such as the surgeon AI and internist AI. The surgeon AI determines whether surgery is appropriate, and the internist AI evaluates whether drug therapy is applicable. Based on the opinions of each specialist AI, the server generates a proposed treatment plan that states, "Chemotherapy after surgery is optimal." This proposed treatment plan is sent to the device and reviewed by members of the cancer board. If the final treatment plan is decided to be "immunotherapy in addition to chemotherapy after surgery," this information is fed back to the server and used for reinforcement learning in the AI ​​model.

[0568] In this way, the present invention can reduce the workload of doctors and reduce regional variations in the quality of medical care.

[0569] The processing flow will be explained below.

[0570] Step 1:

[0571] The terminal connects to the hospital's electronic medical record system and obtains the patient's medical data. Specifically, the patient ID is entered and an API is called to obtain the medical data (e.g., 50-year-old male, diagnosis: stage II lung cancer, medical history: high blood pressure).

[0572] Step 2:

[0573] The device formats the acquired data into an appropriate format and sends it to the server. Specifically, it converts the patient data into JSON format and posts it to the server using an HTTP request.

[0574] Step 3:

[0575] The server saves the received data in the database. Specifically, it executes an SQL query to insert the received patient data into the database.

[0576] Step 4:

[0577] The server initializes each specialist AI (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) in the AI ​​model, which has been fine-tuned with specialized knowledge. Specifically, it loads and configures each specialist AI model.

[0578] Step 5:

[0579] The server reads patient data from the database and inputs the data into each specialist AI. Specifically, it reads the patient data using an SQL query, converts the data into a format, and passes it to each specialist AI.

[0580] Step 6:

[0581] Each AI specialist will analyze patient data from their own perspective and offer their own opinion (treatment plan). Specific actions include the surgeon AI determining whether surgery is appropriate, and the internist AI evaluating whether drug therapy is applicable.

[0582] Step 7:

[0583] The server aggregates the results of the specialist AI discussions and generates the optimal treatment plan. Specifically, it collects the opinions of each specialist AI and applies voting and consensus-forming algorithms to determine the optimal treatment plan.

[0584] Step 8:

[0585] The server sends the generated treatment plan proposal to the terminal. Specifically, it converts the treatment plan proposal into JSON format and posts it to the terminal using an HTTP request.

[0586] Step 9:

[0587] The terminal displays the proposed treatment plan received to the cancer board members. Specifically, the received proposed treatment plan is displayed on the screen so that the cancer board members can refer to it.

[0588] Step 10:

[0589] Users (cancer board members) discuss the proposals and decide on the final treatment plan. Specifically, the cancer board members review the proposed treatment plan and reach a final consensus.

[0590] Step 11:

[0591] The terminal feeds back the final treatment plan it has decided to use to the server. Specifically, it converts the final treatment plan into JSON format and posts it to the server using an HTTP request.

[0592] Step 12:

[0593] The server stores the final treatment plan received in a database and optimizes the AI ​​model using reinforcement learning (RLHF). Specifically, it inserts the feedback data into the database, supplies new data to the AI ​​model, and executes the reinforcement learning algorithm.

[0594] Example 1

[0595] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0596] In the conventional medical system, the process of deciding on a treatment plan takes time, and the inconsistency of expertise and a lack of understanding of the information make it difficult to quickly decide on the optimal treatment plan.In addition, the management and use of collected patient data was not centralized, and collaboration between medical specialties was often ineffective.

[0597] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0598] In this invention, the server includes means for acquiring patient data from an electronic system, means for converting the acquired data into an appropriate format and sending and storing it in a high-performance database, means for holding discussions using generative AI models fine-tuned with multiple pieces of specialized knowledge based on the stored data and for each expert knowledge model to simulate opinions, means for aggregating the discussion results to generate an optimal treatment plan, means for sending the generated treatment plan to the user and displaying it in a dedicated application, and means for the user to decide on a final treatment plan, sending that feedback to the system and storing it in the database, and for reinforcement learning of the generative AI model.This enables centralized management of medical data and rapid decision-making on a treatment plan.

[0599] "Patient data" refers to information about a specific patient, such as diagnostic results, test results, treatment history, and medical history, collected by a medical institution.

[0600] "Electronic system" means a system for managing and processing data in digital form, including computer systems and electronic medical record systems installed in medical institutions.

[0601] A "database" is a high-performance information system for efficiently storing, managing, and searching collected data.

[0602] A "generative AI model" is an artificial intelligence model that has been trained using deep learning techniques to have specific expertise, such as an AI surgeon or AI internist.

[0603] "Discussion" is the process in which multiple generative AI models exchange opinions based on patient data to arrive at the optimal treatment plan.

[0604] A "treatment plan" is a specific medical procedure or treatment plan proposed based on the patient's diagnostic and test results.

[0605] "Users" are medical staff and specialists who use the system.

[0606] "Reinforcement learning" is a machine learning technique that allows artificial intelligence models to improve their performance based on feedback data.

[0607] "Feedback" is the process of returning the results of proposed treatment plans and information after their implementation to the system and using it for subsequent discussions and model updates.

[0608] The present invention provides an AI-based cancer monitoring system that comprehensively collects, processes, analyzes, and provides feedback on medical data, with the aim of quickly determining optimal treatment plans. Specific embodiments of this system are described below.

[0609] System Configuration

[0610] The system consists of a terminal that collects and manages patient data, a server that processes the collected data and holds discussions with a generative AI model with specialized knowledge, and a terminal that proposes a final treatment plan and receives feedback.

[0611] Data collection and format conversion

[0612] 1. The terminal uses a PC or tablet used in a medical institution to obtain patient data from an electronic system (electronic medical record system). For example, the medical details (diagnosis results, test results, medical history, etc.) of Patient A, a 50-year-old male, are obtained.

[0613] 2. The terminal converts the acquired patient data into an appropriate format (e.g., JSON format) and sends it to the server. This conversion is performed using a programming language such as Python. For example, the patient data is converted into the following JSON format:

[0614] json

[0615] {

[0616] "patient_id": "12345",

[0617] "age": 50,

[0618] "gender": "male",

[0619] "diagnosis": "lung cancer stage II",

[0620] "medical_history": ["hypertension", "diabetes"],

[0621] "latest_results": {

[0622] "CT_scan": "positive",

[0623] "blood_test": "normal"

[0624] }

[0625] }

[0626] 3. The server stores the received data in a high-performance database (such as MySQL or PostgreSQL) using SQL queries.

[0627] AI Cancer Board Discussion

[0628] 4. The server initializes multiple generative AI models (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) fine-tuned with TensorFlow or PyTorch. Specifically, a command like model = tf.keras.models.load_model('path_to_model') is assumed.

[0629] 5. The server reads the patient data from the database and inputs it into each generative AI model. This process is also implemented using scripts such as Python.

[0630] 6. Each generative AI model provides an opinion (treatment plan) based on the patient data and discusses it with other generative AI models. For example, an internist AI might suggest that "chemotherapy is optimal," and an surgeon AI might respond that "surgery should also be used."

[0631] 7. The server aggregates the results of the discussions and generates the optimal treatment plan. For example, the treatment plan generated might be "chemotherapy after surgery is optimal."

[0632] Treatment plan suggestions and feedback

[0633] 8. The server sends the generated treatment plan to the device and displays it in a dedicated application. This process uses an HTTP POST request and a secure communication protocol (HTTPS).

[0634] 9. The device displays the received treatment plan proposal to the user (cancer board member). This is done using a web application or mobile application.

[0635] 10. Users (cancer board members) discuss the proposed treatment plan and decide on the final treatment plan. This process uses dedicated discussion tools and remote video conferencing systems (Zoom, Teams, etc.).

[0636] 11. The device sends the final treatment plan back to the server. The feedback data is also sent in JSON format.

[0637] 12. The server stores the feedback data in a database again and optimizes the generative AI model using reinforcement learning with human feedback (RLHF). This advances the AI ​​model's learning, improving the accuracy of the next discussion and treatment plan proposal.

[0638] Specific examples

[0639] For example, if Patient A is a 50-year-old man diagnosed with stage II lung cancer, the data is retrieved from the electronic system by the terminal, converted to JSON format, and sent to the server. The server distributes the data to each generative AI model, with the surgeon AI determining whether surgery is appropriate and the internist AI evaluating whether drug therapy is applicable. As a result of the discussion, the server generates a treatment plan proposal stating that "chemotherapy after surgery is optimal" and sends it to the terminal. The cancer board members consider this and ultimately decide on a treatment plan of "immunotherapy in addition to chemotherapy after surgery," and this information is fed back to the server. The AI ​​model uses this feedback to perform reinforcement learning, improving the accuracy of future proposals.

[0640] Prompt Sentence Examples

[0641] "Patient A is a 50-year-old man who has been diagnosed with stage II lung cancer. His medical information includes his latest test results and medical history. Each AI specialist should propose a treatment plan based on this information."

[0642] In this way, the present invention can reduce the workload of doctors and reduce regional variations in the quality of medical care.

[0643] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0644] Step 1:

[0645] The terminal retrieves patient data from an electronic system (electronic medical record system). The input is the patient ID and treatment request, and the output is the raw patient data. Specifically, the terminal sends a GET / patients / {patient_id} request to the electronic medical record system via the API to retrieve the patient's medical information (for example, the diagnosis results, test results, medical history, etc. of 50-year-old male Patient A).

[0646] Step 2:

[0647] The terminal converts the acquired patient data into an appropriate format (for example, JSON format). The input is raw patient data, and the output is JSON formatted patient data. Specifically, a script running on the terminal parses the raw data and converts it into JSON format as follows:

[0648] json

[0649] {

[0650] "patient_id": "12345",

[0651] "age": 50,

[0652] "gender": "male",

[0653] "diagnosis": "lung cancer stage II",

[0654] "medical_history": ["hypertension", "diabetes"],

[0655] "latest_results": {

[0656] "CT_scan": "positive",

[0657] "blood_test": "normal"

[0658] }

[0659] }

[0660] Step 3:

[0661] The terminal sends the converted data to the server. The input is JSON-formatted patient data, and the output is a data transfer completion notification to the server. Specifically, the terminal uses an HTTP POST request to send data to POST / api / patient_data.

[0662] Step 4:

[0663] The server stores the received data in a database (for example, MySQL or PostgreSQL). The input is patient data in JSON format, and the output is a notification that the data has been saved. Specifically, a backend program running on the server inserts the received data into the database. For example, it executes the SQL query INSERT INTO patient_data (data) VALUES ('{...}').

[0664] Step 5:

[0665] The server initializes multiple generative AI models. The input is the file path of each AI model, and the output is an instance of the initialized AI model. Specifically, the server loads the AI ​​model using the TensorFlow or PyTorch API as follows: model = tf.keras.models.load_model('path_to_model').

[0666] Step 6:

[0667] The server reads patient data from the database and inputs it into each generative AI model. The input is the patient ID, and the output is the result of applying the data to the AI ​​model. Specifically, the server executes the SQL query SELECT FROM patient_data WHERE patient_id = '12345' and inputs the obtained data into each generative AI model.

[0668] Step 7:

[0669] Each generative AI model provides its own opinion (treatment plan) based on patient data and discusses it with other generative AI models. The input is the patient data and opinions from other AI models, and the output is the results of the discussion. Specifically, the system simulates the process in which each generative AI model performs processing, integrates the results, and exchanges opinions.

[0670] Step 8:

[0671] The server aggregates the discussion results and generates the optimal treatment plan. The input is the discussion results, and the output is the optimal treatment plan. Specifically, the server analyzes the output of the AI ​​model and generates the treatment plan that "chemotherapy after surgery is optimal."

[0672] Step 9:

[0673] The server sends the proposed treatment plan to the terminal. The input is the optimal treatment plan, and the output is a notification of data transfer completion to the terminal. Specifically, the server sends the treatment plan data to the terminal using an HTTP POST request.

[0674] Step 10:

[0675] The proposed treatment plan received by the device is displayed to the user. The input is the proposed treatment plan, and the output is the display of the treatment plan to the user. Specifically, the web application or mobile application on the device displays the received data in a GUI.

[0676] Step 11:

[0677] Users (cancer board members) discuss the proposed treatment plans and decide on the final treatment plan. The input is the proposed treatment plan, and the output is the final treatment plan. In concrete terms, users use discussion tools or remote video conferencing systems to hold discussions, and then enter and save their final decision in a form within the system.

[0678] Step 12:

[0679] The terminal feeds back the final treatment plan it has decided to the server. The input is the final treatment plan, and the output is a notification to the server that feedback data has been sent. Specifically, the terminal sends the feedback data to the server using an HTTP POST request.

[0680] Step 13:

[0681] The server stores the feedback data in a database again and optimizes the generative AI model using reinforcement learning. The input is the feedback data, and the output is the updated result of the AI ​​model based on reinforcement learning. Specifically, the server retrains the AI ​​model based on the feedback data and executes code such as model.fit(feedback_data).

[0682] (Application example 1)

[0683] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0684] In conventional cancer board systems, the process of generating collective opinions and confirming treatment plans takes a lot of time and effort, making it difficult for medical professionals to respond quickly.In addition, it is difficult for specialists in remote locations to confirm treatment plans in real time, making it difficult to provide efficient medical care.

[0685] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0686] In this invention, the server includes a means for collecting patient data, a means for conducting discussions based on the collected data using an artificial intelligence model fine-tuned with multiple specialized knowledge, and a means for generating a treatment plan from the discussion results. This allows the proposed treatment plan to be displayed on a mobile communication terminal where it can be checked by a specialist in real time, enabling quick and efficient confirmation of the treatment plan for each patient and appropriate treatment.

[0687] "Patient data" refers to information about individual patients collected by a medical facility, including diagnostic results, test results, medical history, etc.

[0688] "Means of collection" refers to the mechanism for obtaining patient data from an electronic medical record system, etc., converting it into an appropriate format, and entering it into the system.

[0689] An "artificial intelligence model" refers to a machine learning model that has specific expertise and is fine-tuned based on that expertise; in the medical field, it simulates the judgment of a medical specialist.

[0690] "Means for conducting discussions" refers to the process in which multiple artificial intelligence models work together to analyze patient data, automatically exchange opinions from the perspective of each specialist, and discuss the optimal treatment plan.

[0691] "Means for generating a treatment plan" refers to the process of aggregating the results of discussions by the artificial intelligence model and creating a proposal for the most effective treatment method for the patient.

[0692] "Means for proposing to medical staff" refers to a mechanism for displaying the generated treatment plan to medical staff in an appropriate format and encouraging their confirmation and discussion.

[0693] "Feedback data" refers to the information that is input into the system regarding the treatment plan finalized by medical staff and stored for later use in reinforcement learning of the artificial intelligence model.

[0694] "Mobile communication terminals" include portable devices such as smartphones and tablets that allow specialists to check treatment plans in real time wherever they are.

[0695] The present invention relates to a system for efficiently processing patient data and proposing optimal treatment plans to specialists. The system includes a means for collecting patient data, a means for discussing the collected data with an artificial intelligence model fine-tuned by multiple specialized knowledge, and a means for generating optimal treatment plans.

[0696] System Configuration

[0697] This system consists of the following main elements:

[0698] Terminal: This terminal retrieves patient data from the electronic medical record system, converts it into an appropriate format, and sends it to the server.

[0699] Server: Discussions are held using AI models fine-tuned with multiple areas of expertise (e.g., AI surgeons, AI internists, AI radiologists, AI pathologists, etc.) to generate optimal treatment plans.

[0700] Mobile communication terminal: A device (such as a smartphone or tablet) that allows specialists to check the generated treatment plan in real time.

[0701] Data Collection Phase

[0702] 1. The terminal retrieves patient data from the electronic medical record system. For example, information on the diagnosis, test results, and medical history of a 50-year-old male patient A.

[0703] 2. The terminal converts the acquired patient data into an appropriate format (e.g., JSON format) and sends it to the server.

[0704] 3. The server stores the received data in the database.

[0705] AI Cancer Board Phase

[0706] 4. The server initializes an artificial intelligence model that is fine-tuned with multiple areas of expertise.

[0707] 5. The server reads patient data from the database and inputs the data into each specialist AI.

[0708] 6. Each AI specialist will propose a treatment plan based on patient data and discuss it with other AI specialists.

[0709] 7. The server aggregates the discussion results and generates an optimal treatment plan.

[0710] Proposal Phase

[0711] 8. The server sends the proposed treatment plan to the mobile communication terminal.

[0712] 9. The proposed treatment plan received by the mobile communication terminal is displayed to the specialist in real time, who then confirms and considers it.

[0713] Feedback Phase

[0714] 10. The final treatment plan will be decided based on the treatment plan reviewed by the specialist.

[0715] 11. The terminal feeds back the final treatment plan to the server.

[0716] 12. The server stores the feedback data in a database and optimizes the artificial intelligence model using reinforcement learning (RLHF).

[0717] Specific examples

[0718] For example, consider the case where Patient A is a 50-year-old man diagnosed with stage II lung cancer. This data is first obtained from the electronic medical record system by the device and sent to the server. The server receives this data and distributes it to various specialist AI models, such as the surgeon AI and internist AI. The surgeon AI determines whether surgery is appropriate, and the internist AI evaluates whether drug therapy is applicable. Based on the opinions of each specialist AI, the server generates a proposed treatment plan stating that "chemotherapy after surgery is optimal." This proposed treatment plan is sent to the mobile communication device and reviewed in real time by specialists.

[0719] Prompt Sentence Examples

[0720] "Please suggest a treatment plan for Patient A."

[0721] This system enables medical staff to quickly and accurately confirm treatment plans and take appropriate action, improving efficiency in medical settings and providing optimal care to patients.

[0722] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0723] Step 1:

[0724] The terminal retrieves patient data from the electronic medical record system.

