Parkinson's disease patient home information management system based on Internet hospital

Abstract

The invention relates to the technical field of intelligent medical treatment and remote medical treatment, and discloses a Parkinson's disease patient home information management system based on an internet hospital. Comprising the following steps: (1) a multi-source illness state acquisition module for monitoring illness states in real time by combining wearable equipment and a mobile terminal through a diary recording template and scale self-evaluation based on an internet hospital; (2) a medication management module which comprises medication reminding, illness change and side effect recording; (3) a data analysis and early warning module which uses incremental learning and model fine tuning technologies, optimizes a prediction model, analyzes the correlation between medication parameters and illness state changes, and realizes early warning of illness state deterioration; and (4) a remote collaborative diagnosis and treatment module which constructs a visual data platform, establishes a two-way communication channel and realizes immediate intervention. The system is based on an internet hospital, and an intelligent solution is provided for long-range management of Parkinson's disease by innovatively constructing a'monitoring-analysis-early warning-intervention 'closed-loop management system.

Description

Technical Field

[0001] This invention relates to the fields of intelligent healthcare and telemedicine, and discloses an information management system for monitoring, analyzing, providing early warnings, and communicating the condition of Parkinson's disease patients at home based on an internet hospital. Background Technology

[0002] Parkinson's disease is a common neurodegenerative disease, typically manifesting as motor disturbances, tremors, rigidity, and gait instability. As the disease progresses and with prolonged medication use, symptom fluctuations and motor complications such as dyskinesia can occur, severely impacting patients' quality of life. Parkinson's patients often face mobility issues, and their children's busy work schedules make it difficult to accompany them to appointments, resulting in very limited follow-up visits. Continuing medication at home becomes the primary treatment method, leading to rapid deterioration of the condition. Therefore, enabling patients to effectively monitor their condition and manage their medication at home, promptly identifying changes in their condition and adverse drug reactions, communicating with their doctors regularly, and adjusting treatment plans in a timely manner are crucial for better improving their condition and achieving long-term, effective management of the disease.

[0003] Existing treatment models generally suffer from the following shortcomings: 1. Patients have difficulty traveling and cannot make close and regular follow-ups. Staying at home and continuing to take medication as usual has become the main treatment method. 2. Most patients are middle-aged or elderly and lack their children to accompany them. They often forget or make mistakes in the dosage and timing of medication, which significantly affects the treatment effect. 3. Traditional methods can only obtain partial data through consultation and physical examination, and cannot capture subtle fluctuations in patients' daily symptoms and motor function at home. The daily medication, efficacy feedback and side effects records of patients are not detailed enough, which affects the treatment effect and safety. 4. Due to the lack of real-time data monitoring and intelligent analysis, existing technologies are unable to detect adverse reactions such as exercise complications in a timely manner, thus missing the best time for intervention. 5. Doctor-patient interaction is also limited by the face-to-face consultation model, making it difficult for doctors to adjust treatment plans in a timely manner or to remotely monitor and intervene in high-risk patients. 6. The lack of a systematic and standardized data management platform makes it impossible to integrate high-quality real-world data, thus limiting long-term research and in-depth analysis of treatment effects.

[0004] Therefore, those skilled in the art provide a home information management system for Parkinson's disease patients based on an internet hospital to solve the problems mentioned above. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a home-based information management system for Parkinson's disease patients based on an internet hospital, thereby resolving the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a home-based information management system for Parkinson's disease patients based on an internet hospital, comprising: The multi-source disease data collection module, based on the Internet hospital, records the daily changes in the patient's condition. It uses AI and video to demonstrate motor function and self-assessment scales, helping patients record their condition in detail.

[0007] The medication management module provides medication reminders, efficacy feedback, and side effect reporting functions based on daily records, helping patients to use medication correctly and report their condition in a timely manner.

[0008] The data analysis and early warning module analyzes the correlation between changes in the patient's condition and medication use based on the feedback data, and provides timely warnings of adverse reactions to exercise complications.

[0009] The remote collaborative diagnosis and treatment module, based on the analyzed data, supports doctors in assessing the condition and adjusting the treatment plan, and quickly identifies high-risk patients and provides treatment suggestions.

