AI diagnosis system based on multiple rounds of dialogues

Abstract

The invention relates to the technical field of artificial intelligence, and discloses a multi-round dialogue-based AI diagnosis system, which comprises the following modules: a data acquisition module for acquiring personal health information of a user; the data fusion module is used for fusing the health data of different sources of the user to form a unified health data file; the intelligent inquiry analysis module is used for analyzing the health description input by the user through a natural language processing technology and performing comprehensive analysis in combination with historical health data of the user; the diagnosis reasoning module is used for performing reasoning based on deep learning and a medical knowledge base, and automatically generating a diagnosis result in combination with symptoms, disease history, medication conditions and the like of the patient; by adopting an intelligent inquiry system technology based on deep learning and natural language processing, the effect of automatically generating a personalized health diagnosis and treatment scheme is achieved, so that the problem that an existing system cannot comprehensively consider historical medical history, disease evolution and medication conditions of a user is solved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an AI diagnosis system based on multi-round dialogue. Background Art

[0002] As the medical industry continues to improve in terms of digitalization and informatization, intelligent medical consultation systems have gradually become an effective tool for improving the quality of medical services and optimizing the diagnosis and treatment process. Although existing health consultation systems have made some progress in some basic applications, they still have significant deficiencies in comprehensive analysis and intelligence. Many existing medical consultation systems typically employ simple question-and-answer models or rely on static databases for diagnostic solutions. These systems typically rely on users to answer questions using standardized symptom descriptions, and the system then matches these answers based on a fixed rule base and provides a diagnosis. This approach may be effective in dealing with simple health issues, but it lacks the ability to handle complex disease symptoms and personalized health management. In addition, the data isolation problem faced by existing health consultation systems is particularly prominent. Patients' health data usually comes from multiple different systems, including the hospital's electronic health records, drug prescription records, patients' self-reported symptoms, family health records, etc. However, these data are often stored in isolation in various systems and cannot be effectively integrated. The existing consultation system has a low level of intelligence and cannot dynamically adjust or provide personalized suggestions based on patients' health data and symptom descriptions. Most systems can only respond based on fixed rules or templates and lack the ability to flexibly adjust to real-time data changes. Even if multiple rounds of consultations are conducted in a question-and-answer mode, it is difficult for the system to respond in real time based on the user's health background, symptom changes, and feedback.

[0003] Therefore, the present invention proposes an AI diagnosis system based on multi-round dialogue to address the shortcomings of the existing technology. Summary of the Invention

[0004] In response to the shortcomings of the existing technology, the present invention provides an AI diagnostic system based on multi-round dialogues, which solves the problems that the AI ​​diagnostic system based on multi-round dialogues lacks the ability to deal with complex disease symptoms and personalized health management, and the data isolation problem faced by the existing health consultation system is particularly prominent.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: an AI diagnosis system based on multi-round dialogue, including the following modules: Data collection module: used to collect users' personal health information; Data fusion module: used to fuse the user's health data from different sources to form a unified health data file; Intelligent consultation analysis module: used to analyze the health description entered by the user through natural language processing technology and conduct a comprehensive analysis based on the user's historical health data; Diagnostic reasoning module: used to perform reasoning based on deep learning and medical knowledge base, and automatically generate diagnostic results based on the patient's symptoms, medical history, and medication status; Health advice generation module: Based on the user's health diagnosis results, the system will provide personalized diet advice, exercise advice, and treatment plans; Feedback and tracking module: used to allow users to provide feedback on the system's diagnostic results, thereby optimizing the diagnostic process through the continuous learning capabilities of artificial intelligence.

[0006] Preferably, the data acquisition module includes: A user input unit is used to receive health information provided by the user through text input or voice input to upload medical reports; The OCR unit is used to extract text data from hospital consultation reports uploaded by users and identify key information in the reports; A symptom description processing unit, used to convert the symptom description input by the user into a structured data format for use by the subsequent analysis module; Medication record unit, used to record the user's medication history and format the medication information into standard data for processing; The health record generation unit is used to integrate various types of collected health data and form a unified user health record.

[0007] Preferably, the data fusion module includes: Data preprocessing unit, used to standardize, remove noise and fill missing data from different sources; A data integration unit, used to integrate a user's various health data into a unified health data file; A health record storage unit, used for storing the fused health data in a database; The data consistency checking unit is used to check the logical consistency between various types of health data.

[0008] Preferably, the intelligent medical inquiry analysis module includes: The natural language processing unit is used to perform word segmentation, entity recognition, and sentiment analysis on the symptom description entered by the user to extract the core symptom information; a symptom prioritization unit, for generating a priority list of symptoms based on the urgency and relevance of the symptoms described by the user; A historical health data matching unit, which is used to match the user's symptom description with their historical health data, infer the possible evolution of the symptoms, and generate a list of possible health problems based on the matching results; Medical knowledge base query unit, used to query the medical knowledge base based on symptom descriptions and user historical data, conduct further speculation and analysis, and determine potential diagnostic problems; The multi-round medical consultation generation unit is used to automatically generate further medical questions based on symptom descriptions, historical health data, and query results from the medical knowledge base to supplement missing key information.

[0009] Preferably, the diagnostic reasoning module includes: The deep learning reasoning unit, based on a deep neural network model, performs multi-level reasoning analysis on the user's input symptom description, historical health data, and medication records to generate preliminary diagnosis results; Medical knowledge base reasoning unit, used to correct and supplement deep learning reasoning results by combining the medical professional knowledge base; The reasoning feedback unit dynamically adjusts the weight coefficients in the reasoning process based on real-time feedback from users, optimizes the diagnosis results, and modifies the reasoning rules in a timely manner according to actual conditions; The reasoning result verification unit is used to perform a posteriori verification on the automatically generated diagnosis results to ensure their rationality and eliminate possible misdiagnosis.