[0725] Input: Patient basic information, diagnosis results, test results, medical history, etc.

[0726] Output: Retrieved patient data

[0727] Specific operation: Access the electronic medical record system, search for the specified patient's data, extract the necessary information, and convert this data into an appropriate format (e.g., JSON format).

[0728] Step 2:

[0729] The terminal transmits the acquired patient data to the server.

[0730] Input: Patient data in JSON format

[0731] Output: Patient data sent to the server

[0732] Specific operation: The terminal uses a network connection to send the converted patient data to the server, using a data communication protocol such as an HTTP POST request.

[0733] Step 3:

[0734] The server stores the received data in a database.

[0735] Input: Patient data sent to the server

[0736] Output: Patient data stored in a database

[0737] Specific operation: The server parses the received JSON format data and stores the contents in a database appropriately. It uses a database management system (e.g., MySQL or PostgreSQL).

[0738] Step 4:

[0739] The server initializes an artificial intelligence model that is fine-tuned with multiple areas of expertise.

[0740] Input: Patient data stored in the database

[0741] Output: Initialized artificial intelligence model

[0742] Specific operation: The server loads models such as surgeon AI, internist AI, radiologist AI, and pathologist AI, and prepares patient data to be input into these models.

[0743] Step 5:

[0744] The server reads patient data from the database and inputs the data into each specialist AI.

[0745] Input: Patient data read from database

[0746] Output: Patient data input to each specialist AI

[0747] Specific operation: The server extracts specific patient data from the database and inputs this data into each specialist AI. The input data is used as a prompt for each specialist AI.

[0748] Step 6:

[0749] Each AI specialist will propose a treatment plan based on patient data and hold discussions with other AI specialists.

[0750] Input: Patient data entered into the specialist AI

[0751] Output: Treatment options after discussion

[0752] Specific operation: Each AI specialist generates a treatment plan based on the prompts and exchanges opinions with other AI specialists. Through this process, discussions are held and candidate treatment plans are generated for consensus.

[0753] Step 7:

[0754] The server aggregates the discussion results and generates an optimal treatment plan.

[0755] Input: Treatment plan candidates

[0756] Output: Optimal treatment plan

[0757] Specific operation: The server selects the most suitable treatment plan from multiple treatment plan candidates using a statistical or algorithmic method and generates it as the optimal treatment plan.

[0758] Step 8:

[0759] The server transmits the generated proposed treatment plan to the mobile communication terminal.

[0760] Input: Optimal treatment plan

[0761] Output: Treatment plan sent to mobile communication terminal

[0762] Specific operation: The server sends the optimal treatment plan to the mobile communication device. HTTP POST and WebSocket are used as communication protocols.

[0763] Step 9:

[0764] The proposed treatment plan received by the mobile communication terminal is displayed in real time to the specialist, who then checks and considers it.

[0765] Input: Treatment plan sent to mobile communication terminal

[0766] Output: Confirmation and review results by specialists

[0767] Specific operation: The mobile communication terminal displays the received treatment plan on the screen in real time, and a specialist doctor reviews it and considers its appropriateness.

[0768] Step 10:

[0769] The final treatment plan will be decided based on the treatment options reviewed by the specialist.

[0770] Input: Confirmation and review results by specialists

[0771] Output: Final treatment plan

[0772] Specific operation: The specialists discuss the treatment plans displayed on the screen and select the most appropriate treatment plan.

[0773] Step 11:

[0774] The terminal feeds back the final treatment plan to the server.

[0775] Input: Final treatment plan

[0776] Output: Feedback sent to the server

[0777] Specific operation: The terminal inputs the final treatment plan and sends it to the server as feedback.

[0778] Step 12:

[0779] The server stores the feedback data in a database and optimizes the artificial intelligence model using reinforcement learning (RLHF).

[0780] Input: Feedback of final treatment plan

[0781] Output: An artificial intelligence model optimized by reinforcement learning

[0782] Specific operation: The server stores the feedback data in a database and uses this data to perform reinforcement learning on the artificial intelligence model, improving the accuracy and effectiveness of the model.

[0783] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0784] This invention specifically describes an AI-based cancer board system that combines an emotion engine that recognizes the user's emotions. The following explains the system's program processing in natural language, with specific examples.

[0785] System Configuration

[0786] The system consists of a terminal that collects and manages patient data, a server that processes the collected data and holds discussions with an artificial intelligence model with specialized knowledge, a terminal that proposes a final treatment plan and receives feedback, and an emotion engine that recognizes the user's emotions.

[0787] Data Collection Phase

[0788] 1. The terminal obtains patient data from the electronic medical record system. For example, it obtains the medical details (diagnosis results, test results, medical history, etc.) of Patient A, a 50-year-old male.

[0789] 2. The terminal converts the acquired patient data into an appropriate format (e.g., JSON format) and sends it to the server.

[0790] 3. The server stores the received data in the database.

[0791] AI Cancer Board Phase

[0792] 4. The server initializes multiple artificial intelligence models (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) that are fine-tuned with specialized knowledge.

[0793] 5. The server reads patient data from the database and inputs the data into each specialist AI.

[0794] 6. Each AI specialist will provide its own opinion (treatment plan) based on patient data and discuss it with other AI specialists.

[0795] 7. The server aggregates the discussion results and generates an optimal treatment plan.

[0796] Proposal Phase

[0797] 8. The server sends the proposed treatment plan to the terminal.

[0798] 9. The proposed treatment plan received by the terminal is displayed to the cancer board members (users).

[0799] 10. Users (Cancer Board members) discuss the proposed treatment plan, and the emotion engine analyzes users' emotions in real time, taking their emotional state into account when deciding on the final treatment plan.

[0800] Feedback Phase

[0801] 11. The terminal feeds back the final treatment plan decided upon to the server.

[0802] 12. The server stores the feedback data in a database and optimizes the AI ​​model using reinforcement learning (RLHF).

[0803] Emotional Engine Phase

[0804] 13. The device uses an emotion engine to collect and analyze the emotions of Cancer Board members (users) in real time. Specifically, the device evaluates the user's emotions based on speech content, facial expressions, heart rate, and other biometric data.

[0805] 14. The device sends the analysis results to the server, which are used to generate the final treatment plan and feedback for the discussion.

[0806] Specific examples

[0807] For example, consider the case where Patient A is a 50-year-old man diagnosed with stage II lung cancer. This data is first obtained from the electronic medical record system by the device and sent to the server. The server receives this data and distributes it to various specialist AI models, such as the surgeon AI and internist AI. The surgeon AI determines whether surgery is appropriate, and the internist AI evaluates whether drug therapy is applicable. Based on the opinions of each specialist AI, the server generates a proposed treatment plan that states, "Chemotherapy after surgery is optimal." This proposed treatment plan is sent to the device and reviewed by members of the cancer board. During the review, the emotion engine monitors the emotions of medical staff in real time and issues a warning if excessive stress is detected. If the final treatment plan is decided to be "immunotherapy in addition to chemotherapy after surgery," this information is fed back to the server and used for reinforcement learning in the AI ​​model.

[0808] In this way, the present invention not only reduces the workload of doctors and reduces regional variations in the quality of medical care, but also takes into account the emotions and stress of medical staff, providing a more humane and safe medical environment.

[0809] The processing flow will be explained below.

[0810] Step 1:

[0811] The terminal connects to the hospital's electronic medical record system and obtains the patient's medical data. Specifically, the patient ID is entered and an API is called to obtain the medical data (e.g., 50-year-old male, diagnosis: stage II lung cancer, medical history: high blood pressure).

[0812] Step 2:

[0813] The terminal formats the acquired data into an appropriate format (for example, JSON format) and sends it to the server. Specifically, the patient data is converted into JSON format and POSTed to the server using an HTTP request.

[0814] Step 3:

[0815] The server saves the received data in the database. Specifically, it executes an SQL query to insert the received patient data into the database.

[0816] Step 4:

[0817] The server initializes each specialist AI (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) in the AI ​​model, which has been fine-tuned with specialized knowledge. Specifically, it loads each specialist AI model into memory and performs the necessary settings (parameters and model weights).

[0818] Step 5:

[0819] The server reads patient data from the database and inputs it into each specialist AI. Specifically, it reads the patient data using an SQL query, converts it into a different format, and passes the input data to each specialist AI model.

[0820] Step 6:

[0821] Each AI specialist analyzes patient data from their own perspective and offers their own opinion (treatment plan). Specifically, the AI ​​surgeon determines whether surgery is appropriate, while the AI ​​internist evaluates whether drug therapy is applicable. Each AI model generates its own analysis results.

[0822] Step 7:

[0823] The server aggregates the results of the specialist AI discussions and generates the optimal treatment plan. Specifically, it integrates the analysis results and opinions of each specialist AI and determines the optimal treatment plan using a consensus-forming algorithm and weighting coefficients.

[0824] Step 8:

[0825] The server sends the proposed treatment plan to the terminal. Specifically, it converts the proposed treatment plan into JSON format and posts it to the terminal using an HTTP request.

[0826] Step 9:

[0827] The proposed treatment plan received by the device is displayed to the Cancer Board members. Specifically, the proposed treatment plan is displayed on the screen so that the user can confirm the proposal.

[0828] Step 10:

[0829] Users (Cancer Board members) discuss the proposals, and the emotion engine analyzes the user's emotions in real time. Specifically, sensors collect biometric data such as speech content, facial expressions, and heart rate, and an algorithm is run to evaluate the user's emotional state. A final treatment plan is then decided.

[0830] Step 11:

[0831] The terminal feeds back the final treatment plan it has decided to use to the server. Specifically, it converts the final treatment plan into JSON format and posts it to the server using an HTTP request.

[0832] Step 12:

[0833] The server stores the final treatment plan received in a database and optimizes the AI ​​model using reinforcement learning (RLHF). Specifically, it inserts the feedback data into the database, supplies new data to the AI ​​model, and executes the reinforcement learning algorithm.

[0834] Step 13:

[0835] The device uses an emotion engine to collect and analyze the emotions of Cancer Board members (users) in real time. Specifically, it evaluates the user's emotions based on speech content, facial expressions, heart rate, and other biometric data.

[0836] Step 14:

[0837] The device sends the analysis results to the server, which uses them to generate the final treatment plan and feedback for the discussion. Specifically, the emotion evaluation results generated by the emotion engine are converted into JSON format and posted to the server using an HTTP request.

[0838] Example 2

[0839] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0840] Conventional medical systems focus on collecting patient data and proposing treatment plans, but it is difficult to consider the emotions and stress levels of medical staff, which can affect the quality and safety of medical care. Furthermore, there are insufficient methods for utilizing collected feedback data and optimizing artificial intelligence models using reinforcement learning. A system that can solve these problems is needed.

[0841] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0842] In this invention, the server includes means for collecting patient data, means for holding discussions based on the collected data using an AI model fine-tuned with multiple specialized knowledge, means for generating a treatment plan from the discussion results, means for proposing the generated treatment plan to medical staff, means for saving the proposal results as feedback data and for reinforcement learning the AI ​​model, means for recognizing user emotions and analyzing emotional data, and means for utilizing the emotional data for feedback and adjustment of the treatment plan. This makes it possible to continuously propose optimal treatment plans while taking into account the emotions and stress of medical staff.

[0843] "Patient data" refers to all information about patients collected in the healthcare system, including diagnostic results, test results, medical history, and treatment history.

[0844] An "electronic medical record system" is a medical information system for managing, storing, and sharing patient data in digital format.

[0845] An "artificial intelligence model" is a machine learning algorithm that has specific expertise and uses that expertise to analyze data and make decisions.

[0846] "Discussion" is a process in which multiple artificial intelligence models exchange their opinions and judgments to determine the optimal treatment plan.

[0847] A "treatment policy" is a plan that indicates the most appropriate methods and procedures for medical examination and treatment according to the patient's condition.

[0848] "Medical staff" refers to professional workers who treat and care for patients in medical settings, such as doctors, nurses, and medical technicians.

[0849] "Feedback data" is data collected based on proposed treatment plans and implementation results, and is information used to improve artificial intelligence models.

[0850] "Reinforcement learning" is a machine learning technique that allows artificial intelligence models to learn through experience and improve their performance.

[0851] An "emotion engine" is a software or hardware mechanism that recognizes a user's emotions and analyzes them in real time.

[0852] "Emotion data" is data that indicates the emotional state of a user, collected from the content of the user's speech, facial expressions, heart rate, and the like.

[0853] This invention is an AI cancer board system that collects and analyzes patient data, and proposes optimal treatment plans through discussions using artificial intelligence models with multiple specialized knowledge. This system comprehensively supports everything from collecting patient data to proposing treatment plans, providing feedback, and analyzing the emotions of medical staff.

[0854] System Configuration

[0855] The system consists of the following components:

[0856] 1. Terminal for collecting and managing patient data

[0857] 2. A server that processes the collected data and holds discussions with AI models with specialized knowledge

[0858] 3. A device that proposes the final treatment plan and receives feedback

[0859] 4. Emotion engine that recognizes user emotions

[0860] Data Collection Phase

[0861] 1. The terminal retrieves patient data from the electronic medical record system. The terminal accesses the electronic medical record system and retrieves data such as diagnosis results, test results, and medical history of Patient A, a 50-year-old male.

[0862] 2. The device converts the acquired patient data into an appropriate format, such as JSON, and sends it to a server via the Internet.

[0863] 3. The server saves the received data in the database and logs the message "Patient A's data has been saved."

[0864] AI Cancer Board Phase

[0865] 1. The server initializes multiple AI models, including surgeon AI, internist AI, radiologist AI, and pathologist AI, and fine-tunes each model with specialized knowledge.

[0866] 2. The server reads patient data from the database and inputs the data into each specialist AI.

[0867] 3. Each AI specialist will provide its own opinion (treatment plan) based on the patient data, discuss it with other AI specialists, and decide on the final recommendation.

[0868] Proposal Phase

[0869] 1. The server sends the generated treatment plan proposal to the terminal and notifies the terminal that "chemotherapy after surgery is optimal."

[0870] 2. The terminal displays this proposed treatment plan to the cancer board members and uses it as material for discussion.

[0871] Feedback Phase

[0872] 1. Users (Cancer Board members) will discuss the proposed treatment plan and decide on the final treatment plan.

[0873] 2. The terminal feeds back the final treatment plan to the server.

[0874] 3. The server stores this feedback data in a database and uses reinforcement learning (RLHF) techniques to optimize the AI ​​model.

[0875] Emotional Engine Phase

[0876] 1. The device uses an emotion engine to collect and analyze the user's emotions in real time. Specifically, it evaluates the user's emotions based on speech content, facial expressions, heart rate, and other biometric data.

[0877] 2. The device sends the analysis results to the server, which is used to generate the final treatment plan and provide feedback for the discussion.

[0878] Specific examples

[0879] For example, if Patient A, a 50-year-old male, is diagnosed with stage II lung cancer, the following process will occur:

[0880] 1. The terminal retrieves this data from the electronic medical record system and sends it to the server via the Internet.

[0881] 2. The server distributes this data to each specialist AI model, such as surgeon AI and internist AI.

[0882] 3. The AI ​​surgeon judges that "surgery is appropriate," and the AI ​​physician gives the opinion that "drug therapy is applicable." Based on these opinions, the final treatment plan is generated: "Chemotherapy after surgery is optimal."

[0883] 4. The device displays this proposed treatment plan to cancer board members and encourages discussion.

[0884] 5. During the discussion, the emotion engine monitors the emotions of medical staff in real time and may warn them that "excessive stress has been detected."

[0885] 6. If the final treatment plan is decided to be "immunotherapy in addition to chemotherapy after surgery," this information is again fed back to the server and used for reinforcement learning of the AI ​​model.

[0886] Examples of prompt statements

[0887] Enter the following prompt into the generative AI model:

[0888] Patient A is a 50-year-old man diagnosed with stage II lung cancer. The following data is provided: diagnosis, test results, medical history, etc. Each AI specialist must propose a treatment plan based on this data and decide on the final treatment plan.

[0889] In this way, the present invention can reduce the workload of doctors, reduce regional variations in the quality of medical care, take into consideration the emotions and stress of medical staff, and provide a more humane and safe medical environment.

[0890] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0891] Step 1:

[0892] The terminal retrieves patient data from the electronic medical record system.

[0893] Input: User-specified patient ID

[0894] Processing: The terminal accesses the electronic medical record system and retrieves data related to the specified patient ID, such as diagnosis results, test results, medical history, etc. For example, Patient A is a 50-year-old male who has been diagnosed with stage II lung cancer.

[0895] Output: Patient data (diagnosis results, test results, medical history, etc.)

[0896] Step 2:

[0897] The terminal converts the acquired patient data into an appropriate format (e.g., JSON format).

[0898] Input: Retrieved patient data

[0899] Processing: The terminal converts the patient data into JSON format, for example, { "name": "Patient A", "age": 50, "diagnosis": "Lung cancer stage II"}.

[0900] Output: Transformed patient data (JSON format)

[0901] Step 3:

[0902] The terminal transmits the converted data to the server.

[0903] Input: Transformed patient data (JSON format)

[0904] Processing: The device sends data to the server via the Internet.

[0905] Output: Data sent to the server

[0906] Step 4:

[0907] The server stores the received data in a database.

[0908] Input: Submitted patient data (JSON format)

[0909] Processing: The server connects to the database and saves the patient data. It leaves a log stating "Patient A's data has been saved."

[0910] Output: Patient data stored in a database

[0911] Step 5:

[0912] The server initializes an artificial intelligence model that is fine-tuned with expert knowledge.

[0913] Input: Fine-tuned AI model information

[0914] Processing: The server initializes multiple AI models, such as surgeon AI, internist AI, radiologist AI, and pathologist AI, and leaves a log stating, "Each AI model has been initialized."

[0915] Output: Initialized AI model

[0916] Step 6:

[0917] The server reads patient data from the database and inputs the data into each specialist AI.