[0010] Preferably, the multi-source disease acquisition module includes: The patient data collection unit automatically collects objective motion data from patients using wearable devices, and simultaneously collects subjective self-reported data from patients via mobile terminals, including: Step 1.1 employs a data fusion algorithm to weight and integrate objective and subjective data, calculating a comprehensive disease score. The fusion algorithm formula is as follows: , in, This indicates the overall condition score. This represents the objective score obtained from sensor data. This indicates the scores obtained by the patient through self-reporting. Indicates the weighting parameter; Step 1.2 Update the weight parameters online using the recursive least squares method. To adapt to changes in data characteristics, the gain vector is calculated as follows: , in, This represents the error covariance matrix from the previous step. Represents the input vector. This represents the input difference vector. The formula for the weight parameter update algorithm, representing the forgetting factor, is as follows: , in, = , indicating the first The observed or target value of the step. This represents the updated weight parameters; The formula for the error covariance matrix update algorithm is as follows: .

[0011] Preferably, the AI-assisted movement function display unit intelligently analyzes the collected patient movement video data, extracts movement features, and intuitively displays the movement status. The implementation steps are as follows: Step 1.3 involves segmenting the acquired video into frames, scaling the images, converting them to grayscale, and denoising them. A CNN is then used to extract features from the preprocessed video frames. The algorithm formula is as follows: , in, Indicates the first The input feature map of the layer, Indicates the first Layer convolution kernel, Indicates the first Layer bias term, This represents the activation function. Indicates the upper level The input or feature map at location The value, This indicates the position index of the output feature map. A two-dimensional index representing the convolution kernel; Step 1.4 uses the features extracted by CNN and the results of motion state analysis to generate a motor function score, a motion trend graph, and abnormal detection prompts, which show the patient's motor ability and changes. The patient's medical record display unit uses a database or real-time data streaming platform to integrate data from multiple sources, forming a complete record of the patient's medical data. It generates a detailed and intuitive patient medical record for patients to manage themselves and doctors to monitor remotely. The system automatically generates a daily medical record based on the integrated data.

[0012] Preferably, the medication management module includes: The medication reminder unit, through data collection and analysis, ensures that patients take their medications on time and in the correct dosage. Its implementation steps and key technologies are as follows: Step 2.1 The patient enters the drug name, dosage, and medication time information. The system automatically records the patient's medication plan and sends daily medication reminders through the reminder system. Step 2.11 Generate a detailed medication schedule based on the patient's medication plan and the doctor's advice. The system will automatically push medication reminder notifications at the set time points. Step 2.12 Remind patients to take their medication on time via SMS, APP notifications, and sound prompts. Monitor patients in real time to ensure they take their medication on time. If they fail to take their medication on time, send a reminder notification again.

[0013] Preferably, the drug efficacy feedback unit monitors the patient's response to the drug in real time and helps the doctor adjust the treatment plan. Its implementation steps and algorithm are as follows: Step 2.2 Acquire the patient's physiological data through wearable sensors, smart devices, or physiological monitoring devices; Step 2.21 Patients record feedback on the effects of the medication through mobile apps, online questionnaires, etc. Step 2.22 Monitor the improvement of patients' motor function through motion capture and posture assessment. This is especially important for patients with certain neurological disorders, where improved motor function is a key indicator of drug efficacy. The algorithms for motion capture and posture assessment include: The system tracks the patient's movements in real time using sensors, calculating parameters such as speed, acceleration, and angle. The acceleration calculation formula is as follows: , in, Represents the acceleration vector. This represents the patient's position vector in three-dimensional space. Indicates time, Operators for differentiation; Formula for calculating angular velocity:

[0014] in, Indicates angular velocity. Indicates angle; Step 2.23 Use psychological assessment scales and emotion recognition tools to monitor the patient's psychological state and emotional changes, and assess the effect of medication on psychological state; Step 2.24 uses a Bayesian inference algorithm to update the probability estimate of drug efficacy by combining the patient's historical data and real-time monitoring information. The Bayesian inference algorithm formula is as follows: , in, Represents given data Post-drug effects The posterior probability, Indicates the effect of a given drug In this case, the observed data The likelihood function, Indicates drug efficacy The prior probability, Indicates observed data The marginal probability.

