Online inquiry intelligent diagnosis system based on deep learning model

Abstract

The invention relates to the technical field of online medical treatment, in particular to an online inquiry intelligent diagnosis system based on a deep learning model, which comprises a patient interaction end, an intelligent diagnosis end and a hospital database end. The patient interaction terminal collects patient information and generates an inquiry request; the intelligent diagnosis terminal performs disease derivation and diagnosis and treatment scheme generation in combination with historical medical records and a medical knowledge base; and the hospital database end provides medical record storage and knowledge graph updating support. The system optimizes the diagnosis accuracy through a deep learning model processing platform, and has the functions of aging optimization and automatic registration. The problem of resource shortage of doctors can be relieved, the diagnosis efficiency and the medical record data utilization rate are improved, and efficient and accurate online inquiry service is provided for patients.

Description

Technical Field

[0001] This invention relates to the field of online medicine, specifically to an intelligent online consultation and diagnosis system based on a deep learning model. Background Technology

[0002] With the rapid growth of healthcare needs, online consultation technology has been widely used in internet hospitals and telemedicine platforms. However, the current consultation process still relies primarily on human doctors; even in consultations that do not require direct contact with the patient's body, doctors still need to make diagnoses in person. While online consultations offer convenience to patients, the quality varies greatly due to limited doctor resources, inaccurate patient descriptions, and a lack of systematic medical record analysis capabilities, making it difficult to meet the needs of large-scale, efficient consultations. Furthermore, existing online consultation systems have not significantly reduced the workload of doctors; hospitals still need to allocate a large number of medical staff for initial screening, further exacerbating the strain on medical resources.

[0003] Existing technologies have the following shortcomings: First, the problem of strained doctor resources is particularly prominent. Each doctor has limited working time and energy, making it difficult to handle a large number of consultation requests in a short period of time. Second, some online consultation systems rely on rule engines or simple keyword matching technology, which cannot deeply analyze the actual symptoms of patients, resulting in low accuracy of diagnostic results. Third, existing systems generally lack effective utilization of historical case data, making it difficult to combine medical record data for intelligent diagnosis, affecting the scientific validity and reliability of consultation results. Finally, doctors still need to manually respond to each patient's consultation request, which not only increases the workload of doctors but also reduces the overall consultation efficiency of the hospital.

[0004] To address the aforementioned problems, there is an urgent need for an online consultation system that can fully utilize artificial intelligence technology to improve consultation efficiency and accuracy through intelligent means, while reducing the workload of doctors. This invention aims to overcome the shortcomings of existing technologies and proposes a novel online consultation design based on AI technology to achieve more efficient and accurate medical services. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent online consultation and diagnosis system based on a deep learning model, to address the problems of limited doctor resources, low diagnostic efficiency, and insufficient medical record data support in existing online consultation systems. An intelligent online consultation and diagnosis system based on a deep learning model includes a patient interaction terminal, an intelligent diagnosis terminal, and a hospital database terminal. These three terminals operate collaboratively through a distributed computing network, within which a deep learning model processing platform is deployed. The patient interaction terminal is used to collect patient information and symptom descriptions, generate consultation requests, and upload them to the processing platform. The intelligent diagnosis terminal receives consultation requests and performs intelligent analysis based on historical medical record data to output preliminary diagnostic results. The hospital database terminal stores and manages patients' historical medical record data and a medical knowledge base, providing data support for intelligent diagnosis.

[0006] The patient interaction module includes a user registration module, a voice input module, a symptom analysis module, and a department recommendation module. The user registration module allows patients to register using their name, ID number, and mobile phone number, and completes identity verification via facial recognition technology. The voice input module allows patients to describe their symptoms via voice; the system uses natural language processing technology to analyze the voice content and extract key symptom information. The symptom analysis module uses a priority keyword analysis algorithm to semantically understand the patient's symptom description and generate structured symptom data. The department recommendation module analyzes the patient's symptoms based on a deep learning model and recommends the most likely departments for the patient to choose from. Attached Figure Description

[0007] Appendix Figure 1 System overall architecture diagram Detailed Implementation

[0008] This invention provides an online intelligent diagnostic system based on a deep learning model, combined with the attached... Figure 1 The system architecture and module structure shown are explained in detail, along with their specific implementation methods. In practical applications, this system achieves efficient and accurate online consultation services through the collaborative operation of the patient interaction terminal, intelligent diagnosis terminal, hospital database terminal, and deep learning model processing platform.