[0918] Input: Patient data stored in the database

[0919] Processing: The server reads the patient data and inputs it into each specialist AI.

[0920] Output: Patient data input into each AI model

[0921] Step 7:

[0922] Each specialist AI will give their own opinion (treatment plan) based on patient data.

[0923] Input: Patient data entered into each specialist AI

[0924] Processing: The AI ​​surgeon determines that "surgery is appropriate," and the AI ​​internist gives the opinion that "drug therapy is applicable."

[0925] Output: Treatment plan issued by each AI

[0926] Step 8:

[0927] Each AI specialist will hold discussions with other AI specialists.

[0928] Input: Opinion of each specialist AI (treatment plan)

[0929] Processing: Each AI exchanges opinions within the server and decides on the final recommendation: "Chemotherapy after surgery is optimal."

[0930] Output: Optimal treatment plan

[0931] Step 9:

[0932] The proposed treatment plan generated by the server is sent to the terminal.

[0933] Input: Optimal treatment plan

[0934] Processing: The server sends the treatment plan to the terminal.

[0935] Output: Treatment plan sent to the terminal

[0936] Step 10:

[0937] The proposed treatment plan received by the device is displayed to the cancer board members (users).

[0938] Input: Treatment plan sent from the server

[0939] Processing: The device displays the treatment plan and notifies the user, "We suggest chemotherapy after surgery."

[0940] Output: Display of proposed treatment plan

[0941] Step 11:

[0942] Users will discuss the proposed treatment plan.

[0943] Input: proposed treatment plan

[0944] Process: Users hold a discussion based on the treatment plan. The discussion takes the form of "Do you have any additional opinions regarding this treatment plan?"

[0945] Output: Discussion transcript

[0946] Step 12:

[0947] The emotion engine analyzes user emotions in real time.

[0948] Input: User's speech, facial expressions, heart rate, and other biometric data

[0949] Processing: The emotion engine analyzes this data and assesses the user's stress level and emotional state.

[0950] Output: Parsed emotion data

[0951] Step 13:

[0952] The terminal feeds back the final treatment plan it has decided to use to the server.

[0953] Input: Cancer Board's final treatment plan

[0954] Processing: The terminal sends the final treatment plan to the server via the Internet.

[0955] Output: Final treatment plan sent to the server

[0956] Step 14:

[0957] The server stores the feedback data in a database and optimizes the AI ​​model through reinforcement learning.

[0958] Input: Feedback of final treatment plan data

[0959] Processing: The server stores this in a database and runs reinforcement learning (RLHF) to optimize the AI ​​model.

[0960] Output: Optimized AI model

[0961] (Application example 2)

[0962] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0963] The objective of this invention is to provide a technology that utilizes an emotion engine built into a conventional system for analyzing and discussing patient data to evaluate the emotional state of medical staff and passengers in real time, determine optimal treatment plans, and adjust autonomous driving, thereby improving the quality of medical care and the safety and comfort of passengers.

[0964] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0965] In this invention, the server includes means for collecting patient data, means for holding discussions based on the collected data using an AI model fine-tuned with multiple specialized knowledge, means for generating a treatment plan from the discussion results, means for proposing the generated treatment plan to medical staff, means for saving the proposal results as feedback data and for reinforcement learning the AI ​​model, means for acquiring the collected data using an in-vehicle terminal and evaluating the emotional state of passengers using an emotion engine, and means for adjusting the autonomous driving engine based on the evaluated emotional state. This makes it possible to determine a treatment plan that takes emotional states into consideration and achieve safe and comfortable autonomous driving while reducing the workload of medical staff.

[0966] "Patient data" refers to data that includes medical information such as a patient's diagnosis results, test results, medical history, and so on.

[0967] "Means of collection" refers to the means of obtaining and storing patient data from electronic medical record systems, etc.

[0968] A "fine-tuned AI model" is an AI model that has been specialised to have specialized knowledge, with parameters tuned for a specific purpose or role.

[0969] A "means for conducting discussions" is a means for multiple artificial intelligence models to exchange opinions with each other and execute a process to arrive at the optimal decision.

[0970] "Means for generating a treatment plan" refers to a means for determining a treatment plan and method based on the results of the discussion.

[0971] The "means for making suggestions to medical staff" is a means for presenting the generated treatment plan to medical staff.

[0972] "Means for reinforcement learning" refers to means for performing a learning process to improve the performance of an artificial intelligence model based on feedback data.

[0973] The "Emotion Engine" is an artificial intelligence model that recognizes and evaluates the emotional state of passengers and users in real time.

[0974] An "autonomous driving engine" is an artificial intelligence model and its execution system for autonomously driving and controlling a vehicle.

[0975] An "in-vehicle terminal" is a hardware device installed in an autonomous vehicle for acquiring and processing data.

[0976] "Collected data" refers to biometric data such as heart rate and facial expression data acquired by the in-vehicle terminal, as well as other associated data.

[0977] The "means for evaluating" is the means by which the emotion engine analyzes the emotional state of the passenger based on the collected data.

[0978] The "adjusting means" is a means for changing the autonomous vehicle's driving strategy based on the assessed emotional state.

[0979] MODE FOR CARRYING OUT THE INVENTION

[0980] System Configuration

[0981] The system of this invention combines passenger emotion recognition in an autonomous vehicle with autonomous driving adjustment. This system consists of a terminal that collects and manages patient data, a server that discusses the collected data with multiple AI models with specialized knowledge, an in-vehicle terminal that proposes a final driving policy and receives feedback, and an emotion engine that recognizes the user's emotions.

[0982] Data Collection Phase

[0983] The server collects biometric data such as facial expressions and heart rate of passengers through cameras and heart rate sensors on the in-vehicle terminals. This data is converted into an appropriate format (e.g., JSON format) and sent to the server.

[0984] Emotion Recognition Phase

[0985] The server stores the received biometric data in a database. It then initializes the emotion engine and inputs the collected data. The emotion engine evaluates the passenger's emotional state in real time and passes the results back to the server.

[0986] Autonomous driving phase

[0987] Based on the evaluation results of the emotion engine, the server will instruct the autonomous driving engine to adjust its driving policy. For example, if it determines that a passenger is feeling stressed, the autonomous driving engine can slow down the vehicle and play relaxing music in the car, ensuring a safe and comfortable riding experience for passengers.

[0988] Feedback Phase

[0989] The final driving plan is fed back from the on-board terminal to the server and stored in a database. The server uses the feedback data to reinforce learning of the artificial intelligence model and improve the system's performance.

[0990] Hardware and Software Used

[0991] Hardware: In-vehicle cameras, heart rate sensors, in-vehicle terminals

[0992] Software: EmotionRecognizer (emotion engine), OpenCV (image processing library), Requests (HTTP request library)

[0993] Specific examples

[0994] For example, when Passenger A boards an autonomous vehicle, the in-vehicle terminal uses a camera and heart rate sensor to capture facial expression and heart rate data. This data is sent to the server, where the emotion engine on the server analyzes Passenger A's emotional state. If Passenger A is determined to be stressed, the server will instruct the autonomous driving engine to slow down the vehicle and play relaxing music. This process allows Passenger A to enjoy a safer and more comfortable riding experience.

[0995] Prompt Sentence Examples

[0996] "Build a system that can identify passenger emotions in real time based on in-vehicle camera footage and heart rate data, and adjust the autonomous driving speed and route accordingly. For example, if a passenger feels stressed, the system can slow down the vehicle for safety and play relaxing music."

[0997] As described above, the system of the present invention enables operational management of autonomous vehicles that takes into account the emotional state of passengers, thereby improving comfort and safety.

[0998] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0999] Step 1:

[1000] The device uses a camera and heart rate sensor to capture passengers' facial expressions and heart rate data in real time.

[1001] Input: Passenger facial video, heart rate data

[1002] Data processing and calculation: The video data from the camera is divided into frames and facial features are extracted. The heart rate sensor periodically measures the heart rate and outputs it as digital data.

[1003] Output: Collected facial expression data and heart rate data

[1004] Step 2:

[1005] The facial expression data and heart rate data acquired by the device are converted into JSON format and sent to the server.

[1006] Input: facial expression data, heart rate data

[1007] Data processing and calculation: Serialization process is performed to convert data into JSON format.

[1008] Output: JSON formatted data packet

[1009] Step 3:

[1010] The server saves the received JSON format data in the database.

[1011] Input: JSON data packet

[1012] Data processing and calculation: Performs write operations on the database.

[1013] Output: Biometric data stored in a database

[1014] Step 4:

[1015] The server initializes the emotion engine and inputs biometric data obtained from the database to analyze the emotional state.

[1016] Input: Biometric data stored in a database

[1017] Data processing and calculation: The emotion engine (EmotionRecognizer) evaluates the emotional state based on the input data and generates emotion tags.

[1018] Output: Evaluated emotional state (emotion tag)

[1019] Step 5:

[1020] The server receives the evaluation results of the emotion engine and issues instructions to the autonomous driving engine to adjust the driving policy.

[1021] Input: Evaluated emotional state (emotion tag)

[1022] Data processing and calculation: Executes an algorithm that generates operational adjustment instructions corresponding to the emotional state.

[1023] Output: Operation adjustment instruction

[1024] Step 6:

[1025] The self-driving engine receives instructions from the server and adjusts speed and changes the in-car environmental settings.

[1026] Input: Operation adjustment instructions

[1027] Specific actions: Slowing down the vehicle, playing relaxing music, etc.

[1028] Output: Coordinated operating conditions

[1029] Step 7:

[1030] The final operation plan and its results are fed back to the server by the in-vehicle terminal.

[1031] Input: Adjusted operating conditions

[1032] Data processing and calculation: The operation result data is converted back into JSON format and sent to the server.

[1033] Output: Feedback data sent to the server

[1034] Step 8:

[1035] The server stores the feedback data in a database and performs reinforcement learning.

[1036] Input: Feedback data

[1037] Data processing and calculation: Improve the performance of AI models by running reinforcement learning algorithms based on feedback data.

[1038] Output: AI model optimized by reinforcement learning

[1039] Through the above processing steps, the system can analyze passengers' emotional states and optimize the operation of autonomous vehicles in real time.

[1040] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1041] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1042] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[1043] [Third embodiment]

[1044] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[1045] 5, the data processing system 310 includes the data processing device 12 and a headset terminal 314. An example of the data processing device 12 is a server.

[1046] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1047] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[1048] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1049] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1050] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1051] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1052] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1053] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1054] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1055] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[1056] This invention specifically describes a cancer board system that utilizes AI. The following explains the system's program processing in natural language, and includes specific examples.

[1057] System Configuration

[1058] The system consists of a terminal that collects and manages patient data, a server that processes the collected data and holds discussions with an artificial intelligence model with specialized knowledge, and a terminal that proposes a final treatment plan and receives feedback.

[1059] Data Collection Phase

[1060] 1. The terminal obtains patient data from the electronic medical record system. For example, it obtains the medical details (diagnosis results, test results, medical history, etc.) of Patient A, a 50-year-old male.

[1061] 2. The terminal converts the acquired patient data into an appropriate format (e.g., JSON format) and sends it to the server.

[1062] 3. The server stores the received data in the database.

[1063] AI Cancer Board Phase

[1064] 4. The server initializes multiple artificial intelligence models (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) that are fine-tuned with specialized knowledge.

[1065] 5. The server reads patient data from the database and inputs the data into each specialist AI.

[1066] 6. Each AI specialist will provide its own opinion (treatment plan) based on patient data and discuss it with other AI specialists.

[1067] 7. The server aggregates the discussion results and generates an optimal treatment plan.

[1068] Proposal Phase

[1069] 8. The server sends the proposed treatment plan to the terminal.

[1070] 9. The proposed treatment plan received by the terminal is displayed to the cancer board members (users).

[1071] 10. Users (Cancer Board members) discuss the proposed treatment plan and decide on the final treatment plan.

[1072] Feedback Phase

[1073] 11. The terminal feeds back the final treatment plan decided upon to the server.

[1074] 12. The server stores the feedback data in a database and optimizes the artificial intelligence model using reinforcement learning (RLHF).

[1075] Specific examples

[1076] For example, consider the case where Patient A is a 50-year-old man diagnosed with stage II lung cancer. This data is first obtained from the electronic medical record system by the device and sent to the server. The server receives this data and distributes it to various specialist AI models, such as the surgeon AI and internist AI. The surgeon AI determines whether surgery is appropriate, and the internist AI evaluates whether drug therapy is applicable. Based on the opinions of each specialist AI, the server generates a proposed treatment plan that states, "Chemotherapy after surgery is optimal." This proposed treatment plan is sent to the device and reviewed by members of the cancer board. If the final treatment plan is decided to be "immunotherapy in addition to chemotherapy after surgery," this information is fed back to the server and used for reinforcement learning in the AI ​​model.

[1077] In this way, the present invention can reduce the workload of doctors and reduce regional variations in the quality of medical care.

[1078] The processing flow will be explained below.

[1079] Step 1:

[1080] The terminal connects to the hospital's electronic medical record system and obtains the patient's medical data. Specifically, the patient ID is entered and an API is called to obtain the medical data (e.g., 50-year-old male, diagnosis: stage II lung cancer, medical history: high blood pressure).

[1081] Step 2:

[1082] The device formats the acquired data into an appropriate format and sends it to the server. Specifically, it converts the patient data into JSON format and posts it to the server using an HTTP request.

[1083] Step 3:

[1084] The server saves the received data in the database. Specifically, it executes an SQL query to insert the received patient data into the database.

[1085] Step 4:

[1086] The server initializes each specialist AI (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) in the AI ​​model, which has been fine-tuned with specialized knowledge. Specifically, it loads and configures each specialist AI model.

[1087] Step 5:

[1088] The server reads patient data from the database and inputs the data into each specialist AI. Specifically, it reads the patient data using an SQL query, converts the data into a format, and passes it to each specialist AI.

[1089] Step 6:

[1090] Each AI specialist will analyze patient data from their own perspective and offer their own opinion (treatment plan). Specific actions include the surgeon AI determining whether surgery is appropriate, and the internist AI evaluating whether drug therapy is applicable.

[1091] Step 7:

[1092] The server aggregates the results of the specialist AI discussions and generates the optimal treatment plan. Specifically, it collects the opinions of each specialist AI and applies voting and consensus-forming algorithms to determine the optimal treatment plan.

[1093] Step 8:

[1094] The server sends the generated treatment plan proposal to the terminal. Specifically, it converts the treatment plan proposal into JSON format and posts it to the terminal using an HTTP request.

[1095] Step 9:

[1096] The terminal displays the proposed treatment plan received to the cancer board members. Specifically, the received proposed treatment plan is displayed on the screen so that the cancer board members can refer to it.

[1097] Step 10:

[1098] Users (cancer board members) discuss the proposals and decide on the final treatment plan. Specifically, the cancer board members review the proposed treatment plan and reach a final consensus.

[1099] Step 11:

[1100] The terminal feeds back the final treatment plan it has decided to use to the server. Specifically, it converts the final treatment plan into JSON format and posts it to the server using an HTTP request.

[1101] Step 12:

[1102] The server stores the final treatment plan received in a database and optimizes the AI ​​model using reinforcement learning (RLHF). Specifically, it inserts the feedback data into the database, supplies new data to the AI ​​model, and executes the reinforcement learning algorithm.

[1103] Example 1

[1104] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1105] In the conventional medical system, the process of deciding on a treatment plan takes time, and the inconsistency of expertise and a lack of understanding of the information make it difficult to quickly decide on the optimal treatment plan.In addition, the management and use of collected patient data was not centralized, and collaboration between medical specialties was often ineffective.

[1106] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1107] In this invention, the server includes means for acquiring patient data from an electronic system, means for converting the acquired data into an appropriate format and sending and storing it in a high-performance database, means for holding discussions using generative AI models fine-tuned with multiple pieces of specialized knowledge based on the stored data and for each expert knowledge model to simulate opinions, means for aggregating the discussion results to generate an optimal treatment plan, means for sending the generated treatment plan to the user and displaying it in a dedicated application, and means for the user to decide on a final treatment plan, sending that feedback to the system and storing it in the database, and for reinforcement learning of the generative AI model.This enables centralized management of medical data and rapid decision-making on a treatment plan.

[1108] "Patient data" refers to information about a specific patient, such as diagnostic results, test results, treatment history, and medical history, collected by a medical institution.

[1109] "Electronic system" means a system for managing and processing data in digital form, including computer systems and electronic medical record systems installed in medical institutions.

[1110] A "database" is a high-performance information system for efficiently storing, managing, and searching collected data.

[1111] A "generative AI model" is an artificial intelligence model that has been trained using deep learning techniques to have specific expertise, such as an AI surgeon or AI internist.

[1112] "Discussion" is the process in which multiple generative AI models exchange opinions based on patient data to arrive at the optimal treatment plan.

[1113] A "treatment plan" is a specific medical procedure or treatment plan proposed based on the patient's diagnostic and test results.

[1114] "Users" are medical staff and specialists who use the system.

[1115] "Reinforcement learning" is a machine learning technique that allows artificial intelligence models to improve their performance based on feedback data.

[1116] "Feedback" is the process of returning the results of proposed treatment plans and information after their implementation to the system and using it for subsequent discussions and model updates.

[1117] The present invention provides an AI-based cancer monitoring system that comprehensively collects, processes, analyzes, and provides feedback on medical data, with the aim of quickly determining optimal treatment plans. Specific embodiments of this system are described below.

[1118] System Configuration

[1119] The system consists of a terminal that collects and manages patient data, a server that processes the collected data and holds discussions with a generative AI model with specialized knowledge, and a terminal that proposes a final treatment plan and receives feedback.