[0015] Preferably, the side effect reporting unit is used to collect and analyze information on possible side effects of patients to help doctors intervene in a timely manner. The implementation steps and algorithms are as follows: Step 2.3 Patients report side effects of the medication through self-reporting and sensor devices. The reported data includes adverse reactions, their frequency, and severity. Step 2.31: Based on the set side effect threshold, the system determines whether serious side effects exist, analyzes the side effect pattern using a decision tree algorithm, and automatically alerts the doctor. The decision tree algorithm includes: Each node of the decision tree splits the data based on a certain feature value, and the quality of the split is measured by information gain. , in, Represents the dataset entropy, Indicates that feature A takes the value of a subset of Representing a subset Entropy; Step 2.32 uses a multi-level weighted system to score the frequency and severity of side effects, and comprehensively derives a side effect risk score. The formula for calculating the side effect score is as follows: , in, Score the risk of side effects. A score indicating the severity of side effects. A score indicating the frequency of side effects. , This represents the weighting coefficient.

[0016] Preferably, the data analysis and early warning module further includes: The medication history and trend analysis unit reviews the patient's medication history and analyzes the trend of drug efficacy over time. The implementation steps are as follows: Step 3.1 Store the patient's medication history, efficacy feedback, medication reminders, and side effect reports in the database; Step 3.11 Utilize exponential smoothing analysis to track trends in drug efficacy during patient medication use. Based on historical data, provide insights into the changing trends of drug efficacy over time and assist physicians in adjusting treatment plans. The exponential smoothing algorithm includes: In standard exponential smoothing, the smoothing coefficient is a fixed value. In drug efficacy monitoring, the data fluctuates greatly. Adaptive exponential smoothing can dynamically adjust the smoothing coefficient according to the real-time prediction error, so that the model can respond quickly when the data changes abruptly. , , in, Indicates time Actual efficacy observation values, Indicates time Smoothed or predicted values, This represents the smoothed value from the previous time step. Represents the time-varying smoothing coefficient. Represents the basic smoothing coefficient. This represents the error sensitivity parameter.

[0017] Preferably, the data analysis and early warning module further includes: The prediction and trend analysis unit, based on historical and real-time monitoring data, uses the NARX model to analyze and predict the changing trends of drug efficacy and assess the improvement of patients' motor function. Its NARX model combines historical disease data and external inputs, and uses a nonlinear function to capture the complex relationship between the two. The model formula is as follows: , in, Indicates time The disease indicators, Indicates time Drug usage data, Indicates the preceding Real-time patient data, Indicates the preceding Real-time drug use data, Nonlinear mapping function, Indicates the error term; The anomaly detection and early warning unit monitors the patient's motor function improvement and changes in side effects in real time, and identifies potential abnormalities and risks through anomaly detection algorithms. The specific steps are as follows: Step 3.2 Anomaly detection in SVDD involves constructing a hypersphere with a minimum radius in a high-dimensional space to enclose normal data; data exceeding this sphere is considered anomaly. , in, Represents the radius of the hypersphere, and represents the boundary of normal data. The center of the hypersphere is represented by the center of the hypersphere. This represents a slack variable, which allows for a certain amount of outlier data. Represents the regularization parameter; Step 3.21 When the anomaly detection module detects abnormal data, the system will automatically trigger an early warning notification to promptly notify the doctor or relevant personnel for handling. The early warning information may include a detailed description of the abnormal data, a risk assessment, and recommended handling measures; The report generation and decision support unit presents the analysis results to doctors or medical personnel in an easy-to-understand manner, providing information to support decision-making and help adjust treatment plans. Report generation methods include: charts, key indicators, risk levels, and recommended actions. The feedback and optimization unit collects feedback from doctors and patients to continuously optimize models and algorithms, thereby improving the accuracy of prediction and analysis. Feedback methods include: the system collects feedback from doctors on the accuracy of early warnings, the effectiveness of decision-making suggestions, and the actual condition of patients through APP, SMS, online questionnaires, and direct input from medical staff; at the same time, it collects opinions from patients on treatment adjustments and monitoring experiences.