[0009] Firstly, from an overall architecture perspective, the patient interaction terminal, the intelligent diagnosis terminal, and the hospital database terminal are connected through a distributed computing network, and data processing and analysis are performed at the core of a deep learning model processing platform. (See attached image) Figure 1As shown, the patient interaction terminal is responsible for collecting patients' personal information and symptom descriptions, generating consultation requests, and uploading them to the deep learning model processing platform. The intelligent diagnosis terminal receives the consultation request, performs intelligent analysis based on historical medical record data, and outputs preliminary diagnostic results. The hospital database terminal stores and manages patients' historical medical record data and medical knowledge base, providing comprehensive data support for intelligent diagnosis. This distributed architecture not only improves the system's operating efficiency but also ensures data security and privacy.

[0010] In the specific implementation of the patient interaction module, the user registration module completes real-name registration using name, ID number, and mobile phone number, and verifies identity using facial recognition technology. This process ensures the authenticity and uniqueness of patient information, providing a reliable foundation for subsequent diagnosis and treatment. The voice input module allows patients to describe their symptoms by voice. The system uses natural language processing technology to analyze the voice content and extract key symptom information. For example, when a patient describes "I've been coughing for the past few days, accompanied by a low-grade fever," the system automatically identifies "cough" and "low-grade fever" as primary symptoms and converts them into structured data. The symptom parsing module uses a priority keyword analysis algorithm to perform semantic understanding of symptom descriptions and generate structured symptom data. For example, for a description like "chest tightness and shortness of breath," the system will determine "chest tightness" as the primary symptom and "shortness of breath" as a secondary symptom based on priority rules, and encode them according to predefined dimensions. The department recommendation module analyzes patient symptoms based on a deep learning model and recommends the most likely department. For example, if a patient complains of "abdominal pain and nausea," the system may recommend gastroenterology or emergency medicine for the patient to choose from. In addition, the age-friendly optimization module simplifies the operation process through voice-driven interaction, making it easier for elderly patients to use the system.

[0011] The intelligent diagnosis module is the core of the entire system, and its functions are jointly implemented by the medical record matching module, disease inference module, treatment plan generation module, and doctor review module. The medical record matching module retrieves historical medical record data similar to the patient's symptoms from the hospital database and uses a cosine similarity algorithm to calculate the similarity between the symptom vector and the medical record vector. The formula is as follows:

[0012]

[0013] Among them, V p，i V represents the value of the i-th dimension of the patient's symptom vector. h，iThis represents the value of the i-th dimension of the historical medical record vector, where n represents the total number of vector dimensions. Using this algorithm, the system can quickly find historical cases most similar to the current patient's symptoms, providing a basis for subsequent disease inference. The disease inference module infers possible disease types based on the matching results and generates preliminary medication suggestions by combining them with the knowledge graph. For example, if the matching results show that the patient's symptoms are highly similar to historical cases of "acute bronchitis," the system will infer "acute bronchitis" as a possible disease type and suggest the use of antibiotics or cough suppressants. The treatment plan generation module integrates the disease inference results with treatment plans in the medical knowledge base to form a complete treatment report. For example, for "acute bronchitis," the system will generate a comprehensive plan including drug treatment, dietary recommendations, and precautions. The doctor review module allows doctors to review and adjust the intelligent diagnosis results and feed the final diagnosis back to the processing platform. For example, after reviewing the system-generated "acute bronchitis" diagnosis, a doctor may add a suggestion for a lung CT scan based on the patient's specific condition, thereby improving the diagnostic report.