[1120] Data collection and format conversion

[1121] 1. The terminal uses a PC or tablet used in a medical institution to obtain patient data from an electronic system (electronic medical record system). For example, the medical details (diagnosis results, test results, medical history, etc.) of Patient A, a 50-year-old male, are obtained.

[1122] 2. The terminal converts the acquired patient data into an appropriate format (e.g., JSON format) and sends it to the server. This conversion is performed using a programming language such as Python. For example, the patient data is converted into the following JSON format:

[1123] json

[1124] {

[1125] "patient_id": "12345",

[1126] "age": 50,

[1127] "gender": "male",

[1128] "diagnosis": "lung cancer stage II",

[1129] "medical_history": ["hypertension", "diabetes"],

[1130] "latest_results": {

[1131] "CT_scan": "positive",

[1132] "blood_test": "normal"

[1133] }

[1134] }

[1135] 3. The server stores the received data in a high-performance database (such as MySQL or PostgreSQL) using SQL queries.

[1136] AI Cancer Board Discussion

[1137] 4. The server initializes multiple generative AI models (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) fine-tuned with TensorFlow or PyTorch. Specifically, a command like model = tf.keras.models.load_model('path_to_model') is assumed.

[1138] 5. The server reads the patient data from the database and inputs it into each generative AI model. This process is also implemented using scripts such as Python.

[1139] 6. Each generative AI model provides an opinion (treatment plan) based on the patient data and discusses it with other generative AI models. For example, an internist AI might suggest that "chemotherapy is optimal," and an surgeon AI might respond that "surgery should also be used."

[1140] 7. The server aggregates the results of the discussions and generates the optimal treatment plan. For example, the treatment plan generated might be "chemotherapy after surgery is optimal."

[1141] Treatment plan suggestions and feedback

[1142] 8. The server sends the generated treatment plan to the device and displays it in a dedicated application. This process uses an HTTP POST request and a secure communication protocol (HTTPS).

[1143] 9. The device displays the received treatment plan proposal to the user (cancer board member). This is done using a web application or mobile application.

[1144] 10. Users (cancer board members) discuss the proposed treatment plan and decide on the final treatment plan. This process uses dedicated discussion tools and remote video conferencing systems (Zoom, Teams, etc.).

[1145] 11. The device sends the final treatment plan back to the server. The feedback data is also sent in JSON format.

[1146] 12. The server stores the feedback data in a database again and optimizes the generative AI model using reinforcement learning with human feedback (RLHF). This advances the AI ​​model's learning, improving the accuracy of the next discussion and treatment plan proposal.

[1147] Specific examples

[1148] For example, if Patient A is a 50-year-old man diagnosed with stage II lung cancer, the data is retrieved from the electronic system by the terminal, converted to JSON format, and sent to the server. The server distributes the data to each generative AI model, with the surgeon AI determining whether surgery is appropriate and the internist AI evaluating whether drug therapy is applicable. As a result of the discussion, the server generates a treatment plan proposal stating that "chemotherapy after surgery is optimal" and sends it to the terminal. The cancer board members consider this and ultimately decide on a treatment plan of "immunotherapy in addition to chemotherapy after surgery," and this information is fed back to the server. The AI ​​model uses this feedback to perform reinforcement learning, improving the accuracy of future proposals.

[1149] Prompt Sentence Examples

[1150] "Patient A is a 50-year-old man who has been diagnosed with stage II lung cancer. His medical information includes his latest test results and medical history. Each AI specialist should propose a treatment plan based on this information."

[1151] In this way, the present invention can reduce the workload of doctors and reduce regional variations in the quality of medical care.

[1152] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1153] Step 1:

[1154] The terminal retrieves patient data from an electronic system (electronic medical record system). The input is the patient ID and treatment request, and the output is the raw patient data. Specifically, the terminal sends a GET / patients / {patient_id} request to the electronic medical record system via the API to retrieve the patient's medical information (for example, the diagnosis results, test results, medical history, etc. of 50-year-old male Patient A).

[1155] Step 2:

[1156] The terminal converts the acquired patient data into an appropriate format (for example, JSON format). The input is raw patient data, and the output is JSON formatted patient data. Specifically, a script running on the terminal parses the raw data and converts it into JSON format as follows:

[1157] json

[1158] {

[1159] "patient_id": "12345",

[1160] "age": 50,

[1161] "gender": "male",

[1162] "diagnosis": "lung cancer stage II",

[1163] "medical_history": ["hypertension", "diabetes"],

[1164] "latest_results": {

[1165] "CT_scan": "positive",

[1166] "blood_test": "normal"

[1167] }

[1168] }

[1169] Step 3:

[1170] The terminal sends the converted data to the server. The input is JSON-formatted patient data, and the output is a data transfer completion notification to the server. Specifically, the terminal uses an HTTP POST request to send data to POST / api / patient_data.

[1171] Step 4:

[1172] The server stores the received data in a database (for example, MySQL or PostgreSQL). The input is patient data in JSON format, and the output is a notification that the data has been saved. Specifically, a backend program running on the server inserts the received data into the database. For example, it executes the SQL query INSERT INTO patient_data (data) VALUES ('{...}').

[1173] Step 5:

[1174] The server initializes multiple generative AI models. The input is the file path of each AI model, and the output is an instance of the initialized AI model. Specifically, the server loads the AI ​​model using the TensorFlow or PyTorch API as follows: model = tf.keras.models.load_model('path_to_model').

[1175] Step 6:

[1176] The server reads patient data from the database and inputs it into each generative AI model. The input is the patient ID, and the output is the result of applying the data to the AI ​​model. Specifically, the server executes the SQL query SELECT FROM patient_data WHERE patient_id = '12345' and inputs the obtained data into each generative AI model.

[1177] Step 7:

[1178] Each generative AI model provides its own opinion (treatment plan) based on patient data and discusses it with other generative AI models. The input is the patient data and opinions from other AI models, and the output is the results of the discussion. Specifically, the system simulates the process in which each generative AI model performs processing, integrates the results, and exchanges opinions.

[1179] Step 8:

[1180] The server aggregates the discussion results and generates the optimal treatment plan. The input is the discussion results, and the output is the optimal treatment plan. Specifically, the server analyzes the output of the AI ​​model and generates the treatment plan that "chemotherapy after surgery is optimal."

[1181] Step 9:

[1182] The server sends the proposed treatment plan to the terminal. The input is the optimal treatment plan, and the output is a notification of data transfer completion to the terminal. Specifically, the server sends the treatment plan data to the terminal using an HTTP POST request.

[1183] Step 10:

[1184] The proposed treatment plan received by the device is displayed to the user. The input is the proposed treatment plan, and the output is the display of the treatment plan to the user. Specifically, the web application or mobile application on the device displays the received data in a GUI.

[1185] Step 11:

[1186] Users (cancer board members) discuss the proposed treatment plans and decide on the final treatment plan. The input is the proposed treatment plan, and the output is the final treatment plan. In concrete terms, users use discussion tools or remote video conferencing systems to hold discussions, and then enter and save their final decision in a form within the system.

[1187] Step 12:

[1188] The terminal feeds back the final treatment plan it has decided to the server. The input is the final treatment plan, and the output is a notification to the server that feedback data has been sent. Specifically, the terminal sends the feedback data to the server using an HTTP POST request.

[1189] Step 13:

[1190] The server stores the feedback data in a database again and optimizes the generative AI model using reinforcement learning. The input is the feedback data, and the output is the updated result of the AI ​​model based on reinforcement learning. Specifically, the server retrains the AI ​​model based on the feedback data and executes code such as model.fit(feedback_data).

[1191] (Application example 1)

[1192] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1193] In conventional cancer board systems, the process of generating collective opinions and confirming treatment plans takes a lot of time and effort, making it difficult for medical professionals to respond quickly.In addition, it is difficult for specialists in remote locations to confirm treatment plans in real time, making it difficult to provide efficient medical care.

[1194] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1195] In this invention, the server includes a means for collecting patient data, a means for conducting discussions based on the collected data using an artificial intelligence model fine-tuned with multiple specialized knowledge, and a means for generating a treatment plan from the discussion results. This allows the proposed treatment plan to be displayed on a mobile communication terminal where it can be checked by a specialist in real time, enabling quick and efficient confirmation of the treatment plan for each patient and appropriate treatment.

[1196] "Patient data" refers to information about individual patients collected by a medical facility, including diagnostic results, test results, medical history, etc.

[1197] "Means of collection" refers to the mechanism for obtaining patient data from an electronic medical record system, etc., converting it into an appropriate format, and entering it into the system.

[1198] An "artificial intelligence model" refers to a machine learning model that has specific expertise and is fine-tuned based on that expertise; in the medical field, it simulates the judgment of a medical specialist.

[1199] "Means for conducting discussions" refers to the process in which multiple artificial intelligence models work together to analyze patient data, automatically exchange opinions from the perspective of each specialist, and discuss the optimal treatment plan.

[1200] "Means for generating a treatment plan" refers to the process of aggregating the results of discussions by the artificial intelligence model and creating a proposal for the most effective treatment method for the patient.

[1201] "Means for proposing to medical staff" refers to a mechanism for displaying the generated treatment plan to medical staff in an appropriate format and encouraging their confirmation and discussion.

[1202] "Feedback data" refers to the information that is input into the system regarding the treatment plan finalized by medical staff and stored for later use in reinforcement learning of the artificial intelligence model.

[1203] "Mobile communication terminals" include portable devices such as smartphones and tablets that allow specialists to check treatment plans in real time wherever they are.

[1204] The present invention relates to a system for efficiently processing patient data and proposing optimal treatment plans to specialists. The system includes a means for collecting patient data, a means for discussing the collected data with an artificial intelligence model fine-tuned by multiple specialized knowledge, and a means for generating optimal treatment plans.

[1205] System Configuration

[1206] This system consists of the following main elements:

[1207] Terminal: This terminal retrieves patient data from the electronic medical record system, converts it into an appropriate format, and sends it to the server.

[1208] Server: Discussions are held using AI models fine-tuned with multiple areas of expertise (e.g., AI surgeons, AI internists, AI radiologists, AI pathologists, etc.) to generate optimal treatment plans.

[1209] Mobile communication terminal: A device (such as a smartphone or tablet) that allows specialists to check the generated treatment plan in real time.

[1210] Data Collection Phase

[1211] 1. The terminal retrieves patient data from the electronic medical record system. For example, information on the diagnosis, test results, and medical history of a 50-year-old male patient A.

[1212] 2. The terminal converts the acquired patient data into an appropriate format (e.g., JSON format) and sends it to the server.

[1213] 3. The server stores the received data in the database.

[1214] AI Cancer Board Phase

[1215] 4. The server initializes an artificial intelligence model that is fine-tuned with multiple areas of expertise.

[1216] 5. The server reads patient data from the database and inputs the data into each specialist AI.

[1217] 6. Each AI specialist will propose a treatment plan based on patient data and discuss it with other AI specialists.

[1218] 7. The server aggregates the discussion results and generates an optimal treatment plan.

[1219] Proposal Phase

[1220] 8. The server sends the proposed treatment plan to the mobile communication terminal.

[1221] 9. The proposed treatment plan received by the mobile communication terminal is displayed to the specialist in real time, who then confirms and considers it.

[1222] Feedback Phase

[1223] 10. The final treatment plan will be decided based on the treatment plan reviewed by the specialist.

[1224] 11. The terminal feeds back the final treatment plan to the server.

[1225] 12. The server stores the feedback data in a database and optimizes the artificial intelligence model using reinforcement learning (RLHF).

[1226] Specific examples

[1227] For example, consider the case where Patient A is a 50-year-old man diagnosed with stage II lung cancer. This data is first obtained from the electronic medical record system by the device and sent to the server. The server receives this data and distributes it to various specialist AI models, such as the surgeon AI and internist AI. The surgeon AI determines whether surgery is appropriate, and the internist AI evaluates whether drug therapy is applicable. Based on the opinions of each specialist AI, the server generates a proposed treatment plan stating that "chemotherapy after surgery is optimal." This proposed treatment plan is sent to the mobile communication device and reviewed in real time by specialists.

[1228] Prompt Sentence Examples

[1229] "Please suggest a treatment plan for Patient A."

[1230] This system enables medical staff to quickly and accurately confirm treatment plans and take appropriate action, improving efficiency in medical settings and providing optimal care to patients.

[1231] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1232] Step 1:

[1233] The terminal retrieves patient data from the electronic medical record system.

[1234] Input: Patient basic information, diagnosis results, test results, medical history, etc.

[1235] Output: Retrieved patient data

[1236] Specific operation: Access the electronic medical record system, search for the specified patient's data, extract the necessary information, and convert this data into an appropriate format (e.g., JSON format).

[1237] Step 2:

[1238] The terminal transmits the acquired patient data to the server.

[1239] Input: Patient data in JSON format

[1240] Output: Patient data sent to the server

[1241] Specific operation: The terminal uses a network connection to send the converted patient data to the server, using a data communication protocol such as an HTTP POST request.

[1242] Step 3:

[1243] The server stores the received data in a database.

[1244] Input: Patient data sent to the server

[1245] Output: Patient data stored in a database

[1246] Specific operation: The server parses the received JSON format data and stores the contents in a database appropriately. It uses a database management system (e.g., MySQL or PostgreSQL).

[1247] Step 4:

[1248] The server initializes an artificial intelligence model that is fine-tuned with multiple areas of expertise.

[1249] Input: Patient data stored in the database

[1250] Output: Initialized artificial intelligence model

[1251] Specific operation: The server loads models such as surgeon AI, internist AI, radiologist AI, and pathologist AI, and prepares patient data to be input into these models.

[1252] Step 5:

[1253] The server reads patient data from the database and inputs the data into each specialist AI.

[1254] Input: Patient data read from database

[1255] Output: Patient data input to each specialist AI

[1256] Specific operation: The server extracts specific patient data from the database and inputs this data into each specialist AI. The input data is used as a prompt for each specialist AI.

[1257] Step 6:

[1258] Each AI specialist will propose a treatment plan based on patient data and hold discussions with other AI specialists.

[1259] Input: Patient data entered into the specialist AI

[1260] Output: Treatment options after discussion

[1261] Specific operation: Each AI specialist generates a treatment plan based on the prompts and exchanges opinions with other AI specialists. Through this process, discussions are held and candidate treatment plans are generated for consensus.

[1262] Step 7:

[1263] The server aggregates the discussion results and generates an optimal treatment plan.

[1264] Input: Treatment plan candidates

[1265] Output: Optimal treatment plan

[1266] Specific operation: The server selects the most suitable treatment plan from multiple treatment plan candidates using a statistical or algorithmic method and generates it as the optimal treatment plan.

[1267] Step 8:

[1268] The server transmits the generated proposed treatment plan to the mobile communication terminal.

[1269] Input: Optimal treatment plan

[1270] Output: Treatment plan sent to mobile communication terminal

[1271] Specific operation: The server sends the optimal treatment plan to the mobile communication device. HTTP POST and WebSocket are used as communication protocols.

[1272] Step 9:

[1273] The proposed treatment plan received by the mobile communication terminal is displayed in real time to the specialist, who then checks and considers it.

[1274] Input: Treatment plan sent to mobile communication terminal

[1275] Output: Confirmation and review results by specialists

[1276] Specific operation: The mobile communication terminal displays the received treatment plan on the screen in real time, and a specialist doctor reviews it and considers its appropriateness.

[1277] Step 10:

[1278] The final treatment plan will be decided based on the treatment options reviewed by the specialist.

[1279] Input: Confirmation and review results by specialists

[1280] Output: Final treatment plan

[1281] Specific operation: The specialists discuss the treatment plans displayed on the screen and select the most appropriate treatment plan.

[1282] Step 11:

[1283] The terminal feeds back the final treatment plan to the server.

[1284] Input: Final treatment plan

[1285] Output: Feedback sent to the server

[1286] Specific operation: The terminal inputs the final treatment plan and sends it to the server as feedback.

[1287] Step 12:

[1288] The server stores the feedback data in a database and optimizes the artificial intelligence model using reinforcement learning (RLHF).

[1289] Input: Feedback of final treatment plan

[1290] Output: An artificial intelligence model optimized by reinforcement learning

[1291] Specific operation: The server stores the feedback data in a database and uses this data to perform reinforcement learning on the artificial intelligence model, improving the accuracy and effectiveness of the model.

[1292] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1293] This invention specifically describes an AI-based cancer board system that combines an emotion engine that recognizes the user's emotions. The following explains the system's program processing in natural language, with specific examples.

[1294] System Configuration

[1295] The system consists of a terminal that collects and manages patient data, a server that processes the collected data and holds discussions with an artificial intelligence model with specialized knowledge, a terminal that proposes a final treatment plan and receives feedback, and an emotion engine that recognizes the user's emotions.

[1296] Data Collection Phase

[1297] 1. The terminal obtains patient data from the electronic medical record system. For example, it obtains the medical details (diagnosis results, test results, medical history, etc.) of Patient A, a 50-year-old male.

[1298] 2. The terminal converts the acquired patient data into an appropriate format (e.g., JSON format) and sends it to the server.

[1299] 3. The server stores the received data in the database.

[1300] AI Cancer Board Phase

[1301] 4. The server initializes multiple artificial intelligence models (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) that are fine-tuned with specialized knowledge.

[1302] 5. The server reads patient data from the database and inputs the data into each specialist AI.

[1303] 6. Each AI specialist will provide its own opinion (treatment plan) based on patient data and discuss it with other AI specialists.

[1304] 7. The server aggregates the discussion results and generates an optimal treatment plan.

[1305] Proposal Phase

[1306] 8. The server sends the proposed treatment plan to the terminal.

[1307] 9. The proposed treatment plan received by the terminal is displayed to the cancer board members (users).

[1308] 10. Users (Cancer Board members) discuss the proposed treatment plan, and the emotion engine analyzes users' emotions in real time, taking their emotional state into account when deciding on the final treatment plan.