[0018] Preferably, the remote collaborative diagnosis and treatment module includes: The real-time communication and data transmission unit is used for the real-time and secure transmission of audio, video, text, and medical data between the doctor's and patient's terminals, ensuring low latency and high reliability during the interaction process. Low latency can be achieved using a dynamic latency compensation algorithm, the formula of which is as follows: , in, This indicates the currently measured network latency. Indicates the magnitude of the delay; Indicates the adjustment coefficient; The multimodal interaction fusion unit preprocesses and extracts features from audio, video, and text respectively. It then employs an attention-based algorithm to dynamically weight and fuse features from different modalities, generating a fused interaction evaluation index. The attention calculation formula is as follows: , in, These represent the query, key, and value vectors, respectively. The dimension of the key vector Fusion formula: , in, These represent features extracted from audio, video, and text, respectively. This represents the weights of each mode. This represents the similarity scoring function; The remote diagnosis and intelligent decision-making unit, based on real-time data, historical cases, and abnormal detection results, calls the algorithm in step 4.1 to conduct risk assessment, automatically generates diagnostic reports and treatment suggestions, and supports online doctor consultations.

[0019] Preferably, the feedback and optimization unit collects feedback from doctors and patients during the interaction process, dynamically adjusts system parameters and models, and continuously optimizes the accuracy of remote interaction and diagnosis, as well as the user experience. Its dynamic adjustment method includes: Step 4.2 Continuously update the existing model through incremental learning, using small batches of data for online updates, employing the following update formula: , in, Indicates the model parameters at the current time. Indicates the learning rate. Indicates the current data The corresponding loss function, This represents the gradient of the loss function with respect to the parameters; To prevent the model from forgetting previous knowledge during the learning process on new data, EWC introduces a penalty term: , in, Overall objective function, The loss function representing the new data, This represents the parameter values ​​learned from previous tasks. Represents the information matrix. Indicates the regularization strength; Step 4.3 Based on feedback data from doctors and patients, fine-grained adjustments are made to the model's hyperparameters and weights. When the system detects changes in model error, the learning rate can be dynamically adjusted. , in, Represents the original learning rate. This indicates the amount of error change detected in the feedback. Indicates the adjustment factor; For regularization or threshold parameters, the exponential decay method can also be used: , in, Indicates the current regularization parameter. Indicates the adjustment coefficient; Step 4.4 Integrate the prediction results of multiple models in the same task, make full use of the advantages of each model, and perform a weighted average of the prediction results of multiple models, as shown in the following formula: , in, Indicates the first Each model is based on the input data. The prediction This indicates the number of models participating in the fusion. This represents the corresponding fusion weights. It can be dynamically adjusted based on the performance and feedback of each model on historical data; The prediction results of each model are dynamically weighted using an attention mechanism: , in, The scoring function for the prediction result of the i-th model.

[0020] This invention provides a home-based information management system for Parkinson's disease patients based on an internet hospital. It has the following beneficial effects: 1. This invention continuously tracks changes in a patient's condition through real-time data collection from wearable devices and mobile terminals. With the help of IoT hospital platform technology, the patient's daily condition data can be automatically recorded and uploaded, and both doctors and patients can view it in real time, ensuring the timeliness and accuracy of information updates. This real-time monitoring can greatly reduce the risk of a patient's condition being overlooked.

[0021] 2. This invention comprehensively collects patients' exercise data, drug efficacy feedback, and side effect reports, and uses advanced data fusion and analysis methods to accurately capture subtle fluctuations in patients' daily symptoms. Using incremental learning and model fine-tuning technology, the system can adjust and optimize the prediction model in real time, thereby more sensitively detecting subtle changes in the condition.

[0022] 3. This invention utilizes multiple sensors and intelligent devices to monitor the patient's physiological state in real time. Combined with a data analysis and early warning module, it automatically detects drug side effects and exercise complications. Through the NARX model and anomaly detection algorithm, the system can issue timely warnings when potential problems occur, allowing doctors to respond quickly and avoid missing the best intervention opportunity.

[0023] 4. This invention, through its remote collaborative diagnosis and treatment module, utilizes real-time audio and video communication, data transmission, and multimodal interactive fusion technology to break through the limitations of traditional face-to-face consultations. Doctors can obtain patients' real-time data anytime, anywhere, and interact with patients through an online platform. The system supports remote consultations, allowing doctors to quickly assess the condition and adjust treatment plans. It can also promptly identify high-risk patients and provide targeted treatment suggestions.