[0014] The hospital database functionality is implemented through a medical record storage module, a knowledge graph update module, and a network extension module. The medical record storage module stores patients' historical medical data and ensures data security through encryption algorithms. For example, patients' personal information and medical records are encrypted and accessible only to authorized personnel. The knowledge graph update module regularly updates the medical knowledge graph to ensure the accuracy of intelligent diagnosis. For example, when new medical research findings are published, the system automatically integrates relevant information into the knowledge graph, thereby improving the accuracy of disease diagnosis. When the medical record database lacks relevant cases, the network extension module can query the latest medical literature and drug information online and integrate them into the diagnostic report. For example, if the system cannot find historical cases matching the patient's symptoms, it will search online for the latest treatment guidelines or drug instructions to provide doctors with reference. The deep learning model processing platform also includes a dynamic optimization unit, used to continuously optimize the model parameters based on the doctor's review results, improving the accuracy of subsequent diagnoses. The optimization process uses the gradient descent algorithm, as shown in the following formula:

[0015]

[0016] Where θ represents the model parameters and η represents the learning rate. This represents the gradient of the loss function with respect to the parameters. Using this algorithm, the system can gradually adjust the model parameters based on feedback from doctors, making it more adaptable to real-world application scenarios. For example, if a doctor repeatedly corrects the system's misdiagnosis of a certain disease, the dynamic optimization unit will focus on adjusting relevant parameters, thereby reducing the occurrence of similar errors.

[0017] In practical applications, patients describe their symptoms and generate consultation requests through the patient interaction interface. The intelligent diagnosis module combines historical medical records and a medical knowledge base for intelligent analysis, outputting preliminary diagnostic results. For example, a patient accesses the system with "persistent headaches and blurred vision." After describing their symptoms via voice input, the system generates structured data and recommends a neurology specialist. The medical record matching module retrieves similar cases from the hospital database and finds that many of these cases are related to "migraines." Therefore, the disease inference module deduces "migraine" as a possible disease type and suggests the use of painkillers and rest therapy. The treatment plan generation module further integrates relevant information to form a complete report including medication and lifestyle adjustments. Doctors confirm or adjust the results through the doctor review module. The final diagnosis is recorded and stored in the hospital database for reference in subsequent consultations.

[0018] Furthermore, the automated registration form generation module automatically generates registration forms based on the doctor's reviewed diagnosis, reducing manual operations and improving registration efficiency. For example, when a doctor confirms that a patient needs further examination, the system automatically generates a registration form and sends it to the patient's mobile phone; the patient simply needs to follow the prompts to go to the hospital. The age-friendly optimization module simplifies the operation process through voice-driven interaction, making the system more convenient for elderly patients. For example, elderly patients can complete registration, symptom description, and department selection through voice commands, eliminating the need for complex keyboard input or screen operations.

[0019] In summary, this invention achieves efficient patient information collection and symptom analysis through a patient interaction terminal, accurate disease deduction and treatment plan generation through an intelligent diagnostic terminal, and comprehensive historical medical record support by connecting to the data systems of multiple hospitals through a hospital database terminal. In practical implementation, the modules work closely together to ensure the system's efficiency and accuracy. For example, a young patient accesses the system with symptoms of "fever and sore throat," describes their symptoms through the voice input module, and the system generates structured data and recommends an ENT specialist. The medical record matching module retrieves similar cases from the hospital database, finding that many are related to "acute pharyngitis." Therefore, the disease deduction module deduces "acute pharyngitis" as a possible disease type and suggests antibiotics and increased fluid intake. The treatment plan generation module further integrates relevant information to form a complete report including medication and dietary recommendations. Doctors confirm or adjust the results through the doctor review module, and the final diagnosis is recorded and stored in the hospital database for future reference. Simultaneously, the automated registration form generation module automatically generates registration forms based on the doctor's reviewed diagnosis, reducing manual operations and improving registration efficiency.