[0010] Preferably, the health advice generation module includes: Personalized health assessment unit, which assesses the user's overall health status based on the user's health records, diagnosis results, age, gender, and lifestyle information; A dietary recommendation generation unit generates personalized dietary recommendations based on the user's health status and diagnosis results; An exercise recommendation generation unit generates personalized exercise recommendations based on diagnostic results and health assessments, covering the type, intensity, and frequency of exercise suitable for the patient; The drug treatment plan generation unit generates personalized drug treatment plans based on the user's diagnosis results, medication history, and current health status, and recommends appropriate drugs to the patient; The treatment plan optimization unit learns from patient feedback and continuously optimizes diet, exercise, and medication recommendations to improve the accuracy and effectiveness of personalized treatment.

[0011] Preferably, the feedback and tracking module includes: A user feedback collection unit, used to receive user feedback on diagnosis results and health recommendations; A continuous learning unit that continuously optimizes the diagnostic reasoning and health advice generation process through machine learning algorithms based on user feedback; The health status tracking unit is used to regularly track the user's health status and provide up-to-date health advice by periodically updating the health profile; The feedback result analysis unit is used to analyze user feedback and changes in health status, automatically adjust the system's diagnostic rules and recommended strategies, and form a closed-loop health management service.

[0012] Preferably, the medical consultation generating unit automatically adjusts subsequent medical consultation strategies and questions based on the feedback information and preliminary diagnosis results provided by the user, ensuring that subsequent medical consultations can specifically supplement the missing information, thereby optimizing the comprehensiveness and accuracy of the diagnostic data.

[0013] Preferably, the feedback and tracking module regularly updates the user's health data file according to changes in the user's health status and provides the user with real-time health management services, forming a personalized long-term health tracking and optimization process to ensure the continuity and accuracy of health management.

[0014] The present invention also provides an AI diagnosis method based on multi-round dialogue, comprising the following steps: Collect users' personal health information; Integrate health data from different sources to form a unified health data archive; Analyze the health descriptions entered by users through natural language processing technology and conduct a comprehensive analysis based on historical health data; Perform reasoning analysis based on deep learning and medical knowledge base to automatically generate diagnostic results; Generate personalized health recommendations based on diagnostic results and provide users with relevant treatment plans; Optimize diagnosis and health recommendations based on user feedback and continuously track users' health status.

[0015] The present invention provides an AI diagnostic system based on multi-round dialogues. It has the following beneficial effects: 1. This invention utilizes an intelligent medical consultation system based on deep learning and natural language processing to automatically generate personalized health diagnoses and treatment plans. Compared to existing diagnostic systems based on simple questions and answers or static databases, this invention intelligently analyzes and provides more accurate personalized diagnoses based on a user's disease symptoms, historical health data, and hospital reports. This addresses the shortcomings of existing systems, which fail to comprehensively consider a user's medical history, disease progression, and medication use.

[0016] 2. This invention effectively integrates health data from various sources through a data fusion module, achieving comprehensive analysis and comprehensive diagnosis. Unlike existing diagnostic systems that rely on a single data source, this invention integrates multiple data sources and enables in-depth analysis, thus avoiding the limitations of traditional systems that often fail to provide comprehensive health assessments and treatment recommendations.

[0017] 3. This invention utilizes an intelligent consultation analysis module to automatically generate further consultation questions based on the user's input health data and historical health records, supplementing missing information in real time and achieving comprehensive optimization of consultation data. Compared to the static consultation process in existing systems, this invention addresses the shortcomings of traditional systems in effectively exploring user health details and quickly adjusting consultation strategies during the diagnosis process through multiple rounds of dynamic consultation and intelligent feedback mechanisms.

[0018] 4. The feedback and tracking module of this invention continuously collects user feedback and, combined with the self-learning capabilities of artificial intelligence, continuously optimizes the diagnostic process and health recommendations. Unlike existing systems that rely on manual review and fixed rules, this invention can adjust inference rules and treatment plans in real time based on changes in the user's health status, addressing the shortcomings of traditional systems such as the lack of flexibility and inability to dynamically optimize based on user feedback. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is a system architecture diagram of the present invention; Figure 2 This is a system architecture diagram of the present invention; Figure 3 This is a diagram of the data fusion module of the present invention; Figure 4 This is a diagram of the intelligent medical inquiry analysis module of the present invention; Figure 5 This is a diagram of the diagnostic reasoning module of the present invention; Figure 6 Generate a module diagram for the health advice of the present invention; Figure 7 This is a diagram of the feedback and tracking module of the present invention; Figure 8 This is a diagram of a data acquisition module of the present invention; Figure 9 It is a flow chart of the steps of the present invention. DETAILED DESCRIPTION

[0020] The following will clearly and completely describe the technical solution of the present invention in conjunction with the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0021] Please see the attached Figure 1 -Attached Figure 8 , an embodiment of the present invention provides an AI diagnosis system based on multi-round dialogue, including: Data collection module: used to collect users' personal health information; The data collection module is mainly used to collect users' health information from multiple sources and convert this information into structured data. Advanced technical means are used to standardize the data to ensure the quality and format consistency of the input data.

[0022] In general, the data collection module is designed to facilitate and improve user interaction. It supports text input, voice input, and data collection through uploading hospital reports. The following is a detailed description of each unit in the data collection module.