[0024] 5. This invention integrates data from different sensors, devices, and patient self-reports through a centralized data management platform to form a comprehensive and standardized medical record. Patients' daily medical data, medication efficacy feedback, and exercise analysis can all be centrally managed, ensuring the integrity, accuracy, and sustainability of the data. In addition, it provides doctors with a tool for real-time monitoring and analysis, promoting data-driven treatment adjustments and the analysis and research of long-term treatment effects. Attached Figure Description

[0025] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0026] To enable those skilled in the art to understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort should fall within the scope of protection of the present invention.

[0027] The present invention will now be described in detail with reference to the accompanying drawings: Example: Please see the appendix Figure 1 This invention provides a home-based information management system for Parkinson's disease patients based on an internet hospital, comprising: The multi-source disease data collection module, based on the Internet hospital, records the daily changes in the patient's condition. It uses AI and video to demonstrate motor function and self-assessment scales, helping patients record their condition in detail.

[0028] The medication management module provides medication reminders, efficacy feedback, and side effect reporting functions based on daily records, helping patients to use medication correctly and report their condition in a timely manner.

[0029] The data analysis and early warning module analyzes the correlation between changes in the patient's condition and medication use based on the feedback data, and provides timely warnings of adverse reactions to exercise complications.

[0030] The remote collaborative diagnosis and treatment module, based on the analyzed data, supports doctors in assessing the condition and adjusting the treatment plan, and quickly identifies high-risk patients and provides treatment suggestions.

[0031] S1, the multi-source disease data acquisition module includes: The patient data collection unit automatically collects objective motion data from patients using wearable devices, and simultaneously collects subjective self-reported data from patients via mobile terminals, including: Step 1.1 employs a data fusion algorithm to weight and integrate objective and subjective data, calculating a comprehensive disease score. The fusion algorithm formula is as follows: , in, This indicates the overall condition score. This represents the objective score obtained from sensor data. This indicates the scores obtained by the patient through self-reporting. Indicates the weighting parameter; Step 1.2 Update the weight parameters online using the recursive least squares method. To adapt to changes in data characteristics, the gain vector is calculated as follows: , in, This represents the error covariance matrix from the previous step. Represents the input vector. This represents the input difference vector. The formula for the weight parameter update algorithm, representing the forgetting factor, is as follows: , in, = , indicating the first The observed or target value of the step. This represents the updated weight parameters; The formula for the error covariance matrix update algorithm is as follows: .

[0032] The AI-assisted exercise function demonstration unit intelligently analyzes the collected patient exercise video data, extracts exercise features, and intuitively displays the exercise status. The implementation steps are as follows: Step 1.3 involves segmenting the acquired video into frames, scaling the images, converting them to grayscale, and denoising them. A CNN is then used to extract features from the preprocessed video frames. The algorithm formula is as follows: , in, Indicates the first The input feature map of the layer, Indicates the first Layer convolution kernel, Indicates the first Layer bias term, This represents the activation function. Indicates the upper level The input or feature map at location The value, This indicates the position index of the output feature map. A two-dimensional index representing the convolution kernel; Step 1.4 uses the features extracted by CNN and the results of motion state analysis to generate a motor function score, a motion trend graph, and abnormal detection prompts, which show the patient's motor ability and changes. The patient's medical record display unit uses a database or real-time data streaming platform to integrate data from multiple sources, forming a complete record of the patient's medical data. It generates a detailed and intuitive patient medical record for patients to manage themselves and doctors to monitor remotely. The system automatically generates a daily medical record based on the integrated data.

[0033] Self-assessment of the scale using voice: Step 1: User Authentication and Bootstrapping Patients log in to the system via voiceprint recognition. The system calls a pre-stored voiceprint model (such as an i-vector + x-vector hybrid architecture) for identity matching, with an error rate of <2%.

[0034] After logging in, a voice prompt will play: "Hello, today's medical assessment will now begin. Please ensure the environment is quiet. You will have 10 seconds to answer each question." Step 2: Loading the dynamic scale and voice interaction The scale template is automatically matched according to the patient's disease type (example: Parkinson's disease patient loads the UPDRS scale voice version).

[0035] Problem reporting logic: The system plays questions via TTS (Text-to-Speech), such as: "How often have you felt your hands trembling in the past week? 1 means very rarely, and 5 means very frequently." The patient answered in natural language, such as "Several times a day, especially when I pick up a water cup."

[0036] Speech parsing technology: Use a bidirectional LSTM+CRF model to extract keywords (e.g., “several times a day” corresponds to a score of 4, and “take a water cup” triggers an action-related tag).