Claims

1. An online intelligent diagnostic system based on a deep learning model, characterized in that, The system includes a patient interaction terminal, an intelligent diagnosis terminal, and a hospital database terminal. These three terminals operate collaboratively through a distributed computing network, within which a deep learning model processing platform is deployed. The patient interaction terminal collects patient information and symptom descriptions to generate consultation requests and uploads them to the processing platform. The intelligent diagnosis terminal receives consultation requests and performs intelligent analysis based on historical medical record data to output preliminary diagnostic results. The hospital database terminal stores and manages patients' historical medical record data and a medical knowledge base to provide data support for intelligent diagnosis.

2. The online intelligent diagnostic system based on a deep learning model according to claim 1, characterized in that, The patient interaction module includes a user registration module, a voice input module, a symptom analysis module, and a department recommendation module. The user registration module allows patients to register using their real names, ID numbers, and mobile phone numbers, and complete identity verification through facial recognition technology. The voice input module allows patients to describe their symptoms via voice, and the system uses natural language processing technology to analyze the voice content and extract key symptom information. The symptom analysis module uses a priority keyword analysis algorithm to perform semantic understanding of the patient's symptom description and generate structured symptom data. The department recommendation module uses a deep learning model to analyze the patient's symptoms and recommend the most likely departments for the patient to choose from.

3. The online intelligent diagnostic system based on a deep learning model according to claim 2, characterized in that, The intelligent diagnostic module includes a medical record matching module, a disease inference module, a treatment plan generation module, and a doctor review module. The medical record matching module is used to obtain historical medical record data with similar symptoms to the patient from the hospital database and calculate the similarity between the symptom vector and the medical record vector using the cosine similarity algorithm. The disease inference module infers possible disease types based on the matching results and generates preliminary medication suggestions by combining them with a knowledge graph. The treatment plan generation module integrates the disease inference results with treatment plans in the medical knowledge base to form a complete diagnosis and treatment report. The doctor review module is used by doctors to review and adjust the intelligent diagnostic results and feed the final diagnosis results back to the processing platform.

4. The online intelligent diagnostic system based on a deep learning model according to claim 3, characterized in that, The hospital database includes a medical record storage module, a knowledge graph update module, and a network expansion module. The medical record storage module stores patients' historical medical record data and ensures data security through encryption algorithms. The knowledge graph update module is used to regularly update the medical knowledge graph to ensure the accuracy of intelligent diagnosis. When the medical record database lacks relevant cases, the network expansion module allows the system to connect to the network to query the latest medical literature and drug information and integrate them into the diagnostic report.

5. The online intelligent diagnostic system based on a deep learning model according to claim 4, characterized in that, The deep learning model processing platform also includes a dynamic optimization unit, which is used to continuously optimize the model parameters based on the doctor's review results to improve the accuracy of subsequent diagnoses. The optimization process adopts the gradient descent algorithm.

6. The online intelligent diagnostic system based on a deep learning model according to claim 5, characterized in that, The patient interaction terminal also includes an age-friendly optimization module, which simplifies the operation process and improves the ease of use for elderly patients through voice-driven interaction. The intelligent diagnosis terminal also includes an automated registration form generation module, which automatically generates registration forms based on the diagnosis results reviewed by the doctor, reducing manual operation and improving registration efficiency.

7. The online intelligent diagnostic system based on a deep learning model according to claim 6, characterized in that, The medical record matching module uses a cosine similarity algorithm to calculate the similarity between the patient's symptom vector and the historical medical record vector, as shown in the following formula: Where V p，i V represents the value of the i-th dimension of the patient's symptom vector. h，i This represents the value of the i-th dimension of the historical medical record vector, where n represents the total number of dimensions in the vector.

8. The online intelligent diagnostic system based on a deep learning model according to claim 7, characterized in that, The dynamic optimization unit uses the gradient descent algorithm to optimize the model parameters, as shown in the following formula: Where θ represents the model parameters and η represents the learning rate. This represents the gradient of the loss function with respect to the parameters.

Images