[0023] In this embodiment, the data acquisition module includes the following units: a user input unit is used to receive health information input by the user through various means. Specifically, the user can provide personal health data through text input, voice input, or uploading a medical report. For example, the user can enter his or her own symptom description through the keyboard on the mobile terminal or PC, or enter the symptom information through voice recognition technology. In order to support multiple languages ​​and different dialects, the system converts the voice content into text data through a voice recognition algorithm to facilitate subsequent analysis. For text input, the system has designed a clear input box to guide the user to provide symptom information step by step.

[0024] In some embodiments, the system may also provide guiding questions to help users more accurately enter symptom information and further improve data accuracy. For example, the system may ask users more targeted questions based on the initial symptoms they enter to help the system accurately identify the disease type.

[0025] In one possible implementation, the Optical Character Recognition (OCR) unit extracts text data from user-uploaded hospital medical reports or other health-related documents. OCR technology converts unstructured image data into editable and processable text. This unit accurately identifies key information in reports, such as patient names, disease diagnoses, medication records, and test results. The OCR unit removes noise, identifies edges, and recognizes characters in the image data to ensure accurate data extraction.

[0026] Specifically, when performing document recognition, the OCR unit first performs image enhancement processing on the input image, using binarization technology to remove unnecessary noise and enhance the contrast of key information. For recognized text content, the OCR unit uses a language model and medical vocabulary library to correct the recognition results, further improving recognition accuracy. Ultimately, OCR technology extracts key information from the document and converts it into structured data for use by subsequent modules.

[0027] The symptom description processing unit is primarily responsible for preprocessing and analyzing symptom descriptions entered by users. Specifically, the system utilizes natural language processing (NLP) technology to perform operations such as word segmentation, named entity recognition (NER), and sentiment analysis on user-entered text. By structured processing of user free text input, the system can accurately identify key information such as symptoms, duration, and severity.

[0028] For example, when processing a symptom description like "I've had a headache recently, which lasted for two days and is getting worse," the system will break it down into "headache" (symptom), "lasted for two days" (duration), and "felt getting worse" (symptom severity). This information is converted into structured data for subsequent analysis modules to process.

[0029] The Medication Record Unit records and standardizes user-provided medication information. Users can enter historical medication records or upload prescriptions, providing information such as medication name, dosage, and frequency of use. The system then converts this information into a standardized format for subsequent analysis.

[0030] In one embodiment, the medication record unit can also automatically recommend possible medications based on the user's symptoms and inquire about the user's medication history. In this way, the system can predict the patient's likely medication needs based on existing health data and provide the user with accurate medication recommendations.

[0031] The health record generation unit is the core of the data collection module. It integrates various types of collected health data into a unified user health record. This record includes information such as the user's symptom description, medication history, and hospital reports, and integrates them into a comprehensive health data file.

[0032] In some embodiments, the health profile generation unit uses data fusion technology to match and integrate data from different sources. By standardizing, deduplicating, and supplementing different data sources (such as symptom descriptions, historical medication records, and hospital reports), the system can generate a complete and unified health profile for each user for subsequent analysis modules. This profile not only includes basic symptom and historical health data, but also updates and improves in real time as the user's health changes.

[0033] In order to perform data standardization, the system uses the following mathematical model to normalize the data. Assume that the input original data is The standardized data is , then we have the following formula: ;in, Indicates the user-provided Health data, represents the mean of the data set, is the standard deviation of the data set.

[0034] Data fusion module: used to fuse the user's health data from different sources to form a unified health data file; The Data Fusion module processes, integrates, and standardizes collected data, ensuring that subsequent analysis and reasoning steps are performed on a consistent data basis. This module seamlessly integrates multiple types of data, including user symptoms, historical health records, medication records, and hospital reports, providing a solid foundation for personalized diagnosis, treatment plan generation, and health recommendations.

[0035] In this embodiment, the data fusion module primarily consists of the following units: A data preprocessing unit. Data from different sources may differ in format, unit, and semantics. Therefore, the data preprocessing unit standardizes, removes noise, and fills in missing data. Through this process, the system removes useless or erroneous data and supplements missing health information, ensuring the consistency and quality of the dataset.

[0036] Specifically, the data integration unit is responsible for merging data from multiple sources to form a unified health data file. This unit not only needs to consider the uniformity of the data format, but also needs to process data from different time points and different dimensions. Various health information, such as symptom descriptions entered by users, historical medication records, and diagnosis results in hospital reports, may be stored in different forms and structures. The data integration unit matches, sorts, and merges this data, converts it into a standard format, and summarizes this information in the user's health file.

[0037] For example, the symptoms a user enters might be real-time data, while the hospital's consultation report contains the diagnosis results at a specific point in time. The data integration unit must effectively match these two data types in time series, enabling the system to capture a complete historical health record. After integration, the data is stored in a structured format for subsequent processing by the system.

[0038] The health record storage unit is responsible for storing the consolidated health data in a database. Health data storage must ensure integrity, traceability, and efficient access. To meet these requirements, the storage unit utilizes efficient data storage technologies, such as relational or NoSQL databases, to facilitate fast data access and query.

[0039] The health record storage unit is responsible for not only storing current health data but also regularly updating the data archive based on changes in the user's health status. In this embodiment, the system ensures that each diagnosis and health advice is generated based on the latest data through automatic or user-initiated updates.

[0040] The data consistency checker ensures consistency between data from different sources. Data conflicts can occur, especially when integrating data from different systems. For example, a diagnosis in a hospital report might not match a user's input symptom description, or data in a medication history might not match the symptom description.