[0037] Simultaneously analyze the consistency between speech emotion features (such as fundamental frequency and speech rate) and semantics. If a contradiction is detected (such as answering "no pain" but speaking in a painful tone), mark it as having low confidence and requiring manual review.

[0038] Step 3: Multimodal Action Verification (Optional) When the voice response involves motor symptoms (such as "unsteady gait"), the system triggers motion verification: Voice command: "Please stand up and walk five steps slowly. Your phone's camera will record your gait." The coordinates of skeletal nodes were extracted using the OpenPose algorithm, and parameters such as stride length and torso tilt angle were calculated and cross-validated with the speech self-assessment results.

[0039] Step 4: Real-time feedback and adaptive adjustment If the patient's response takes too long or is unclear, the system will initiate multiple rounds of clarification dialogue: Example: The patient answers "Sometimes I feel dizzy," and the system asks: "Does the dizziness accompany tinnitus? 1. Yes, 2. No."

[0040] Logical jump mechanism: If the scores of the same type of question are stable for three consecutive times (standard deviation < 0.5), then the frequency of questions in that dimension will be reduced.

[0041] Step 5: Data Fusion and Report Generation The speech analysis score, action analysis data, and historical records are input into the multilayer perceptron (MLP) model, which outputs a comprehensive severity index.

[0042] The system generates a voice summary report: "Your motor function score today has decreased by 12% compared to last week. We recommend adjusting your medication dosage and contacting your doctor." Specifically, by collecting real-time data through wearable devices and mobile terminals, changes in the patient's condition can be continuously tracked. With the help of Internet of Things (IoT) technology, the patient's daily condition data can be automatically recorded and uploaded, and both doctors and patients can view it in real time, ensuring the timeliness and accuracy of information updates. This real-time monitoring can greatly reduce the risk of the patient's condition being overlooked.

[0043] The medication management module includes: The medication reminder unit, through data collection and analysis, ensures that patients take their medications on time and in the correct dosage. Its implementation steps and key technologies are as follows: Step 2.1 The patient enters the drug name, dosage, and medication time information. The system automatically records the patient's medication plan and sends daily medication reminders through the reminder system. Step 2.11 Generate a detailed medication schedule based on the patient's medication plan and the doctor's advice. The system will automatically push medication reminder notifications at the set time points. Step 2.12 Remind patients to take their medication on time via SMS, APP notifications, and sound prompts. Monitor patients in real time to ensure they take their medication on time. If they fail to take their medication on time, send a reminder notification again.

[0044] The drug efficacy feedback unit monitors the patient's response to medication in real time and helps doctors adjust treatment plans. Its implementation steps and algorithms are as follows: Step 2.2 Acquire the patient's physiological data through wearable sensors, smart devices, or physiological monitoring devices; Step 2.21 Patients record feedback on the effects of the medication through mobile apps, online questionnaires, etc. Step 2.22 Monitor the improvement of patients' motor function through motion capture and posture assessment. This is especially important for patients with certain neurological disorders, where improved motor function is a key indicator of drug efficacy. The algorithms for motion capture and posture assessment include: The system tracks the patient's movements in real time using sensors, calculating parameters such as speed, acceleration, and angle. The acceleration calculation formula is as follows: , in, Represents the acceleration vector. This represents the patient's position vector in three-dimensional space. Indicates time, Operators for differentiation; Formula for calculating angular velocity:

[0045] in, Indicates angular velocity. Indicates angle; Step 2.23 Use psychological assessment scales and emotion recognition tools to monitor the patient's psychological state and emotional changes, and assess the effect of medication on psychological state; Step 2.24 uses a Bayesian inference algorithm to update the probability estimate of drug efficacy by combining the patient's historical data and real-time monitoring information. The Bayesian inference algorithm formula is as follows: , in, Represents given data Post-drug effects The posterior probability, Indicates the effect of a given drug In this case, the observed data The likelihood function, Indicates drug efficacy The prior probability, Indicates observed data The marginal probability.