[1309] Feedback Phase

[1310] 11. The terminal feeds back the final treatment plan decided upon to the server.

[1311] 12. The server stores the feedback data in a database and optimizes the AI ​​model using reinforcement learning (RLHF).

[1312] Emotional Engine Phase

[1313] 13. The device uses an emotion engine to collect and analyze the emotions of Cancer Board members (users) in real time. Specifically, the device evaluates the user's emotions based on speech content, facial expressions, heart rate, and other biometric data.

[1314] 14. The device sends the analysis results to the server, which are used to generate the final treatment plan and feedback for the discussion.

[1315] Specific examples

[1316] For example, consider the case where Patient A is a 50-year-old man diagnosed with stage II lung cancer. This data is first obtained from the electronic medical record system by the device and sent to the server. The server receives this data and distributes it to various specialist AI models, such as the surgeon AI and internist AI. The surgeon AI determines whether surgery is appropriate, and the internist AI evaluates whether drug therapy is applicable. Based on the opinions of each specialist AI, the server generates a proposed treatment plan that states, "Chemotherapy after surgery is optimal." This proposed treatment plan is sent to the device and reviewed by members of the cancer board. During the review, the emotion engine monitors the emotions of medical staff in real time and issues a warning if excessive stress is detected. If the final treatment plan is decided to be "immunotherapy in addition to chemotherapy after surgery," this information is fed back to the server and used for reinforcement learning in the AI ​​model.

[1317] In this way, the present invention not only reduces the workload of doctors and reduces regional variations in the quality of medical care, but also takes into account the emotions and stress of medical staff, providing a more humane and safe medical environment.

[1318] The processing flow will be explained below.

[1319] Step 1:

[1320] The terminal connects to the hospital's electronic medical record system and obtains the patient's medical data. Specifically, the patient ID is entered and an API is called to obtain the medical data (e.g., 50-year-old male, diagnosis: stage II lung cancer, medical history: high blood pressure).

[1321] Step 2:

[1322] The terminal formats the acquired data into an appropriate format (for example, JSON format) and sends it to the server. Specifically, the patient data is converted into JSON format and POSTed to the server using an HTTP request.

[1323] Step 3:

[1324] The server saves the received data in the database. Specifically, it executes an SQL query to insert the received patient data into the database.

[1325] Step 4:

[1326] The server initializes each specialist AI (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) in the AI ​​model, which has been fine-tuned with specialized knowledge. Specifically, it loads each specialist AI model into memory and performs the necessary settings (parameters and model weights).

[1327] Step 5:

[1328] The server reads patient data from the database and inputs it into each specialist AI. Specifically, it reads the patient data using an SQL query, converts it into a different format, and passes the input data to each specialist AI model.

[1329] Step 6:

[1330] Each AI specialist analyzes patient data from their own perspective and offers their own opinion (treatment plan). Specifically, the AI ​​surgeon determines whether surgery is appropriate, while the AI ​​internist evaluates whether drug therapy is applicable. Each AI model generates its own analysis results.

[1331] Step 7:

[1332] The server aggregates the results of the specialist AI discussions and generates the optimal treatment plan. Specifically, it integrates the analysis results and opinions of each specialist AI and determines the optimal treatment plan using a consensus-forming algorithm and weighting coefficients.

[1333] Step 8:

[1334] The server sends the proposed treatment plan to the terminal. Specifically, it converts the proposed treatment plan into JSON format and posts it to the terminal using an HTTP request.

[1335] Step 9:

[1336] The proposed treatment plan received by the device is displayed to the Cancer Board members. Specifically, the proposed treatment plan is displayed on the screen so that the user can confirm the proposal.

[1337] Step 10:

[1338] Users (Cancer Board members) discuss the proposals, and the emotion engine analyzes the user's emotions in real time. Specifically, sensors collect biometric data such as speech content, facial expressions, and heart rate, and an algorithm is run to evaluate the user's emotional state. A final treatment plan is then decided.

[1339] Step 11:

[1340] The terminal feeds back the final treatment plan it has decided to use to the server. Specifically, it converts the final treatment plan into JSON format and posts it to the server using an HTTP request.

[1341] Step 12:

[1342] The server stores the final treatment plan received in a database and optimizes the AI ​​model using reinforcement learning (RLHF). Specifically, it inserts the feedback data into the database, supplies new data to the AI ​​model, and executes the reinforcement learning algorithm.

[1343] Step 13:

[1344] The device uses an emotion engine to collect and analyze the emotions of Cancer Board members (users) in real time. Specifically, it evaluates the user's emotions based on speech content, facial expressions, heart rate, and other biometric data.

[1345] Step 14:

[1346] The device sends the analysis results to the server, which uses them to generate the final treatment plan and feedback for the discussion. Specifically, the emotion evaluation results generated by the emotion engine are converted into JSON format and posted to the server using an HTTP request.

[1347] Example 2

[1348] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1349] Conventional medical systems focus on collecting patient data and proposing treatment plans, but it is difficult to consider the emotions and stress levels of medical staff, which can affect the quality and safety of medical care. Furthermore, there are insufficient methods for utilizing collected feedback data and optimizing artificial intelligence models using reinforcement learning. A system that can solve these problems is needed.

[1350] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1351] In this invention, the server includes means for collecting patient data, means for holding discussions based on the collected data using an AI model fine-tuned with multiple specialized knowledge, means for generating a treatment plan from the discussion results, means for proposing the generated treatment plan to medical staff, means for saving the proposal results as feedback data and for reinforcement learning the AI ​​model, means for recognizing user emotions and analyzing emotional data, and means for utilizing the emotional data for feedback and adjustment of the treatment plan. This makes it possible to continuously propose optimal treatment plans while taking into account the emotions and stress of medical staff.

[1352] "Patient data" refers to all information about patients collected in the healthcare system, including diagnostic results, test results, medical history, and treatment history.

[1353] An "electronic medical record system" is a medical information system for managing, storing, and sharing patient data in digital format.

[1354] An "artificial intelligence model" is a machine learning algorithm that has specific expertise and uses that expertise to analyze data and make decisions.

[1355] "Discussion" is a process in which multiple artificial intelligence models exchange their opinions and judgments to determine the optimal treatment plan.

[1356] A "treatment policy" is a plan that indicates the most appropriate methods and procedures for medical examination and treatment according to the patient's condition.

[1357] "Medical staff" refers to professional workers who treat and care for patients in medical settings, such as doctors, nurses, and medical technicians.

[1358] "Feedback data" is data collected based on proposed treatment plans and implementation results, and is information used to improve artificial intelligence models.

[1359] "Reinforcement learning" is a machine learning technique that allows artificial intelligence models to learn through experience and improve their performance.

[1360] An "emotion engine" is a software or hardware mechanism that recognizes a user's emotions and analyzes them in real time.

[1361] "Emotion data" is data that indicates the emotional state of a user, collected from the content of the user's speech, facial expressions, heart rate, and the like.

[1362] This invention is an AI cancer board system that collects and analyzes patient data, and proposes optimal treatment plans through discussions using artificial intelligence models with multiple specialized knowledge. This system comprehensively supports everything from collecting patient data to proposing treatment plans, providing feedback, and analyzing the emotions of medical staff.

[1363] System Configuration

[1364] The system consists of the following components:

[1365] 1. Terminal for collecting and managing patient data

[1366] 2. A server that processes the collected data and holds discussions with AI models with specialized knowledge

[1367] 3. A device that proposes the final treatment plan and receives feedback

[1368] 4. Emotion engine that recognizes user emotions

[1369] Data Collection Phase

[1370] 1. The terminal retrieves patient data from the electronic medical record system. The terminal accesses the electronic medical record system and retrieves data such as diagnosis results, test results, and medical history of Patient A, a 50-year-old male.

[1371] 2. The device converts the acquired patient data into an appropriate format, such as JSON, and sends it to a server via the Internet.

[1372] 3. The server saves the received data in the database and logs the message "Patient A's data has been saved."

[1373] AI Cancer Board Phase

[1374] 1. The server initializes multiple AI models, including surgeon AI, internist AI, radiologist AI, and pathologist AI, and fine-tunes each model with specialized knowledge.

[1375] 2. The server reads patient data from the database and inputs the data into each specialist AI.

[1376] 3. Each AI specialist will provide its own opinion (treatment plan) based on the patient data, discuss it with other AI specialists, and decide on the final recommendation.

[1377] Proposal Phase

[1378] 1. The server sends the generated treatment plan proposal to the terminal and notifies the terminal that "chemotherapy after surgery is optimal."

[1379] 2. The terminal displays this proposed treatment plan to the cancer board members and uses it as material for discussion.

[1380] Feedback Phase

[1381] 1. Users (Cancer Board members) will discuss the proposed treatment plan and decide on the final treatment plan.

[1382] 2. The terminal feeds back the final treatment plan to the server.

[1383] 3. The server stores this feedback data in a database and uses reinforcement learning (RLHF) techniques to optimize the AI ​​model.

[1384] Emotional Engine Phase

[1385] 1. The device uses an emotion engine to collect and analyze the user's emotions in real time. Specifically, it evaluates the user's emotions based on speech content, facial expressions, heart rate, and other biometric data.

[1386] 2. The device sends the analysis results to the server, which is used to generate the final treatment plan and provide feedback for the discussion.

[1387] Specific examples

[1388] For example, if Patient A, a 50-year-old male, is diagnosed with stage II lung cancer, the following process will occur:

[1389] 1. The terminal retrieves this data from the electronic medical record system and sends it to the server via the Internet.

[1390] 2. The server distributes this data to each specialist AI model, such as surgeon AI and internist AI.

[1391] 3. The AI ​​surgeon judges that "surgery is appropriate," and the AI ​​physician gives the opinion that "drug therapy is applicable." Based on these opinions, the final treatment plan is generated: "Chemotherapy after surgery is optimal."

[1392] 4. The device displays this proposed treatment plan to cancer board members and encourages discussion.

[1393] 5. During the discussion, the emotion engine monitors the emotions of medical staff in real time and may warn them that "excessive stress has been detected."

[1394] 6. If the final treatment plan is decided to be "immunotherapy in addition to chemotherapy after surgery," this information is again fed back to the server and used for reinforcement learning of the AI ​​model.

[1395] Examples of prompt statements

[1396] Enter the following prompt into the generative AI model:

[1397] Patient A is a 50-year-old man diagnosed with stage II lung cancer. The following data is provided: diagnosis, test results, medical history, etc. Each AI specialist must propose a treatment plan based on this data and decide on the final treatment plan.

[1398] In this way, the present invention can reduce the workload of doctors, reduce regional variations in the quality of medical care, take into consideration the emotions and stress of medical staff, and provide a more humane and safe medical environment.

[1399] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1400] Step 1:

[1401] The terminal retrieves patient data from the electronic medical record system.

[1402] Input: User-specified patient ID

[1403] Processing: The terminal accesses the electronic medical record system and retrieves data related to the specified patient ID, such as diagnosis results, test results, medical history, etc. For example, Patient A is a 50-year-old male who has been diagnosed with stage II lung cancer.

[1404] Output: Patient data (diagnosis results, test results, medical history, etc.)

[1405] Step 2:

[1406] The terminal converts the acquired patient data into an appropriate format (e.g., JSON format).

[1407] Input: Retrieved patient data

[1408] Processing: The terminal converts the patient data into JSON format, for example, { "name": "Patient A", "age": 50, "diagnosis": "Lung cancer stage II"}.

[1409] Output: Transformed patient data (JSON format)

[1410] Step 3:

[1411] The terminal transmits the converted data to the server.

[1412] Input: Transformed patient data (JSON format)

[1413] Processing: The device sends data to the server via the Internet.

[1414] Output: Data sent to the server

[1415] Step 4:

[1416] The server stores the received data in a database.

[1417] Input: Submitted patient data (JSON format)

[1418] Processing: The server connects to the database and saves the patient data. It leaves a log stating "Patient A's data has been saved."

[1419] Output: Patient data stored in a database

[1420] Step 5:

[1421] The server initializes an artificial intelligence model that is fine-tuned with expert knowledge.

[1422] Input: Fine-tuned AI model information

[1423] Processing: The server initializes multiple AI models, such as surgeon AI, internist AI, radiologist AI, and pathologist AI, and leaves a log stating, "Each AI model has been initialized."

[1424] Output: Initialized AI model

[1425] Step 6:

[1426] The server reads patient data from the database and inputs the data into each specialist AI.

[1427] Input: Patient data stored in the database

[1428] Processing: The server reads the patient data and inputs it into each specialist AI.

[1429] Output: Patient data input into each AI model

[1430] Step 7:

[1431] Each specialist AI will give their own opinion (treatment plan) based on patient data.

[1432] Input: Patient data entered into each specialist AI

[1433] Processing: The AI ​​surgeon determines that "surgery is appropriate," and the AI ​​internist gives the opinion that "drug therapy is applicable."

[1434] Output: Treatment plan issued by each AI

[1435] Step 8:

[1436] Each AI specialist will hold discussions with other AI specialists.

[1437] Input: Opinion of each specialist AI (treatment plan)

[1438] Processing: Each AI exchanges opinions within the server and decides on the final recommendation: "Chemotherapy after surgery is optimal."

[1439] Output: Optimal treatment plan

[1440] Step 9:

[1441] The proposed treatment plan generated by the server is sent to the terminal.

[1442] Input: Optimal treatment plan

[1443] Processing: The server sends the treatment plan to the terminal.

[1444] Output: Treatment plan sent to the terminal

[1445] Step 10:

[1446] The proposed treatment plan received by the device is displayed to the cancer board members (users).

[1447] Input: Treatment plan sent from the server

[1448] Processing: The device displays the treatment plan and notifies the user, "We suggest chemotherapy after surgery."

[1449] Output: Display of proposed treatment plan

[1450] Step 11:

[1451] Users will discuss the proposed treatment plan.

[1452] Input: proposed treatment plan

[1453] Process: Users hold a discussion based on the treatment plan. The discussion takes the form of "Do you have any additional opinions regarding this treatment plan?"

[1454] Output: Discussion transcript

[1455] Step 12:

[1456] The emotion engine analyzes user emotions in real time.

[1457] Input: User's speech, facial expressions, heart rate, and other biometric data

[1458] Processing: The emotion engine analyzes this data and assesses the user's stress level and emotional state.

[1459] Output: Parsed emotion data

[1460] Step 13:

[1461] The terminal feeds back the final treatment plan it has decided to use to the server.

[1462] Input: Cancer Board's final treatment plan

[1463] Processing: The terminal sends the final treatment plan to the server via the Internet.

[1464] Output: Final treatment plan sent to the server

[1465] Step 14:

[1466] The server stores the feedback data in a database and optimizes the AI ​​model through reinforcement learning.

[1467] Input: Feedback of final treatment plan data

[1468] Processing: The server stores this in a database and runs reinforcement learning (RLHF) to optimize the AI ​​model.

[1469] Output: Optimized AI model

[1470] (Application example 2)

[1471] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1472] The objective of this invention is to provide a technology that utilizes an emotion engine built into a conventional system for analyzing and discussing patient data to evaluate the emotional state of medical staff and passengers in real time, determine optimal treatment plans, and adjust autonomous driving, thereby improving the quality of medical care and the safety and comfort of passengers.

[1473] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1474] In this invention, the server includes means for collecting patient data, means for holding discussions based on the collected data using an AI model fine-tuned with multiple specialized knowledge, means for generating a treatment plan from the discussion results, means for proposing the generated treatment plan to medical staff, means for saving the proposal results as feedback data and for reinforcement learning the AI ​​model, means for acquiring the collected data using an in-vehicle terminal and evaluating the emotional state of passengers using an emotion engine, and means for adjusting the autonomous driving engine based on the evaluated emotional state. This makes it possible to determine a treatment plan that takes emotional states into consideration and achieve safe and comfortable autonomous driving while reducing the workload of medical staff.

[1475] "Patient data" refers to data that includes medical information such as a patient's diagnosis results, test results, medical history, and so on.

[1476] "Means of collection" refers to the means of obtaining and storing patient data from electronic medical record systems, etc.

[1477] A "fine-tuned AI model" is an AI model that has been specialised to have specialized knowledge, with parameters tuned for a specific purpose or role.

[1478] A "means for conducting discussions" is a means for multiple artificial intelligence models to exchange opinions with each other and execute a process to arrive at the optimal decision.

[1479] "Means for generating a treatment plan" refers to a means for determining a treatment plan and method based on the results of the discussion.

[1480] The "means for making suggestions to medical staff" is a means for presenting the generated treatment plan to medical staff.

[1481] "Means for reinforcement learning" refers to means for performing a learning process to improve the performance of an artificial intelligence model based on feedback data.

[1482] The "Emotion Engine" is an artificial intelligence model that recognizes and evaluates the emotional state of passengers and users in real time.

[1483] An "autonomous driving engine" is an artificial intelligence model and its execution system for autonomously driving and controlling a vehicle.

[1484] An "in-vehicle terminal" is a hardware device installed in an autonomous vehicle for acquiring and processing data.

[1485] "Collected data" refers to biometric data such as heart rate and facial expression data acquired by the in-vehicle terminal, as well as other associated data.

[1486] The "means for evaluating" is the means by which the emotion engine analyzes the emotional state of the passenger based on the collected data.

[1487] The "adjusting means" is a means for changing the autonomous vehicle's driving strategy based on the assessed emotional state.

[1488] MODE FOR CARRYING OUT THE INVENTION

[1489] System Configuration

[1490] The system of this invention combines passenger emotion recognition in an autonomous vehicle with autonomous driving adjustment. This system consists of a terminal that collects and manages patient data, a server that discusses the collected data with multiple AI models with specialized knowledge, an in-vehicle terminal that proposes a final driving policy and receives feedback, and an emotion engine that recognizes the user's emotions.