[0041] The data consistency check unit compares and analyzes the data to identify conflicts or inconsistencies and correct them according to the rules. Specifically, the system may make adjustments based on factors such as the credibility of the data source and the chronological order of the data to ensure that the final health data truly reflects the user's health status.

[0042] In order to standardize and integrate data from different sources, the system may use some mathematical models. Suppose there are two health data sets from different sources. and , these data sets may contain different information dimensions. The mathematical formula for data integration can be expressed as: ; in, Represents the final result vector or value of the synthesis, represents the first input vector or value, represents the second input vector or value, Is a weight coefficient, which indicates the influence weight of different data sources in the integration process, usually , and Represents health data from different sources.

[0043] In another implementation, when data types differ or are inconsistent, the system may use a weighted average approach to process the data. For example, if the time point of the symptom description is inconsistent with the time point of the hospital diagnosis report, the system may assign a higher weight to the diagnosis report to ensure the accuracy of the diagnosis result.

[0044] Alternatively, the data fusion module can dynamically adjust the data integration process based on real-time user feedback. For example, if a user provides new feedback on a symptom or historical health data, the system can update and adjust the health record in real time based on this new data, ensuring the real-time and accurate integration of data. The system's adaptive capabilities enable continuous optimization of the data fusion process based on feedback from different users.

[0045] Intelligent medical consultation analysis module: used to analyze the health description entered by the user through natural language processing technology, and conduct a comprehensive analysis in combination with the user's historical health data.

[0046] By combining natural language processing (NLP) technology, matching analysis of historical health data, and querying a medical knowledge base, combined with deep learning algorithms, the system comprehensively assesses symptoms, providing users with personalized diagnostic information and further consultation recommendations. The goal of the Intelligent Consultation Analysis module is to accurately and comprehensively analyze user-entered symptoms, assisting medical professionals in making scientific diagnoses and generating appropriate health recommendations based on the diagnostic results.

[0047] In this embodiment, the intelligent medical consultation analysis module includes the following units: A natural language processing unit provides symptom descriptions via text or voice input. These inputs are unstructured text data that requires natural language processing (NLP) technology to convert. The natural language processing unit's primary task is to convert the user's symptom description into structured data, including symptom analysis, entity recognition, and sentiment analysis. Using NLP technology, the system can extract key information such as symptom information, symptom duration, and severity from the user's natural language text input, facilitating subsequent analysis.

[0048] Specifically, for a symptom description like "I've had chest pain and a feeling of pressure for two days," the system will identify "chest pain" as the symptom, "two days" as the duration, and "pressure" as the symptom characteristics. All of this information will be converted into structured data for processing by subsequent modules.

[0049] In some embodiments, to further improve the accuracy of symptom descriptions, the system may also perform sentiment analysis in context. For example, by analyzing the user's emotional language, the system can determine the severity of the symptoms and provide support for subsequent consultation strategies.

[0050] The symptom prioritization unit ranks different symptoms based on the symptom descriptions entered by the user, prioritizing the most urgent or relevant symptoms. This ranking process relies on a combination of key information in the symptom description and historical health data.

[0051] Alternatively, the symptom prioritization unit may use a rule-based algorithm, combined with disease prioritization rules in the medical knowledge base, to determine the order in which symptoms should be addressed. For example, for a symptom like "chest pain," if the user's description lasts for a long time and is accompanied by other serious symptoms, the system will mark it as an urgent symptom and prioritize it. In symptom prioritization, the system takes into account multiple factors, such as the severity of the symptoms, duration, relevance of historical data, support from the medical knowledge base, etc. Assume that these factors are , priority score The calculation formula is as follows: ; in, represents the severity score of symptoms, Represents the duration of symptoms, Represents the matching degree of historical health data, Represents the support and weight coefficient in the medical knowledge base , , ,and Indicates the relative contribution of each factor to the symptom priority score. It represents the urgency of symptom treatment. Symptoms with higher scores will be given priority by the system.

[0052] The historical health data matching unit matches the user's symptom description with their historical health data, inferring the likely evolution of symptoms and generating a list of related health issues. This unit combines the user's previous health records, such as past medical history, family medical history, and medication history, to analyze symptom trends and assess whether symptoms are related to known diseases.

[0053] In one possible implementation, the historical health data matching unit compares the symptom data stored in the user's health record to find similar historical symptoms and disease patterns. In this way, the system can infer the potential cause of the symptoms and provide clues for subsequent diagnosis; Historical data matching is a key step in the intelligent medical consultation analysis module. The system uses the similarity between a user's historical health data and their current symptoms to infer potential health issues. One commonly used formula for calculating similarity is the cosine similarity formula.

[0054] Assume the current symptoms are , the historical symptom dataset is { , then the similarity The calculation formula is: ; in, A vector representation of the current symptoms, Indicates historical symptoms The vector representation of and They are vectors and The modulus length is calculated as follows: ; in, Represents a vector The model, is a vector Middle The value of the dimension, Represents a vector dimension.

[0055] The medical knowledge base query unit further infers the disease type by querying a pre-established medical knowledge base, combining symptoms with the user's historical health data. Medical knowledge bases typically contain extensive knowledge about diseases, the relationships between symptoms and diseases, and treatment options. By querying the knowledge base, the system can determine the likelihood of a given symptom and provide a reference for subsequent consultation and diagnosis.

[0056] Specifically, this query unit not only relies on fixed rules but also incorporates deep learning models for reasoning, updating the latest medical knowledge related to the user's health status in real time. For example, the latest research results on certain diseases may be dynamically added to the medical knowledge base, thereby improving the system's diagnostic accuracy.