[0046] The side effect reporting unit is used to collect and analyze information on possible side effects from patients, helping doctors to intervene in a timely manner. The implementation steps and algorithms are as follows: Step 2.3 Patients report side effects of the medication through self-reporting and sensor devices. The reported data includes adverse reactions, their frequency, and severity. Step 2.31: Based on the set side effect threshold, the system determines whether serious side effects exist, analyzes the side effect pattern using a decision tree algorithm, and automatically alerts the doctor. The decision tree algorithm includes: Each node of the decision tree splits the data based on a certain feature value, and the quality of the split is measured by information gain. , in, Represents the dataset entropy, Indicates that feature A takes the value of a subset of Representing a subset Entropy; Step 2.32 uses a multi-level weighted system to score the frequency and severity of side effects, and comprehensively derives a side effect risk score. The formula for calculating the side effect score is as follows: , in, Score the risk of side effects. A score indicating the severity of side effects. A score indicating the frequency of side effects. , This represents the weighting coefficient.

[0047] The data analysis and early warning module includes: The medication history and trend analysis unit reviews a patient's medication history and analyzes the trend of drug efficacy over time. The implementation steps are as follows: Step 3.1 Store the patient's medication history, efficacy feedback, medication reminders, and side effect reports in the database; Step 3.11 Utilize exponential smoothing analysis to track trends in drug efficacy during patient medication use. Based on historical data, provide insights into the changing trends of drug efficacy over time and assist physicians in adjusting treatment plans. The exponential smoothing algorithm includes: In standard exponential smoothing, the smoothing coefficient is a fixed value. In drug efficacy monitoring, the data fluctuates greatly. Adaptive exponential smoothing can dynamically adjust the smoothing coefficient according to the real-time prediction error, so that the model can respond quickly when the data changes abruptly. , , in, Indicates time Actual efficacy observation values, Indicates time Smoothed or predicted values, This represents the smoothed value from the previous time step. Represents the time-varying smoothing coefficient. Represents the basic smoothing coefficient. This represents the error sensitivity parameter.

[0048] Specifically, this invention comprehensively collects patients' exercise data, drug efficacy feedback, and side effect reports, and employs advanced data fusion and analysis methods to accurately capture subtle fluctuations in patients' daily symptoms. Using incremental learning and model fine-tuning techniques, the system can adjust and optimize the prediction model in real time, thereby more sensitively detecting subtle changes in the condition.

[0049] The data analysis and early warning module also includes: The prediction and trend analysis unit, based on historical and real-time monitoring data, uses the NARX model to analyze and predict the changing trends of drug efficacy and assess the improvement of patients' motor function. Its NARX model combines historical disease data and external inputs, and uses a nonlinear function to capture the complex relationship between the two. The model formula is as follows: , in, Indicates time The disease indicators, Indicates time Drug usage data, Indicates the preceding Real-time patient data, Indicates the preceding Real-time drug use data, Nonlinear mapping function, Indicates the error term; The anomaly detection and early warning unit monitors the patient's motor function improvement and changes in side effects in real time, and identifies potential abnormalities and risks through anomaly detection algorithms. The specific steps are as follows: Step 3.2 Anomaly detection in SVDD involves constructing a hypersphere with a minimum radius in a high-dimensional space to enclose normal data; data exceeding this sphere is considered anomaly. , in, Represents the radius of the hypersphere, and represents the boundary of normal data. The center of the hypersphere is represented by the center of the hypersphere. This represents a slack variable, which allows for a certain amount of outlier data. Represents the regularization parameter; Step 3.21 When the anomaly detection module detects abnormal data, the system will automatically trigger an early warning notification to promptly notify the doctor or relevant personnel for handling. The early warning information may include a detailed description of the abnormal data, a risk assessment, and recommended handling measures; The report generation and decision support unit presents the analysis results to doctors or medical personnel in an easy-to-understand manner, providing information to support decision-making and help adjust treatment plans. Report generation methods include: charts, key indicators, risk levels, and recommended actions. The feedback and optimization unit collects feedback from doctors and patients to continuously optimize models and algorithms, thereby improving the accuracy of prediction and analysis. Feedback methods include: the system collects feedback from doctors on the accuracy of early warnings, the effectiveness of decision-making suggestions, and the actual condition of patients through APP, SMS, online questionnaires, and direct input from medical staff; at the same time, it collects opinions from patients on treatment adjustments and monitoring experiences.