[1491] Data Collection Phase

[1492] The server collects biometric data such as facial expressions and heart rate of passengers through cameras and heart rate sensors on the in-vehicle terminals. This data is converted into an appropriate format (e.g., JSON format) and sent to the server.

[1493] Emotion Recognition Phase

[1494] The server stores the received biometric data in a database. It then initializes the emotion engine and inputs the collected data. The emotion engine evaluates the passenger's emotional state in real time and passes the results back to the server.

[1495] Autonomous driving phase

[1496] Based on the evaluation results of the emotion engine, the server will instruct the autonomous driving engine to adjust its driving policy. For example, if it determines that a passenger is feeling stressed, the autonomous driving engine can slow down the vehicle and play relaxing music in the car, ensuring a safe and comfortable riding experience for passengers.

[1497] Feedback Phase

[1498] The final driving plan is fed back from the on-board terminal to the server and stored in a database. The server uses the feedback data to reinforce learning of the artificial intelligence model and improve the system's performance.

[1499] Hardware and Software Used

[1500] Hardware: In-vehicle cameras, heart rate sensors, in-vehicle terminals

[1501] Software: EmotionRecognizer (emotion engine), OpenCV (image processing library), Requests (HTTP request library)

[1502] Specific examples

[1503] For example, when Passenger A boards an autonomous vehicle, the in-vehicle terminal uses a camera and heart rate sensor to capture facial expression and heart rate data. This data is sent to the server, where the emotion engine on the server analyzes Passenger A's emotional state. If Passenger A is determined to be stressed, the server will instruct the autonomous driving engine to slow down the vehicle and play relaxing music. This process allows Passenger A to enjoy a safer and more comfortable riding experience.

[1504] Prompt Sentence Examples

[1505] "Build a system that can identify passenger emotions in real time based on in-vehicle camera footage and heart rate data, and adjust the autonomous driving speed and route accordingly. For example, if a passenger feels stressed, the system can slow down the vehicle for safety and play relaxing music."

[1506] As described above, the system of the present invention enables operational management of autonomous vehicles that takes into account the emotional state of passengers, thereby improving comfort and safety.

[1507] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1508] Step 1:

[1509] The device uses a camera and heart rate sensor to capture passengers' facial expressions and heart rate data in real time.

[1510] Input: Passenger facial video, heart rate data

[1511] Data processing and calculation: The video data from the camera is divided into frames and facial features are extracted. The heart rate sensor periodically measures the heart rate and outputs it as digital data.

[1512] Output: Collected facial expression data and heart rate data

[1513] Step 2:

[1514] The facial expression data and heart rate data acquired by the device are converted into JSON format and sent to the server.

[1515] Input: facial expression data, heart rate data

[1516] Data processing and calculation: Serialization process is performed to convert data into JSON format.

[1517] Output: JSON formatted data packet

[1518] Step 3:

[1519] The server saves the received JSON format data in the database.

[1520] Input: JSON data packet

[1521] Data processing and calculation: Performs write operations on the database.

[1522] Output: Biometric data stored in a database

[1523] Step 4:

[1524] The server initializes the emotion engine and inputs biometric data obtained from the database to analyze the emotional state.

[1525] Input: Biometric data stored in a database

[1526] Data processing and calculation: The emotion engine (EmotionRecognizer) evaluates the emotional state based on the input data and generates emotion tags.

[1527] Output: Evaluated emotional state (emotion tag)

[1528] Step 5:

[1529] The server receives the evaluation results of the emotion engine and issues instructions to the autonomous driving engine to adjust the driving policy.

[1530] Input: Evaluated emotional state (emotion tag)

[1531] Data processing and calculation: Executes an algorithm that generates operational adjustment instructions corresponding to the emotional state.

[1532] Output: Operation adjustment instruction

[1533] Step 6:

[1534] The self-driving engine receives instructions from the server and adjusts speed and changes the in-car environmental settings.

[1535] Input: Operation adjustment instructions

[1536] Specific actions: Slowing down the vehicle, playing relaxing music, etc.

[1537] Output: Coordinated operating conditions

[1538] Step 7:

[1539] The final operation plan and its results are fed back to the server by the in-vehicle terminal.

[1540] Input: Adjusted operating conditions

[1541] Data processing and calculation: The operation result data is converted back into JSON format and sent to the server.

[1542] Output: Feedback data sent to the server

[1543] Step 8:

[1544] The server stores the feedback data in a database and performs reinforcement learning.

[1545] Input: Feedback data

[1546] Data processing and calculation: Improve the performance of AI models by running reinforcement learning algorithms based on feedback data.

[1547] Output: AI model optimized by reinforcement learning

[1548] Through the above processing steps, the system can analyze passengers' emotional states and optimize the operation of autonomous vehicles in real time.

[1549] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1550] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1551] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1552] [Fourth embodiment]

[1553] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1554] 7, a data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[1555] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1556] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1557] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1558] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1559] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1560] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1561] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1562] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1563] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1564] In the robot 414, the processor 46 performs the reception output process. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1565] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1566] This invention specifically describes a cancer board system that utilizes AI. The following explains the system's program processing in natural language, and includes specific examples.

[1567] System Configuration

[1568] The system consists of a terminal that collects and manages patient data, a server that processes the collected data and holds discussions with an artificial intelligence model with specialized knowledge, and a terminal that proposes a final treatment plan and receives feedback.

[1569] Data Collection Phase

[1570] 1. The terminal obtains patient data from the electronic medical record system. For example, it obtains the medical details (diagnosis results, test results, medical history, etc.) of Patient A, a 50-year-old male.

[1571] 2. The terminal converts the acquired patient data into an appropriate format (e.g., JSON format) and sends it to the server.

[1572] 3. The server stores the received data in the database.

[1573] AI Cancer Board Phase

[1574] 4. The server initializes multiple artificial intelligence models (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) that are fine-tuned with specialized knowledge.

[1575] 5. The server reads patient data from the database and inputs the data into each specialist AI.

[1576] 6. Each AI specialist will provide its own opinion (treatment plan) based on patient data and discuss it with other AI specialists.

[1577] 7. The server aggregates the discussion results and generates an optimal treatment plan.

[1578] Proposal Phase

[1579] 8. The server sends the proposed treatment plan to the terminal.

[1580] 9. The proposed treatment plan received by the terminal is displayed to the cancer board members (users).

[1581] 10. Users (Cancer Board members) discuss the proposed treatment plan and decide on the final treatment plan.

[1582] Feedback Phase

[1583] 11. The terminal feeds back the final treatment plan decided upon to the server.

[1584] 12. The server stores the feedback data in a database and optimizes the artificial intelligence model using reinforcement learning (RLHF).

[1585] Specific examples

[1586] For example, consider the case where Patient A is a 50-year-old man diagnosed with stage II lung cancer. This data is first obtained from the electronic medical record system by the device and sent to the server. The server receives this data and distributes it to various specialist AI models, such as the surgeon AI and internist AI. The surgeon AI determines whether surgery is appropriate, and the internist AI evaluates whether drug therapy is applicable. Based on the opinions of each specialist AI, the server generates a proposed treatment plan that states, "Chemotherapy after surgery is optimal." This proposed treatment plan is sent to the device and reviewed by members of the cancer board. If the final treatment plan is decided to be "immunotherapy in addition to chemotherapy after surgery," this information is fed back to the server and used for reinforcement learning in the AI ​​model.

[1587] In this way, the present invention can reduce the workload of doctors and reduce regional variations in the quality of medical care.

[1588] The processing flow will be explained below.

[1589] Step 1:

[1590] The terminal connects to the hospital's electronic medical record system and obtains the patient's medical data. Specifically, the patient ID is entered and an API is called to obtain the medical data (e.g., 50-year-old male, diagnosis: stage II lung cancer, medical history: high blood pressure).

[1591] Step 2:

[1592] The device formats the acquired data into an appropriate format and sends it to the server. Specifically, it converts the patient data into JSON format and posts it to the server using an HTTP request.

[1593] Step 3:

[1594] The server saves the received data in the database. Specifically, it executes an SQL query to insert the received patient data into the database.

[1595] Step 4:

[1596] The server initializes each specialist AI (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) in the AI ​​model, which has been fine-tuned with specialized knowledge. Specifically, it loads and configures each specialist AI model.

[1597] Step 5:

[1598] The server reads patient data from the database and inputs the data into each specialist AI. Specifically, it reads the patient data using an SQL query, converts the data into a format, and passes it to each specialist AI.

[1599] Step 6:

[1600] Each AI specialist will analyze patient data from their own perspective and offer their own opinion (treatment plan). Specific actions include the surgeon AI determining whether surgery is appropriate, and the internist AI evaluating whether drug therapy is applicable.

[1601] Step 7:

[1602] The server aggregates the results of the specialist AI discussions and generates the optimal treatment plan. Specifically, it collects the opinions of each specialist AI and applies voting and consensus-forming algorithms to determine the optimal treatment plan.

[1603] Step 8:

[1604] The server sends the generated treatment plan proposal to the terminal. Specifically, it converts the treatment plan proposal into JSON format and posts it to the terminal using an HTTP request.

[1605] Step 9:

[1606] The terminal displays the proposed treatment plan received to the cancer board members. Specifically, the received proposed treatment plan is displayed on the screen so that the cancer board members can refer to it.

[1607] Step 10:

[1608] Users (cancer board members) discuss the proposals and decide on the final treatment plan. Specifically, the cancer board members review the proposed treatment plan and reach a final consensus.

[1609] Step 11:

[1610] The terminal feeds back the final treatment plan it has decided to use to the server. Specifically, it converts the final treatment plan into JSON format and posts it to the server using an HTTP request.

[1611] Step 12:

[1612] The server stores the final treatment plan received in a database and optimizes the AI ​​model using reinforcement learning (RLHF). Specifically, it inserts the feedback data into the database, supplies new data to the AI ​​model, and executes the reinforcement learning algorithm.

[1613] Example 1

[1614] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1615] In the conventional medical system, the process of deciding on a treatment plan takes time, and the inconsistency of expertise and a lack of understanding of the information make it difficult to quickly decide on the optimal treatment plan.In addition, the management and use of collected patient data was not centralized, and collaboration between medical specialties was often ineffective.

[1616] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1617] In this invention, the server includes means for acquiring patient data from an electronic system, means for converting the acquired data into an appropriate format and sending and storing it in a high-performance database, means for holding discussions using generative AI models fine-tuned with multiple pieces of specialized knowledge based on the stored data and for each expert knowledge model to simulate opinions, means for aggregating the discussion results to generate an optimal treatment plan, means for sending the generated treatment plan to the user and displaying it in a dedicated application, and means for the user to decide on a final treatment plan, sending that feedback to the system and storing it in the database, and for reinforcement learning of the generative AI model.This enables centralized management of medical data and rapid decision-making on a treatment plan.

[1618] "Patient data" refers to information about a specific patient, such as diagnostic results, test results, treatment history, and medical history, collected by a medical institution.

[1619] "Electronic system" means a system for managing and processing data in digital form, including computer systems and electronic medical record systems installed in medical institutions.

[1620] A "database" is a high-performance information system for efficiently storing, managing, and searching collected data.

[1621] A "generative AI model" is an artificial intelligence model that has been trained using deep learning techniques to have specific expertise, such as an AI surgeon or AI internist.

[1622] "Discussion" is the process in which multiple generative AI models exchange opinions based on patient data to arrive at the optimal treatment plan.

[1623] A "treatment plan" is a specific medical procedure or treatment plan proposed based on the patient's diagnostic and test results.

[1624] "Users" are medical staff and specialists who use the system.

[1625] "Reinforcement learning" is a machine learning technique that allows artificial intelligence models to improve their performance based on feedback data.

[1626] "Feedback" is the process of returning the results of proposed treatment plans and information after their implementation to the system and using it for subsequent discussions and model updates.

[1627] The present invention provides an AI-based cancer monitoring system that comprehensively collects, processes, analyzes, and provides feedback on medical data, with the aim of quickly determining optimal treatment plans. Specific embodiments of this system are described below.

[1628] System Configuration

[1629] The system consists of a terminal that collects and manages patient data, a server that processes the collected data and holds discussions with a generative AI model with specialized knowledge, and a terminal that proposes a final treatment plan and receives feedback.

[1630] Data collection and format conversion

[1631] 1. The terminal uses a PC or tablet used in a medical institution to obtain patient data from an electronic system (electronic medical record system). For example, the medical details (diagnosis results, test results, medical history, etc.) of Patient A, a 50-year-old male, are obtained.

[1632] 2. The terminal converts the acquired patient data into an appropriate format (e.g., JSON format) and sends it to the server. This conversion is performed using a programming language such as Python. For example, the patient data is converted into the following JSON format:

[1633] json

[1634] {

[1635] "patient_id": "12345",

[1636] "age": 50,

[1637] "gender": "male",

[1638] "diagnosis": "lung cancer stage II",

[1639] "medical_history": ["hypertension", "diabetes"],

[1640] "latest_results": {

[1641] "CT_scan": "positive",

[1642] "blood_test": "normal"

[1643] }

[1644] }

[1645] 3. The server stores the received data in a high-performance database (such as MySQL or PostgreSQL) using SQL queries.

[1646] AI Cancer Board Discussion

[1647] 4. The server initializes multiple generative AI models (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) fine-tuned with TensorFlow or PyTorch. Specifically, a command like model = tf.keras.models.load_model('path_to_model') is assumed.

[1648] 5. The server reads the patient data from the database and inputs it into each generative AI model. This process is also implemented using scripts such as Python.

[1649] 6. Each generative AI model provides an opinion (treatment plan) based on the patient data and discusses it with other generative AI models. For example, an internist AI might suggest that "chemotherapy is optimal," and an surgeon AI might respond that "surgery should also be used."

[1650] 7. The server aggregates the results of the discussions and generates the optimal treatment plan. For example, the treatment plan generated might be "chemotherapy after surgery is optimal."

[1651] Treatment plan suggestions and feedback

[1652] 8. The server sends the generated treatment plan to the device and displays it in a dedicated application. This process uses an HTTP POST request and a secure communication protocol (HTTPS).

[1653] 9. The device displays the received treatment plan proposal to the user (cancer board member). This is done using a web application or mobile application.

[1654] 10. Users (cancer board members) discuss the proposed treatment plan and decide on the final treatment plan. This process uses dedicated discussion tools and remote video conferencing systems (Zoom, Teams, etc.).

[1655] 11. The device sends the final treatment plan back to the server. The feedback data is also sent in JSON format.

[1656] 12. The server stores the feedback data in a database again and optimizes the generative AI model using reinforcement learning with human feedback (RLHF). This advances the AI ​​model's learning, improving the accuracy of the next discussion and treatment plan proposal.

[1657] Specific examples

[1658] For example, if Patient A is a 50-year-old man diagnosed with stage II lung cancer, the data is retrieved from the electronic system by the terminal, converted to JSON format, and sent to the server. The server distributes the data to each generative AI model, with the surgeon AI determining whether surgery is appropriate and the internist AI evaluating whether drug therapy is applicable. As a result of the discussion, the server generates a treatment plan proposal stating that "chemotherapy after surgery is optimal" and sends it to the terminal. The cancer board members consider this and ultimately decide on a treatment plan of "immunotherapy in addition to chemotherapy after surgery," and this information is fed back to the server. The AI ​​model uses this feedback to perform reinforcement learning, improving the accuracy of future proposals.

[1659] Prompt Sentence Examples

[1660] "Patient A is a 50-year-old man who has been diagnosed with stage II lung cancer. His medical information includes his latest test results and medical history. Each AI specialist should propose a treatment plan based on this information."

[1661] In this way, the present invention can reduce the workload of doctors and reduce regional variations in the quality of medical care.

[1662] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1663] Step 1:

[1664] The terminal retrieves patient data from an electronic system (electronic medical record system). The input is the patient ID and treatment request, and the output is the raw patient data. Specifically, the terminal sends a GET / patients / {patient_id} request to the electronic medical record system via the API to retrieve the patient's medical information (for example, the diagnosis results, test results, medical history, etc. of 50-year-old male Patient A).

[1665] Step 2:

[1666] The terminal converts the acquired patient data into an appropriate format (for example, JSON format). The input is raw patient data, and the output is JSON formatted patient data. Specifically, a script running on the terminal parses the raw data and converts it into JSON format as follows:

[1667] json

[1668] {

[1669] "patient_id": "12345",

[1670] "age": 50,

[1671] "gender": "male",

[1672] "diagnosis": "lung cancer stage II",

[1673] "medical_history": ["hypertension", "diabetes"],

[1674] "latest_results": {

[1675] "CT_scan": "positive",

[1676] "blood_test": "normal"

[1677] }

[1678] }

[1679] Step 3:

[1680] The terminal sends the converted data to the server. The input is JSON-formatted patient data, and the output is a data transfer completion notification to the server. Specifically, the terminal uses an HTTP POST request to send data to POST / api / patient_data.

[1681] Step 4:

[1682] The server stores the received data in a database (for example, MySQL or PostgreSQL). The input is patient data in JSON format, and the output is a notification that the data has been saved. Specifically, a backend program running on the server inserts the received data into the database. For example, it executes the SQL query INSERT INTO patient_data (data) VALUES ('{...}').

[1683] Step 5:

[1684] The server initializes multiple generative AI models. The input is the file path of each AI model, and the output is an instance of the initialized AI model. Specifically, the server loads the AI ​​model using the TensorFlow or PyTorch API as follows: model = tf.keras.models.load_model('path_to_model').

[1685] Step 6:

[1686] The server reads patient data from the database and inputs it into each generative AI model. The input is the patient ID, and the output is the result of applying the data to the AI ​​model. Specifically, the server executes the SQL query SELECT FROM patient_data WHERE patient_id = '12345' and inputs the obtained data into each generative AI model.