[0057] The multi-round consultation generation unit automatically generates further questions based on the system's initial analysis of symptoms and comparisons with historical data. These questions are typically highly targeted, aiming to fill in missing pieces of the user's health information. Through multiple rounds of dialogue with the user, the system can gain a deeper understanding of the symptoms, thereby improving the accuracy of the diagnosis.

[0058] For example, when the system analyzes that the symptoms described by the user are related to certain known diseases, it may automatically ask more details, such as "Does your chest pain radiate to your left arm?" or "Do you feel short of breath or have difficulty breathing?" In this way, the system can adjust the content of the consultation based on real-time feedback to ensure the comprehensiveness and accuracy of the diagnostic data.

[0059] Diagnostic reasoning module: used to perform reasoning based on deep learning and medical knowledge base, and automatically generate diagnostic results based on the patient's symptoms, medical history, medication status, etc. The diagnostic reasoning module is responsible for performing deep reasoning and generating preliminary diagnostic results based on user-provided symptoms, historical health data, and information from the medical knowledge base. This module's functions include not only basic symptom reasoning but also complex dynamic adjustments and optimization processes to ensure accurate and personalized diagnostic results. By combining deep learning with the medical knowledge base, this module is able to comprehensively analyze complex health data to infer potential diseases.

[0060] In this embodiment, the diagnostic reasoning module includes multiple subunits, each responsible for tasks such as data reasoning, deep learning analysis, medical knowledge reasoning, dynamic adjustment, and reasoning result verification. The deep learning reasoning unit uses neural network algorithms to perform multi-level reasoning and analysis on the user's health data. Specifically, this unit uses deep learning models such as convolutional neural networks (CNN) and recurrent neural networks (RNN) to comprehensively analyze symptoms, historical health data, and medication records, extracting underlying patterns and regularities. The deep learning reasoning unit can adaptively learn from large amounts of health data, gradually improving the accuracy of reasoning.

[0061] Alternatively, the inference unit may utilize labeled case data during training, optimizing network weights through a backpropagation algorithm to gradually approach the optimal inference result. For example, when identifying and predicting the association between certain symptoms and diseases, the system can learn the relationships between different diseases through training and use this knowledge to reason about user-entered data to generate accurate diagnostic results. The medical knowledge base reasoning unit further refines and supplements disease assumptions by querying the rules and relationship graphs within the medical knowledge base and combining them with the results of deep learning reasoning. The medical knowledge base contains a wealth of information on disease symptom patterns, etiologies, treatment options, and their interrelationships. By combining deep learning results with the medical knowledge base, the system can generate diagnostic conclusions consistent with medical common sense based on different symptom combinations.

[0062] In one possible implementation, the system can also query the possibility of related diseases based on symptom descriptions and historical data, and infer the most appropriate diagnosis for the user's health status based on the correlation between symptoms and diseases. For example, when encountering symptoms such as heart disease or stomach problems, the system can query the association rules in the medical knowledge base to evaluate the possible correlation between the symptoms and these diseases, thereby obtaining a more accurate inference result. In the medical knowledge base reasoning unit, the system infers the possible disease type by calculating the correlation between symptoms and diseases. Assume that the symptom vector is , the disease vector is , then the support between symptoms and diseases It can be calculated by the following formula: ; in, Indicates the The weight of each symptom feature on the final disease inference result, Symptoms and disease The similarity between The support between symptoms and diseases, Indicates the total number of symptom characteristics.

[0063] The inference feedback unit dynamically adjusts the weighting coefficients used in the inference process based on real-time user feedback. This unit optimizes the inference model through this feedback mechanism, ensuring that the diagnosis results are more accurately tailored to the user's actual health status. For example, if a user provides new symptom descriptions or medical history information during the consultation, the inference feedback unit adjusts the weightings used in the inference process based on this new information and updates the diagnosis results, ensuring real-time and accurate inference.

[0064] The inference result verification unit is responsible for performing a posteriori verification on the automatically generated diagnostic results to ensure their rationality and eliminate possible misdiagnosis or missed diagnosis. This unit will check the reliability of the inference results based on the user's health records, medical records, and the system's internal verification rules. For example, if the inference result of a certain symptom does not conform to the common pattern in previous cases, the system will perform anomaly detection and adjust the inference path to ensure that the final output diagnosis result is consistent with the actual health status; In the inference feedback unit, the system dynamically adjusts the weight coefficients in the inference process according to user feedback information. Assume that the weights updated by the system according to the feedback are , then the adjusted inference results It can be calculated by the following formula: ; in, represents the adjusted inference results, is the new weight adjusted based on user feedback information. Indicates the input features, Indicates the total number of features.

[0065] Health advice generation module: Based on the user's health diagnosis results, the system will provide personalized diet advice, exercise advice and treatment plans.

[0066] The Health Advice Generation Module provides users with personalized health advice based on their health diagnosis results, historical health data, and multi-dimensional information such as individual lifestyle habits and environmental factors. These recommendations include diet, exercise, medication, lifestyle, and other aspects, aiming to help users improve their health, prevent disease, and enhance their physical fitness. By combining with the aforementioned intelligent consultation analysis module and diagnostic reasoning module, the Health Advice Generation Module can develop practical and personalized health management plans based on the user's health status and the system's reasoning results.

[0067] In this embodiment, the health advice generation module includes the following units: The personalized health assessment unit first assesses the user's overall health based on information such as the user's health profile, symptom analysis results, historical health data, and lifestyle habits. The results of the health assessment help the system identify the user's health risks and provide a basis for subsequent health advice generation.

[0068] Specifically, the personalized health assessment unit analyzes the user's basic health information (such as age, gender, weight, height, etc.) and symptom information to calculate a health score. This score reflects the user's overall health status and provides a reference for generating personalized health recommendations. By combining criteria from the medical knowledge base, the assessment unit can also classify the user's health risk as high risk, low risk, or normal.

[0069] The dietary advice generation unit provides users with personalized dietary recommendations based on their health status and health assessment results. For example, if the system identifies that a user may be at risk for high blood pressure or diabetes, the dietary advice generation unit will recommend a low-salt, low-sugar diet to avoid exacerbating high blood pressure or diabetes symptoms. The dietary advice generation unit will generate detailed recommendations based on medical guidelines and the user's actual needs, including food types, nutritional combinations, and daily intake.

[0070] In one possible implementation, the dietary recommendation generation unit can also adjust based on the user's dietary preferences and lifestyle habits. For example, if the user prefers certain foods or follows specific dietary cultural habits, the system will flexibly adjust the dietary recommendations to better suit the user's actual situation while ensuring health. The system can also regularly adjust the diet plan based on the user's health status.

[0071] The dietary advice generation unit generates specific dietary advice for the user by analyzing the user's health status and combining it with medical guidelines. The diet plan generated by the system includes food types, nutritional combinations, and daily intake. The specific calculation formula is: ; in, Indicates the The contribution of various foods to health, Indicates the The nutritional composition and intake of various foods, Score the user's dietary recommendations, Indicates the total number of food types.

[0072] The exercise recommendation generation unit develops a personalized exercise plan based on the user's health assessment results and the output of diagnostic reasoning. If the system identifies that the user is overweight or lacks exercise, the exercise recommendation generation unit will recommend an appropriate exercise type, intensity, and frequency. For users at risk of cardiovascular disease, the exercise recommendation generation unit may recommend low-intensity aerobic exercise such as walking, swimming, or yoga; while for users in better health, the system may recommend high-intensity training to help improve cardiopulmonary function and enhance physical strength.

[0073] Optionally, the exercise suggestion generator can also optimize based on the user's exercise preferences. For example, if the user dislikes running, the system can recommend alternative exercise methods such as cycling or rope skipping. Furthermore, exercise suggestions are tailored to the user's daily activity level. For example, the system can create a personalized exercise plan based on the user's daily activity level, health status, and goals (such as weight loss or fitness).

[0074] The medication treatment plan generation unit recommends appropriate medication treatment plans based on the user's health diagnosis results and historical medication records. When generating treatment plans, the system screens and customizes medications based on the user's specific circumstances, such as medical history, allergies, and medication dependencies. For example, for users with hypertension, the system recommends appropriate antihypertensive medications, taking into account dosage and frequency of use to ensure efficacy and minimize side effects.

[0075] In one possible implementation, the medication treatment plan generation unit can also make real-time adjustments based on the user's health feedback. If the user experiences side effects or poor efficacy during treatment, the system will adjust the treatment plan and recommend new medications.

[0076] Feedback and tracking module: used to allow users to provide feedback on the system's diagnostic results, thereby optimizing the diagnostic process through the continuous learning capabilities of artificial intelligence.

[0077] The feedback and tracking module collects user feedback to continuously optimize the system's diagnostic reasoning and health advice generation. Through continuous learning and data updates, this module ensures that diagnostic results and health advice are adjusted based on actual health changes, thereby improving the system's accuracy and personalization.

[0078] In this embodiment, the feedback and tracking module includes the following units, responsible for collecting user feedback, conducting continuous learning, tracking health status, and analyzing feedback results: The user feedback collection unit collects user satisfaction with diagnosis results, health recommendations, and the consultation process by asking users for feedback. This feedback includes user evaluations of diagnostic accuracy, health recommendation effectiveness, and consultation experience. Through real-time user feedback, the system can perceive the discrepancy between diagnosis results and actual health status, providing a basis for subsequent optimization.

[0079] Specifically, users can provide feedback to the system in a variety of ways, including through text, voice, or images. The system can convert this feedback into structured data to facilitate subsequent analysis and learning. To ensure efficient processing of feedback, the system may also design guiding questions to encourage users to provide more useful feedback.

[0080] The continuous learning unit uses machine learning algorithms to continuously optimize diagnostic reasoning and health recommendation generation based on user feedback. This feedback allows the system to identify and learn from deficiencies in diagnostic results, gradually adjusting its reasoning rules and recommendation strategies to ensure continuous improvement over time.

[0081] Alternatively, the continuous learning unit might employ reinforcement learning to train the system model through real-time feedback, enabling the system to dynamically adjust its behavior based on this feedback. For example, if a diagnosis fails to meet expectations, the system can update the model through reinforcement learning to improve future inference accuracy. Through these learning processes, the system can gradually adapt to the health conditions and needs of different users, providing more personalized services.

[0082] In the continuous learning unit, the system dynamically adjusts the weights in the reasoning process based on the feedback information. Assume that the weighted score of the feedback information is , the adjustment factor of the system is , then the updated weight of the system It can be calculated by the following formula: ; in, is the adjusted weight, is the original weight, is the adjustment factor, Represents a weighted score.

[0083] The health tracking unit regularly tracks changes in the user's health and provides periodic updates to the user's health data. Based on user feedback on changes in health status, the system automatically updates health records and adjusts health recommendations. For example, the system can update diagnostic results and treatment recommendations based on new symptom descriptions, changes in condition, or treatment feedback, ensuring that users always receive recommendations that match their current health status.