[1687] Step 7:

[1688] Each generative AI model provides its own opinion (treatment plan) based on patient data and discusses it with other generative AI models. The input is the patient data and opinions from other AI models, and the output is the results of the discussion. Specifically, the system simulates the process in which each generative AI model performs processing, integrates the results, and exchanges opinions.

[1689] Step 8:

[1690] The server aggregates the discussion results and generates the optimal treatment plan. The input is the discussion results, and the output is the optimal treatment plan. Specifically, the server analyzes the output of the AI ​​model and generates the treatment plan that "chemotherapy after surgery is optimal."

[1691] Step 9:

[1692] The server sends the proposed treatment plan to the terminal. The input is the optimal treatment plan, and the output is a notification of data transfer completion to the terminal. Specifically, the server sends the treatment plan data to the terminal using an HTTP POST request.

[1693] Step 10:

[1694] The proposed treatment plan received by the device is displayed to the user. The input is the proposed treatment plan, and the output is the display of the treatment plan to the user. Specifically, the web application or mobile application on the device displays the received data in a GUI.

[1695] Step 11:

[1696] Users (cancer board members) discuss the proposed treatment plans and decide on the final treatment plan. The input is the proposed treatment plan, and the output is the final treatment plan. In concrete terms, users use discussion tools or remote video conferencing systems to hold discussions, and then enter and save their final decision in a form within the system.

[1697] Step 12:

[1698] The terminal feeds back the final treatment plan it has decided to the server. The input is the final treatment plan, and the output is a notification to the server that feedback data has been sent. Specifically, the terminal sends the feedback data to the server using an HTTP POST request.

[1699] Step 13:

[1700] The server stores the feedback data in a database again and optimizes the generative AI model using reinforcement learning. The input is the feedback data, and the output is the updated result of the AI ​​model based on reinforcement learning. Specifically, the server retrains the AI ​​model based on the feedback data and executes code such as model.fit(feedback_data).

[1701] (Application example 1)

[1702] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1703] In conventional cancer board systems, the process of generating collective opinions and confirming treatment plans takes a lot of time and effort, making it difficult for medical professionals to respond quickly.In addition, it is difficult for specialists in remote locations to confirm treatment plans in real time, making it difficult to provide efficient medical care.

[1704] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1705] In this invention, the server includes a means for collecting patient data, a means for conducting discussions based on the collected data using an artificial intelligence model fine-tuned with multiple specialized knowledge, and a means for generating a treatment plan from the discussion results. This allows the proposed treatment plan to be displayed on a mobile communication terminal where it can be checked by a specialist in real time, enabling quick and efficient confirmation of the treatment plan for each patient and appropriate treatment.

[1706] "Patient data" refers to information about individual patients collected by a medical facility, including diagnostic results, test results, medical history, etc.

[1707] "Means of collection" refers to the mechanism for obtaining patient data from an electronic medical record system, etc., converting it into an appropriate format, and entering it into the system.

[1708] An "artificial intelligence model" refers to a machine learning model that has specific expertise and is fine-tuned based on that expertise; in the medical field, it simulates the judgment of a medical specialist.

[1709] "Means for conducting discussions" refers to the process in which multiple artificial intelligence models work together to analyze patient data, automatically exchange opinions from the perspective of each specialist, and discuss the optimal treatment plan.

[1710] "Means for generating a treatment plan" refers to the process of aggregating the results of discussions by the artificial intelligence model and creating a proposal for the most effective treatment method for the patient.

[1711] "Means for proposing to medical staff" refers to a mechanism for displaying the generated treatment plan to medical staff in an appropriate format and encouraging their confirmation and discussion.

[1712] "Feedback data" refers to the information that is input into the system regarding the treatment plan finalized by medical staff and stored for later use in reinforcement learning of the artificial intelligence model.

[1713] "Mobile communication terminals" include portable devices such as smartphones and tablets that allow specialists to check treatment plans in real time wherever they are.

[1714] The present invention relates to a system for efficiently processing patient data and proposing optimal treatment plans to specialists. The system includes a means for collecting patient data, a means for discussing the collected data with an artificial intelligence model fine-tuned by multiple specialized knowledge, and a means for generating optimal treatment plans.

[1715] System Configuration

[1716] This system consists of the following main elements:

[1717] Terminal: This terminal retrieves patient data from the electronic medical record system, converts it into an appropriate format, and sends it to the server.

[1718] Server: Discussions are held using AI models fine-tuned with multiple areas of expertise (e.g., AI surgeons, AI internists, AI radiologists, AI pathologists, etc.) to generate optimal treatment plans.

[1719] Mobile communication terminal: A device (such as a smartphone or tablet) that allows specialists to check the generated treatment plan in real time.

[1720] Data Collection Phase

[1721] 1. The terminal retrieves patient data from the electronic medical record system. For example, information on the diagnosis, test results, and medical history of a 50-year-old male patient A.

[1722] 2. The terminal converts the acquired patient data into an appropriate format (e.g., JSON format) and sends it to the server.

[1723] 3. The server stores the received data in the database.

[1724] AI Cancer Board Phase

[1725] 4. The server initializes an artificial intelligence model that is fine-tuned with multiple areas of expertise.

[1726] 5. The server reads patient data from the database and inputs the data into each specialist AI.

[1727] 6. Each AI specialist will propose a treatment plan based on patient data and discuss it with other AI specialists.

[1728] 7. The server aggregates the discussion results and generates an optimal treatment plan.

[1729] Proposal Phase

[1730] 8. The server sends the proposed treatment plan to the mobile communication terminal.

[1731] 9. The proposed treatment plan received by the mobile communication terminal is displayed to the specialist in real time, who then confirms and considers it.

[1732] Feedback Phase

[1733] 10. The final treatment plan will be decided based on the treatment plan reviewed by the specialist.

[1734] 11. The terminal feeds back the final treatment plan to the server.

[1735] 12. The server stores the feedback data in a database and optimizes the artificial intelligence model using reinforcement learning (RLHF).

[1736] Specific examples

[1737] For example, consider the case where Patient A is a 50-year-old man diagnosed with stage II lung cancer. This data is first obtained from the electronic medical record system by the device and sent to the server. The server receives this data and distributes it to various specialist AI models, such as the surgeon AI and internist AI. The surgeon AI determines whether surgery is appropriate, and the internist AI evaluates whether drug therapy is applicable. Based on the opinions of each specialist AI, the server generates a proposed treatment plan stating that "chemotherapy after surgery is optimal." This proposed treatment plan is sent to the mobile communication device and reviewed in real time by specialists.

[1738] Prompt Sentence Examples

[1739] "Please suggest a treatment plan for Patient A."

[1740] This system enables medical staff to quickly and accurately confirm treatment plans and take appropriate action, improving efficiency in medical settings and providing optimal care to patients.

[1741] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1742] Step 1:

[1743] The terminal retrieves patient data from the electronic medical record system.

[1744] Input: Patient basic information, diagnosis results, test results, medical history, etc.

[1745] Output: Retrieved patient data

[1746] Specific operation: Access the electronic medical record system, search for the specified patient's data, extract the necessary information, and convert this data into an appropriate format (e.g., JSON format).

[1747] Step 2:

[1748] The terminal transmits the acquired patient data to the server.

[1749] Input: Patient data in JSON format

[1750] Output: Patient data sent to the server

[1751] Specific operation: The terminal uses a network connection to send the converted patient data to the server, using a data communication protocol such as an HTTP POST request.

[1752] Step 3:

[1753] The server stores the received data in a database.

[1754] Input: Patient data sent to the server

[1755] Output: Patient data stored in a database

[1756] Specific operation: The server parses the received JSON format data and stores the contents in a database appropriately. It uses a database management system (e.g., MySQL or PostgreSQL).

[1757] Step 4:

[1758] The server initializes an artificial intelligence model that is fine-tuned with multiple areas of expertise.

[1759] Input: Patient data stored in the database

[1760] Output: Initialized artificial intelligence model

[1761] Specific operation: The server loads models such as surgeon AI, internist AI, radiologist AI, and pathologist AI, and prepares patient data to be input into these models.

[1762] Step 5:

[1763] The server reads patient data from the database and inputs the data into each specialist AI.

[1764] Input: Patient data read from database

[1765] Output: Patient data input to each specialist AI

[1766] Specific operation: The server extracts specific patient data from the database and inputs this data into each specialist AI. The input data is used as a prompt for each specialist AI.

[1767] Step 6:

[1768] Each AI specialist will propose a treatment plan based on patient data and hold discussions with other AI specialists.

[1769] Input: Patient data entered into the specialist AI

[1770] Output: Treatment options after discussion

[1771] Specific operation: Each AI specialist generates a treatment plan based on the prompts and exchanges opinions with other AI specialists. Through this process, discussions are held and candidate treatment plans are generated for consensus.

[1772] Step 7:

[1773] The server aggregates the discussion results and generates an optimal treatment plan.

[1774] Input: Treatment plan candidates

[1775] Output: Optimal treatment plan

[1776] Specific operation: The server selects the most suitable treatment plan from multiple treatment plan candidates using a statistical or algorithmic method and generates it as the optimal treatment plan.

[1777] Step 8:

[1778] The server transmits the generated proposed treatment plan to the mobile communication terminal.

[1779] Input: Optimal treatment plan

[1780] Output: Treatment plan sent to mobile communication terminal

[1781] Specific operation: The server sends the optimal treatment plan to the mobile communication device. HTTP POST and WebSocket are used as communication protocols.

[1782] Step 9:

[1783] The proposed treatment plan received by the mobile communication terminal is displayed in real time to the specialist, who then checks and considers it.

[1784] Input: Treatment plan sent to mobile communication terminal

[1785] Output: Confirmation and review results by specialists

[1786] Specific operation: The mobile communication terminal displays the received treatment plan on the screen in real time, and a specialist doctor reviews it and considers its appropriateness.

[1787] Step 10:

[1788] The final treatment plan will be decided based on the treatment options reviewed by the specialist.

[1789] Input: Confirmation and review results by specialists

[1790] Output: Final treatment plan

[1791] Specific operation: The specialists discuss the treatment plans displayed on the screen and select the most appropriate treatment plan.

[1792] Step 11:

[1793] The terminal feeds back the final treatment plan to the server.

[1794] Input: Final treatment plan

[1795] Output: Feedback sent to the server

[1796] Specific operation: The terminal inputs the final treatment plan and sends it to the server as feedback.

[1797] Step 12:

[1798] The server stores the feedback data in a database and optimizes the artificial intelligence model using reinforcement learning (RLHF).

[1799] Input: Feedback of final treatment plan

[1800] Output: An artificial intelligence model optimized by reinforcement learning

[1801] Specific operation: The server stores the feedback data in a database and uses this data to perform reinforcement learning on the artificial intelligence model, improving the accuracy and effectiveness of the model.

[1802] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1803] This invention specifically describes an AI-based cancer board system that combines an emotion engine that recognizes the user's emotions. The following explains the system's program processing in natural language, with specific examples.

[1804] System Configuration

[1805] The system consists of a terminal that collects and manages patient data, a server that processes the collected data and holds discussions with an artificial intelligence model with specialized knowledge, a terminal that proposes a final treatment plan and receives feedback, and an emotion engine that recognizes the user's emotions.

[1806] Data Collection Phase

[1807] 1. The terminal obtains patient data from the electronic medical record system. For example, it obtains the medical details (diagnosis results, test results, medical history, etc.) of Patient A, a 50-year-old male.

[1808] 2. The terminal converts the acquired patient data into an appropriate format (e.g., JSON format) and sends it to the server.

[1809] 3. The server stores the received data in the database.

[1810] AI Cancer Board Phase

[1811] 4. The server initializes multiple artificial intelligence models (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) that are fine-tuned with specialized knowledge.

[1812] 5. The server reads patient data from the database and inputs the data into each specialist AI.

[1813] 6. Each AI specialist will provide its own opinion (treatment plan) based on patient data and discuss it with other AI specialists.

[1814] 7. The server aggregates the discussion results and generates an optimal treatment plan.

[1815] Proposal Phase

[1816] 8. The server sends the proposed treatment plan to the terminal.

[1817] 9. The proposed treatment plan received by the terminal is displayed to the cancer board members (users).

[1818] 10. Users (Cancer Board members) discuss the proposed treatment plan, and the emotion engine analyzes users' emotions in real time, taking their emotional state into account when deciding on the final treatment plan.

[1819] Feedback Phase

[1820] 11. The terminal feeds back the final treatment plan decided upon to the server.

[1821] 12. The server stores the feedback data in a database and optimizes the AI ​​model using reinforcement learning (RLHF).

[1822] Emotional Engine Phase

[1823] 13. The device uses an emotion engine to collect and analyze the emotions of Cancer Board members (users) in real time. Specifically, the device evaluates the user's emotions based on speech content, facial expressions, heart rate, and other biometric data.

[1824] 14. The device sends the analysis results to the server, which are used to generate the final treatment plan and feedback for the discussion.

[1825] Specific examples

[1826] For example, consider the case where Patient A is a 50-year-old man diagnosed with stage II lung cancer. This data is first obtained from the electronic medical record system by the device and sent to the server. The server receives this data and distributes it to various specialist AI models, such as the surgeon AI and internist AI. The surgeon AI determines whether surgery is appropriate, and the internist AI evaluates whether drug therapy is applicable. Based on the opinions of each specialist AI, the server generates a proposed treatment plan that states, "Chemotherapy after surgery is optimal." This proposed treatment plan is sent to the device and reviewed by members of the cancer board. During the review, the emotion engine monitors the emotions of medical staff in real time and issues a warning if excessive stress is detected. If the final treatment plan is decided to be "immunotherapy in addition to chemotherapy after surgery," this information is fed back to the server and used for reinforcement learning in the AI ​​model.

[1827] In this way, the present invention not only reduces the workload of doctors and reduces regional variations in the quality of medical care, but also takes into account the emotions and stress of medical staff, providing a more humane and safe medical environment.

[1828] The processing flow will be explained below.

[1829] Step 1:

[1830] The terminal connects to the hospital's electronic medical record system and obtains the patient's medical data. Specifically, the patient ID is entered and an API is called to obtain the medical data (e.g., 50-year-old male, diagnosis: stage II lung cancer, medical history: high blood pressure).

[1831] Step 2:

[1832] The terminal formats the acquired data into an appropriate format (for example, JSON format) and sends it to the server. Specifically, the patient data is converted into JSON format and POSTed to the server using an HTTP request.

[1833] Step 3:

[1834] The server saves the received data in the database. Specifically, it executes an SQL query to insert the received patient data into the database.

[1835] Step 4:

[1836] The server initializes each specialist AI (surgeon AI, internist AI, radiologist AI, pathologist AI, etc.) in the AI ​​model, which has been fine-tuned with specialized knowledge. Specifically, it loads each specialist AI model into memory and performs the necessary settings (parameters and model weights).

[1837] Step 5:

[1838] The server reads patient data from the database and inputs it into each specialist AI. Specifically, it reads the patient data using an SQL query, converts it into a different format, and passes the input data to each specialist AI model.

[1839] Step 6:

[1840] Each AI specialist analyzes patient data from their own perspective and offers their own opinion (treatment plan). Specifically, the AI ​​surgeon determines whether surgery is appropriate, while the AI ​​internist evaluates whether drug therapy is applicable. Each AI model generates its own analysis results.

[1841] Step 7:

[1842] The server aggregates the results of the specialist AI discussions and generates the optimal treatment plan. Specifically, it integrates the analysis results and opinions of each specialist AI and determines the optimal treatment plan using a consensus-forming algorithm and weighting coefficients.

[1843] Step 8:

[1844] The server sends the proposed treatment plan to the terminal. Specifically, it converts the proposed treatment plan into JSON format and posts it to the terminal using an HTTP request.

[1845] Step 9:

[1846] The proposed treatment plan received by the device is displayed to the Cancer Board members. Specifically, the proposed treatment plan is displayed on the screen so that the user can confirm the proposal.

[1847] Step 10:

[1848] Users (Cancer Board members) discuss the proposals, and the emotion engine analyzes the user's emotions in real time. Specifically, sensors collect biometric data such as speech content, facial expressions, and heart rate, and an algorithm is run to evaluate the user's emotional state. A final treatment plan is then decided.

[1849] Step 11:

[1850] The terminal feeds back the final treatment plan it has decided to use to the server. Specifically, it converts the final treatment plan into JSON format and posts it to the server using an HTTP request.

[1851] Step 12:

[1852] The server stores the final treatment plan received in a database and optimizes the AI ​​model using reinforcement learning (RLHF). Specifically, it inserts the feedback data into the database, supplies new data to the AI ​​model, and executes the reinforcement learning algorithm.

[1853] Step 13:

[1854] The device uses an emotion engine to collect and analyze the emotions of Cancer Board members (users) in real time. Specifically, it evaluates the user's emotions based on speech content, facial expressions, heart rate, and other biometric data.

[1855] Step 14:

[1856] The device sends the analysis results to the server, which uses them to generate the final treatment plan and feedback for the discussion. Specifically, the emotion evaluation results generated by the emotion engine are converted into JSON format and posted to the server using an HTTP request.

[1857] Example 2

[1858] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1859] Conventional medical systems focus on collecting patient data and proposing treatment plans, but it is difficult to consider the emotions and stress levels of medical staff, which can affect the quality and safety of medical care. Furthermore, there are insufficient methods for utilizing collected feedback data and optimizing artificial intelligence models using reinforcement learning. A system that can solve these problems is needed.

[1860] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1861] In this invention, the server includes means for collecting patient data, means for holding discussions based on the collected data using an AI model fine-tuned with multiple specialized knowledge, means for generating a treatment plan from the discussion results, means for proposing the generated treatment plan to medical staff, means for saving the proposal results as feedback data and for reinforcement learning the AI ​​model, means for recognizing user emotions and analyzing emotional data, and means for utilizing the emotional data for feedback and adjustment of the treatment plan. This makes it possible to continuously propose optimal treatment plans while taking into account the emotions and stress of medical staff.