[0084] In some embodiments, the health tracking unit can also periodically check the user's health data to identify potential risks to the user's health status and issue early warnings. For example, if the system detects that the user's symptoms are worsening, the system may remind the user to seek medical attention or adjust treatment plans, thereby effectively managing the user's health.

[0085] In the health status tracking unit, the system predicts the future health status by analyzing the changing trend of the user's health data. Assume that the user's health data is , the historical data of each health indicator is , then the predicted value of health status It can be calculated by the following formula: ; in, represents the predicted health status, is the weight coefficient of each health indicator, Indicates the The value of the current health indicator, Indicates the Historical values ​​of health indicators, Indicates the total number of health indicators.

[0086] The feedback analysis unit is responsible for conducting in-depth analysis of user feedback, identifying patterns and potential issues within the feedback. By analyzing user feedback data, the system can identify common issues across different diagnostic results and health recommendations, thereby optimizing the system's overall diagnostic strategy.

[0087] Specifically, this unit aggregates and analyzes large amounts of user feedback data to identify key factors that influence diagnostic accuracy and the effectiveness of health recommendations. For example, certain symptom descriptions may lead to misdiagnosis, or certain health recommendations may be less effective in specific user groups. By analyzing this data, the system can promptly adjust its inference rules, consultation strategies, and health recommendations, thereby improving its overall performance.

[0088] In the feedback result analysis unit, the system performs weighted calculation on each user feedback to measure the contribution of each feedback to system optimization. Assume that the user feedback information is The weight of each feedback item is , then the system's comprehensive feedback score The calculation formula is: ; in, Indicates feedback item The weight of influence on the final feedback score, represents the weighted comprehensive score, Indicates the The numerical result of the feedback items, Indicates the total number of feedback items.

[0089] Alternatively, in some embodiments, the feedback and tracking module incorporates sentiment analysis technology to more accurately identify emotional tendencies in user feedback. For example, if a user expresses dissatisfaction with a diagnosis, the system can use sentiment analysis to identify this dissatisfaction and prioritize certain decisions in the reasoning process to ensure the diagnosis meets user expectations.

[0090] The AI ​​diagnosis method based on multi-round dialogue described below and the AI ​​diagnosis system based on multi-round dialogue described above can be referenced to each other.

[0091] Please see the attached Figure 9 The present invention also provides an AI diagnosis method based on multi-round dialogue, comprising the following steps: S1. Collect the user's personal health information; S2. Integrate health data from different sources to form a unified health data archive; S3. Analyze the health description entered by the user through natural language processing technology; S4. Perform reasoning analysis based on deep learning and medical knowledge base to automatically generate diagnostic results; S5. Generate personalized health recommendations based on the diagnosis results and provide relevant treatment plans to users; S6. Optimize diagnosis and health recommendations based on user feedback and continuously track user health status.

[0092] In step S1, the user's health data is collected through the data collection module. This data can be input in a variety of ways, including but not limited to manual symptom descriptions, voice input, and uploaded medical reports. Through this process, the system is able to collect comprehensive health information, including symptoms, medication history, past medical history, and hospital reports.

[0093] In step S2, the system integrates data from various sources into a unified health profile through the data fusion module. To ensure data accuracy and consistency, the data fusion module standardizes, removes noise, and fills in missing data. The integrated health profile contains health information from multiple sources and is then used by the subsequent analysis module.

[0094] In step S3, in the intelligent consultation analysis module, the system uses natural language processing (NLP) technology to perform word segmentation, entity recognition, and sentiment analysis on the symptom description entered by the user, extracting useful information such as symptom type, duration, and severity. Simultaneously, the system combines the user's historical health data with symptom analysis and generates a list of possible health issues based on this information.

[0095] In step S4, the diagnostic reasoning module automatically generates a diagnosis using a deep learning model and a medical knowledge base. Based on the trained model, the deep learning reasoning unit comprehensively analyzes the input symptom data, medication records, and historical health data to infer potential health issues. Simultaneously, the medical knowledge base reasoning unit queries relevant medical rules and relationships to further verify and correct the diagnosis.

[0096] In step S5, the system uses the health advice generation module to provide users with personalized diet, exercise, and treatment recommendations based on the diagnosis results. Depending on the user's health status, the system may recommend lifestyle adjustments, increased exercise, specific medications, or medical treatment options. Furthermore, the system can recommend appropriate hospital departments or specialists based on the patient's health status.

[0097] In step S6, the feedback and tracking module allows users to provide feedback on the system's diagnostic results and health recommendations. The system collects this feedback data and, in conjunction with a continuous learning algorithm, continuously optimizes the diagnosis and recommendation generation process. By dynamically adjusting the weights and rules used in the inference process, the system continuously improves its diagnostic accuracy. The system also periodically tracks changes in the user's health and promptly updates health records and recommendations to ensure the continuity and accuracy of health management services.

[0098] The method of this embodiment can be used to execute the above system embodiment. Its principles and technical effects are similar and will not be described in detail here.

[0099] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. The AI ​​diagnosis system based on multi-round dialogue is characterized by: Includes the following modules, Data collection module: used to collect users' personal health information; Data fusion module: used to fuse the user's health data from different sources to form a unified health data file; Intelligent consultation analysis module: used to analyze the health description entered by the user through natural language processing technology and conduct a comprehensive analysis based on the user's historical health data; Diagnostic reasoning module: used to perform reasoning based on deep learning and medical knowledge base, and automatically generate diagnostic results based on the patient's symptoms, medical history, and medication status; Health advice generation module: Based on the user's health diagnosis results, the system will provide personalized diet advice, exercise advice, and treatment plans; Feedback and tracking module: used to allow users to provide feedback on the system's diagnostic results, thereby optimizing the diagnostic process through the continuous learning capabilities of artificial intelligence.