[1862] "Patient data" refers to all information about patients collected in the healthcare system, including diagnostic results, test results, medical history, and treatment history.

[1863] An "electronic medical record system" is a medical information system for managing, storing, and sharing patient data in digital format.

[1864] An "artificial intelligence model" is a machine learning algorithm that has specific expertise and uses that expertise to analyze data and make decisions.

[1865] "Discussion" is a process in which multiple artificial intelligence models exchange their opinions and judgments to determine the optimal treatment plan.

[1866] A "treatment policy" is a plan that indicates the most appropriate methods and procedures for medical examination and treatment according to the patient's condition.

[1867] "Medical staff" refers to professional workers who treat and care for patients in medical settings, such as doctors, nurses, and medical technicians.

[1868] "Feedback data" is data collected based on proposed treatment plans and implementation results, and is information used to improve artificial intelligence models.

[1869] "Reinforcement learning" is a machine learning technique that allows artificial intelligence models to learn through experience and improve their performance.

[1870] An "emotion engine" is a software or hardware mechanism that recognizes a user's emotions and analyzes them in real time.

[1871] "Emotion data" is data that indicates the emotional state of a user, collected from the content of the user's speech, facial expressions, heart rate, and the like.

[1872] This invention is an AI cancer board system that collects and analyzes patient data, and proposes optimal treatment plans through discussions using artificial intelligence models with multiple specialized knowledge. This system comprehensively supports everything from collecting patient data to proposing treatment plans, providing feedback, and analyzing the emotions of medical staff.

[1873] System Configuration

[1874] The system consists of the following components:

[1875] 1. Terminal for collecting and managing patient data

[1876] 2. A server that processes the collected data and holds discussions with AI models with specialized knowledge

[1877] 3. A device that proposes the final treatment plan and receives feedback

[1878] 4. Emotion engine that recognizes user emotions

[1879] Data Collection Phase

[1880] 1. The terminal retrieves patient data from the electronic medical record system. The terminal accesses the electronic medical record system and retrieves data such as diagnosis results, test results, and medical history of Patient A, a 50-year-old male.

[1881] 2. The device converts the acquired patient data into an appropriate format, such as JSON, and sends it to a server via the Internet.

[1882] 3. The server saves the received data in the database and logs the message "Patient A's data has been saved."

[1883] AI Cancer Board Phase

[1884] 1. The server initializes multiple AI models, including surgeon AI, internist AI, radiologist AI, and pathologist AI, and fine-tunes each model with specialized knowledge.

[1885] 2. The server reads patient data from the database and inputs the data into each specialist AI.

[1886] 3. Each AI specialist will provide its own opinion (treatment plan) based on the patient data, discuss it with other AI specialists, and decide on the final recommendation.

[1887] Proposal Phase

[1888] 1. The server sends the generated treatment plan proposal to the terminal and notifies the terminal that "chemotherapy after surgery is optimal."

[1889] 2. The terminal displays this proposed treatment plan to the cancer board members and uses it as material for discussion.

[1890] Feedback Phase

[1891] 1. Users (Cancer Board members) will discuss the proposed treatment plan and decide on the final treatment plan.

[1892] 2. The terminal feeds back the final treatment plan to the server.

[1893] 3. The server stores this feedback data in a database and uses reinforcement learning (RLHF) techniques to optimize the AI ​​model.

[1894] Emotional Engine Phase

[1895] 1. The device uses an emotion engine to collect and analyze the user's emotions in real time. Specifically, it evaluates the user's emotions based on speech content, facial expressions, heart rate, and other biometric data.

[1896] 2. The device sends the analysis results to the server, which is used to generate the final treatment plan and provide feedback for the discussion.

[1897] Specific examples

[1898] For example, if Patient A, a 50-year-old male, is diagnosed with stage II lung cancer, the following process will occur:

[1899] 1. The terminal retrieves this data from the electronic medical record system and sends it to the server via the Internet.

[1900] 2. The server distributes this data to each specialist AI model, such as surgeon AI and internist AI.

[1901] 3. The AI ​​surgeon judges that "surgery is appropriate," and the AI ​​physician gives the opinion that "drug therapy is applicable." Based on these opinions, the final treatment plan is generated: "Chemotherapy after surgery is optimal."

[1902] 4. The device displays this proposed treatment plan to cancer board members and encourages discussion.

[1903] 5. During the discussion, the emotion engine monitors the emotions of medical staff in real time and may warn them that "excessive stress has been detected."

[1904] 6. If the final treatment plan is decided to be "immunotherapy in addition to chemotherapy after surgery," this information is again fed back to the server and used for reinforcement learning of the AI ​​model.

[1905] Examples of prompt statements

[1906] Enter the following prompt into the generative AI model:

[1907] Patient A is a 50-year-old man diagnosed with stage II lung cancer. The following data is provided: diagnosis, test results, medical history, etc. Each AI specialist must propose a treatment plan based on this data and decide on the final treatment plan.

[1908] In this way, the present invention can reduce the workload of doctors, reduce regional variations in the quality of medical care, take into consideration the emotions and stress of medical staff, and provide a more humane and safe medical environment.

[1909] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1910] Step 1:

[1911] The terminal retrieves patient data from the electronic medical record system.

[1912] Input: User-specified patient ID

[1913] Processing: The terminal accesses the electronic medical record system and retrieves data related to the specified patient ID, such as diagnosis results, test results, medical history, etc. For example, Patient A is a 50-year-old male who has been diagnosed with stage II lung cancer.

[1914] Output: Patient data (diagnosis results, test results, medical history, etc.)

[1915] Step 2:

[1916] The terminal converts the acquired patient data into an appropriate format (e.g., JSON format).

[1917] Input: Retrieved patient data

[1918] Processing: The terminal converts the patient data into JSON format, for example, { "name": "Patient A", "age": 50, "diagnosis": "Lung cancer stage II"}.

[1919] Output: Transformed patient data (JSON format)

[1920] Step 3:

[1921] The terminal transmits the converted data to the server.

[1922] Input: Transformed patient data (JSON format)

[1923] Processing: The device sends data to the server via the Internet.

[1924] Output: Data sent to the server

[1925] Step 4:

[1926] The server stores the received data in a database.

[1927] Input: Submitted patient data (JSON format)

[1928] Processing: The server connects to the database and saves the patient data. It leaves a log stating "Patient A's data has been saved."

[1929] Output: Patient data stored in a database

[1930] Step 5:

[1931] The server initializes an artificial intelligence model that is fine-tuned with expert knowledge.

[1932] Input: Fine-tuned AI model information

[1933] Processing: The server initializes multiple AI models, such as surgeon AI, internist AI, radiologist AI, and pathologist AI, and leaves a log stating, "Each AI model has been initialized."

[1934] Output: Initialized AI model

[1935] Step 6:

[1936] The server reads patient data from the database and inputs the data into each specialist AI.

[1937] Input: Patient data stored in the database

[1938] Processing: The server reads the patient data and inputs it into each specialist AI.

[1939] Output: Patient data input into each AI model

[1940] Step 7:

[1941] Each specialist AI will give their own opinion (treatment plan) based on patient data.

[1942] Input: Patient data entered into each specialist AI

[1943] Processing: The AI ​​surgeon determines that "surgery is appropriate," and the AI ​​internist gives the opinion that "drug therapy is applicable."

[1944] Output: Treatment plan issued by each AI

[1945] Step 8:

[1946] Each AI specialist will hold discussions with other AI specialists.

[1947] Input: Opinion of each specialist AI (treatment plan)

[1948] Processing: Each AI exchanges opinions within the server and decides on the final recommendation: "Chemotherapy after surgery is optimal."

[1949] Output: Optimal treatment plan

[1950] Step 9:

[1951] The proposed treatment plan generated by the server is sent to the terminal.

[1952] Input: Optimal treatment plan

[1953] Processing: The server sends the treatment plan to the terminal.

[1954] Output: Treatment plan sent to the terminal

[1955] Step 10:

[1956] The proposed treatment plan received by the device is displayed to the cancer board members (users).

[1957] Input: Treatment plan sent from the server

[1958] Processing: The device displays the treatment plan and notifies the user, "We suggest chemotherapy after surgery."

[1959] Output: Display of proposed treatment plan

[1960] Step 11:

[1961] Users will discuss the proposed treatment plan.

[1962] Input: proposed treatment plan

[1963] Process: Users hold a discussion based on the treatment plan. The discussion takes the form of "Do you have any additional opinions regarding this treatment plan?"

[1964] Output: Discussion transcript

[1965] Step 12:

[1966] The emotion engine analyzes user emotions in real time.

[1967] Input: User's speech, facial expressions, heart rate, and other biometric data

[1968] Processing: The emotion engine analyzes this data and assesses the user's stress level and emotional state.

[1969] Output: Parsed emotion data

[1970] Step 13:

[1971] The terminal feeds back the final treatment plan it has decided to use to the server.

[1972] Input: Cancer Board's final treatment plan

[1973] Processing: The terminal sends the final treatment plan to the server via the Internet.

[1974] Output: Final treatment plan sent to the server

[1975] Step 14:

[1976] The server stores the feedback data in a database and optimizes the AI ​​model through reinforcement learning.

[1977] Input: Feedback of final treatment plan data

[1978] Processing: The server stores this in a database and runs reinforcement learning (RLHF) to optimize the AI ​​model.

[1979] Output: Optimized AI model

[1980] (Application example 2)

[1981] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1982] The objective of this invention is to provide a technology that utilizes an emotion engine built into a conventional system for analyzing and discussing patient data to evaluate the emotional state of medical staff and passengers in real time, determine optimal treatment plans, and adjust autonomous driving, thereby improving the quality of medical care and the safety and comfort of passengers.

[1983] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1984] In this invention, the server includes means for collecting patient data, means for holding discussions based on the collected data using an AI model fine-tuned with multiple specialized knowledge, means for generating a treatment plan from the discussion results, means for proposing the generated treatment plan to medical staff, means for saving the proposal results as feedback data and for reinforcement learning the AI ​​model, means for acquiring the collected data using an in-vehicle terminal and evaluating the emotional state of passengers using an emotion engine, and means for adjusting the autonomous driving engine based on the evaluated emotional state. This makes it possible to determine a treatment plan that takes emotional states into consideration and achieve safe and comfortable autonomous driving while reducing the workload of medical staff.

[1985] "Patient data" refers to data that includes medical information such as a patient's diagnosis results, test results, medical history, and so on.

[1986] "Means of collection" refers to the means of obtaining and storing patient data from electronic medical record systems, etc.

[1987] A "fine-tuned AI model" is an AI model that has been specialised to have specialized knowledge, with parameters tuned for a specific purpose or role.

[1988] A "means for conducting discussions" is a means for multiple artificial intelligence models to exchange opinions with each other and execute a process to arrive at the optimal decision.

[1989] "Means for generating a treatment plan" refers to a means for determining a treatment plan and method based on the results of the discussion.

[1990] The "means for making suggestions to medical staff" is a means for presenting the generated treatment plan to medical staff.

[1991] "Means for reinforcement learning" refers to means for performing a learning process to improve the performance of an artificial intelligence model based on feedback data.

[1992] The "Emotion Engine" is an artificial intelligence model that recognizes and evaluates the emotional state of passengers and users in real time.

[1993] An "autonomous driving engine" is an artificial intelligence model and its execution system for autonomously driving and controlling a vehicle.

[1994] An "in-vehicle terminal" is a hardware device installed in an autonomous vehicle for acquiring and processing data.

[1995] "Collected data" refers to biometric data such as heart rate and facial expression data acquired by the in-vehicle terminal, as well as other associated data.

[1996] The "means for evaluating" is the means by which the emotion engine analyzes the emotional state of the passenger based on the collected data.

[1997] The "adjusting means" is a means for changing the autonomous vehicle's driving strategy based on the assessed emotional state.

[1998] MODE FOR CARRYING OUT THE INVENTION

[1999] System Configuration

[2000] The system of this invention combines passenger emotion recognition in an autonomous vehicle with autonomous driving adjustment. This system consists of a terminal that collects and manages patient data, a server that discusses the collected data with multiple AI models with specialized knowledge, an in-vehicle terminal that proposes a final driving policy and receives feedback, and an emotion engine that recognizes the user's emotions.

[2001] Data Collection Phase

[2002] The server collects biometric data such as facial expressions and heart rate of passengers through cameras and heart rate sensors on the in-vehicle terminals. This data is converted into an appropriate format (e.g., JSON format) and sent to the server.

[2003] Emotion Recognition Phase

[2004] The server stores the received biometric data in a database. It then initializes the emotion engine and inputs the collected data. The emotion engine evaluates the passenger's emotional state in real time and passes the results back to the server.

[2005] Autonomous driving phase

[2006] Based on the evaluation results of the emotion engine, the server will instruct the autonomous driving engine to adjust its driving policy. For example, if it determines that a passenger is feeling stressed, the autonomous driving engine can slow down the vehicle and play relaxing music in the car, ensuring a safe and comfortable riding experience for passengers.

[2007] Feedback Phase

[2008] The final driving plan is fed back from the on-board terminal to the server and stored in a database. The server uses the feedback data to reinforce learning of the artificial intelligence model and improve the system's performance.

[2009] Hardware and Software Used

[2010] Hardware: In-vehicle cameras, heart rate sensors, in-vehicle terminals

[2011] Software: EmotionRecognizer (emotion engine), OpenCV (image processing library), Requests (HTTP request library)

[2012] Specific examples

[2013] For example, when Passenger A boards an autonomous vehicle, the in-vehicle terminal uses a camera and heart rate sensor to capture facial expression and heart rate data. This data is sent to the server, where the emotion engine on the server analyzes Passenger A's emotional state. If Passenger A is determined to be stressed, the server will instruct the autonomous driving engine to slow down the vehicle and play relaxing music. This process allows Passenger A to enjoy a safer and more comfortable riding experience.

[2014] Prompt Sentence Examples

[2015] "Build a system that can identify passenger emotions in real time based on in-vehicle camera footage and heart rate data, and adjust the autonomous driving speed and route accordingly. For example, if a passenger feels stressed, the system can slow down the vehicle for safety and play relaxing music."

[2016] As described above, the system of the present invention enables operational management of autonomous vehicles that takes into account the emotional state of passengers, thereby improving comfort and safety.

[2017] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[2018] Step 1:

[2019] The device uses a camera and heart rate sensor to capture passengers' facial expressions and heart rate data in real time.

[2020] Input: Passenger facial video, heart rate data

[2021] Data processing and calculation: The video data from the camera is divided into frames and facial features are extracted. The heart rate sensor periodically measures the heart rate and outputs it as digital data.

[2022] Output: Collected facial expression data and heart rate data

[2023] Step 2:

[2024] The facial expression data and heart rate data acquired by the device are converted into JSON format and sent to the server.

[2025] Input: facial expression data, heart rate data

[2026] Data processing and calculation: Serialization process is performed to convert data into JSON format.

[2027] Output: JSON formatted data packet

[2028] Step 3:

[2029] The server saves the received JSON format data in the database.

[2030] Input: JSON data packet

[2031] Data processing and calculation: Performs write operations on the database.

[2032] Output: Biometric data stored in a database

[2033] Step 4:

[2034] The server initializes the emotion engine and inputs biometric data obtained from the database to analyze the emotional state.

[2035] Input: Biometric data stored in a database

[2036] Data processing and calculation: The emotion engine (EmotionRecognizer) evaluates the emotional state based on the input data and generates emotion tags.

[2037] Output: Evaluated emotional state (emotion tag)

[2038] Step 5:

[2039] The server receives the evaluation results of the emotion engine and issues instructions to the autonomous driving engine to adjust the driving policy.

[2040] Input: Evaluated emotional state (emotion tag)

[2041] Data processing and calculation: Executes an algorithm that generates operational adjustment instructions corresponding to the emotional state.

[2042] Output: Operation adjustment instruction

[2043] Step 6:

[2044] The self-driving engine receives instructions from the server and adjusts speed and changes the in-car environmental settings.

[2045] Input: Operation adjustment instructions

[2046] Specific actions: Slowing down the vehicle, playing relaxing music, etc.

[2047] Output: Coordinated operating conditions

[2048] Step 7:

[2049] The final operation plan and its results are fed back to the server by the in-vehicle terminal.

[2050] Input: Adjusted operating conditions

[2051] Data processing and calculation: The operation result data is converted back into JSON format and sent to the server.

[2052] Output: Feedback data sent to the server

[2053] Step 8:

[2054] The server stores the feedback data in a database and performs reinforcement learning.

[2055] Input: Feedback data

[2056] Data processing and calculation: Improve the performance of AI models by running reinforcement learning algorithms based on feedback data.

[2057] Output: AI model optimized by reinforcement learning

[2058] Through the above processing steps, the system can analyze passengers' emotional states and optimize the operation of autonomous vehicles in real time.

[2059] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[2060] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[2061] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[2062] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[2063] FIG. 9 is a diagram illustrating an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and actions arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[2064] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[2065] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[2066] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[2067] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[2068] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion ma...

Claims

1. a means for collecting patient data; A means for conducting discussions using an artificial intelligence model fine-tuned with multiple areas of expertise based on collected data; A means for generating a treatment plan from the discussion results; A means for proposing the generated treatment plan to medical staff; A means for storing the proposal results as feedback data and using the feedback data to reinforce learning the artificial intelligence model; A system including:

2. 2. The system according to claim 1, wherein the collected patient data is obtained from an electronic medical record system, converted into an appropriate format, and stored in a database.

3. The system according to claim 1, wherein each artificial intelligence model, fine-tuned with specialized knowledge, simulates the role of a specialist and holds discussions for each role.

Images