2. The AI ​​diagnostic system based on multi-round dialogue according to claim 1, characterized in that: The data acquisition module includes: A user input unit is used to receive health information provided by the user through text input or voice input to upload medical reports; The OCR unit is used to extract text data from hospital consultation reports uploaded by users and identify key information in the reports; A symptom description processing unit, used to convert the symptom description input by the user into a structured data format for use by the subsequent analysis module; Medication record unit, used to record the user's medication history and format the medication information into standard data for processing; The health record generation unit is used to integrate various types of collected health data and form a unified user health record.

3. The AI ​​diagnostic system based on multi-round dialogue according to claim 1, characterized in that: The data fusion module includes: Data preprocessing unit, used to standardize, remove noise and fill missing data from different sources; A data integration unit, used to integrate a user's various health data into a unified health data file; A health record storage unit, used for storing the fused health data in a database; The data consistency checking unit is used to check the logical consistency between various types of health data.

4. The AI ​​diagnostic system based on multi-round dialogue according to claim 1, characterized in that: The intelligent medical inquiry analysis module includes: The natural language processing unit is used to perform word segmentation, entity recognition, and sentiment analysis on the symptom description entered by the user to extract the core symptom information; a symptom prioritization unit, for generating a priority list of symptoms based on the urgency and relevance of the symptoms described by the user; A historical health data matching unit, which is used to match the user's symptom description with their historical health data, infer the possible evolution of the symptoms, and generate a list of possible health problems based on the matching results; Medical knowledge base query unit, used to query the medical knowledge base based on symptom descriptions and user historical data, conduct further speculation and analysis, and determine potential diagnostic problems; The multi-round medical consultation generation unit is used to automatically generate further medical questions based on symptom descriptions, historical health data, and query results from the medical knowledge base to supplement missing key information.

5. The AI ​​diagnostic system based on multi-round dialogue according to claim 1, characterized in that: The diagnostic reasoning module includes: The deep learning reasoning unit, based on a deep neural network model, performs multi-level reasoning analysis on the user's input symptom description, historical health data, and medication records to generate preliminary diagnosis results; Medical knowledge base reasoning unit, used to correct and supplement deep learning reasoning results by combining the medical professional knowledge base; The reasoning feedback unit dynamically adjusts the weight coefficients in the reasoning process based on real-time feedback from users, optimizes the diagnosis results, and modifies the reasoning rules in a timely manner according to actual conditions; The reasoning result verification unit is used to perform a posteriori verification on the automatically generated diagnosis results to ensure their rationality and eliminate possible misdiagnosis.

6. The AI ​​diagnostic system based on multi-round dialogue according to claim 1, characterized in that: The health advice generation module includes: Personalized health assessment unit, which assesses the user's overall health status based on the user's health records, diagnosis results, age, gender, and lifestyle information; A dietary recommendation generation unit generates personalized dietary recommendations based on the user's health status and diagnosis results; An exercise recommendation generation unit generates personalized exercise recommendations based on diagnostic results and health assessments, covering the type, intensity, and frequency of exercise suitable for the patient; The drug treatment plan generation unit generates personalized drug treatment plans based on the user's diagnosis results, medication history, and current health status, and recommends appropriate drugs to the patient; The treatment plan optimization unit learns from patient feedback and continuously optimizes diet, exercise, and medication recommendations to improve the accuracy and effectiveness of personalized treatment.

7. The AI ​​diagnostic system based on multi-round dialogue according to claim 1, characterized in that: The feedback and tracking module includes: A user feedback collection unit, used to receive user feedback on diagnosis results and health recommendations; A continuous learning unit that continuously optimizes the diagnostic reasoning and health advice generation process through machine learning algorithms based on user feedback; The health status tracking unit is used to regularly track the user's health status and provide up-to-date health advice by periodically updating the health profile; The feedback result analysis unit is used to analyze user feedback and changes in health status, automatically adjust the system's diagnostic rules and recommended strategies, and form a closed-loop health management service.

8. The AI ​​diagnostic system based on multi-round dialogue according to claim 4, characterized in that: The questionnaire generation unit automatically adjusts subsequent questionnaire strategies and questions based on the feedback information and preliminary diagnosis results provided by the user, ensuring that subsequent questionnaires can specifically supplement the missing information, thereby optimizing the comprehensiveness and accuracy of the diagnostic data.

9. The AI ​​diagnostic system based on multi-round dialogue according to claim 7, characterized in that: The feedback and tracking module regularly updates the user's health data file according to changes in the user's health status and provides the user with real-time health management services, forming a personalized long-term health tracking and optimization process to ensure the continuity and accuracy of health management.

10. An AI diagnosis method based on multi-round dialogue, applied to the AI ​​diagnosis system based on multi-round dialogue according to any one of claims 1 to 9, characterized in that: The following steps are involved: Collect users' personal health information; Integrate health data from different sources to form a unified health data archive; Analyze the health descriptions entered by users through natural language processing technology and conduct a comprehensive analysis based on historical health data; Perform reasoning analysis based on deep learning and medical knowledge base to automatically generate diagnostic results; Generate personalized health recommendations based on diagnostic results and provide users with relevant treatment plans; Optimize diagnosis and health recommendations based on user feedback and continuously track users' health status.

Images