[0050] Specifically, this invention utilizes multiple sensors and intelligent devices to monitor the patient's physiological state in real time, and combines data analysis and early warning modules to automatically detect drug side effects and exercise complications. Through the NARX model and anomaly detection algorithm, the system can issue timely warnings when potential problems occur, allowing doctors to respond quickly and avoid missing the best intervention opportunity.

[0051] The remote collaborative diagnosis and treatment module includes: The real-time communication and data transmission unit is used for the real-time and secure transmission of audio, video, text, and medical data between the doctor's and patient's terminals, ensuring low latency and high reliability during the interaction process. Low latency can be achieved using a dynamic latency compensation algorithm, the formula of which is as follows: , in, This indicates the currently measured network latency. Indicates the magnitude of the delay; Indicates the adjustment coefficient; The multimodal interaction fusion unit preprocesses and extracts features from audio, video, and text respectively. It then employs an attention-based algorithm to dynamically weight and fuse features from different modalities, generating a fused interaction evaluation index. The attention calculation formula is as follows: , in, These represent the query, key, and value vectors, respectively. The dimension of the key vector Fusion formula: , in, These represent features extracted from audio, video, and text, respectively. This represents the weights of each mode. This represents the similarity scoring function; The remote diagnosis and intelligent decision-making unit, based on real-time data, historical cases, and abnormal detection results, calls the algorithm in step 4.1 to conduct risk assessment, automatically generates diagnostic reports and treatment suggestions, and supports online doctor consultations.

[0052] The feedback and optimization unit collects feedback from doctors and patients during interactions, dynamically adjusts system parameters and models, and continuously optimizes the accuracy of remote interaction and diagnosis, as well as the user experience. Its dynamic adjustment methods include: Step 4.2 Continuously update the existing model through incremental learning, using small batches of data for online updates, employing the following update formula: , in, Indicates the model parameters at the current time. Indicates the learning rate. Indicates the current data The corresponding loss function, This represents the gradient of the loss function with respect to the parameters; To prevent the model from forgetting previous knowledge during the learning process on new data, EWC introduces a penalty term: , in, Overall objective function, The loss function representing the new data, This represents the parameter values ​​learned from previous tasks. Represents the information matrix. Indicates the regularization strength; Step 4.3 Based on feedback data from doctors and patients, fine-grained adjustments are made to the model's hyperparameters and weights. When the system detects changes in model error, the learning rate can be dynamically adjusted. , in, Represents the original learning rate. This indicates the amount of error change detected in the feedback. Indicates the adjustment factor; For regularization or threshold parameters, the exponential decay method can also be used: , in, Indicates the current regularization parameter. Indicates the adjustment coefficient; Step 4.4 Integrate the prediction results of multiple models in the same task, make full use of the advantages of each model, and perform a weighted average of the prediction results of multiple models, as shown in the following formula: , in, Indicates the first Each model is based on the input data. The prediction This indicates the number of models participating in the fusion. This represents the corresponding fusion weights. It can be dynamically adjusted based on the performance and feedback of each model on historical data; The prediction results of each model are dynamically weighted using an attention mechanism: , in, The scoring function for the prediction result of the i-th model.

[0053] Specifically, this invention, through a remote collaborative diagnosis and treatment module, breaks through the limitations of traditional face-to-face consultations by utilizing real-time audio and video communication, data transmission, and multimodal interactive fusion technology. Doctors can access patients' real-time data anytime, anywhere, and interact with patients through an online platform. The system supports remote consultations, allowing doctors to quickly assess conditions and adjust treatment plans. It can also promptly identify high-risk patients and provide targeted treatment recommendations. Through a centralized data management platform, data from different sensors, devices, and patient self-reports are integrated to form comprehensive and standardized medical records. Patients' daily medical data, medication efficacy feedback, and exercise analysis can all be centrally managed, ensuring data integrity, accuracy, and sustainability. Furthermore, it provides doctors with a tool for real-time monitoring and analysis, promoting data-driven treatment adjustments and long-term treatment effect analysis research.

[0054] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A home-based information management system for Parkinson's disease patients based on an internet hospital, characterized in that, The system includes a data analysis and early warning module, which includes: The medication history and trend analysis unit reviews a patient's medication history and analyzes the trend of drug efficacy over time. The specific steps are as follows: The database stores patients' medication history, efficacy feedback, medication reminders, and side effect reports; it uses exponential smoothing analysis to analyze the trend of changes in drug efficacy during patient medication, and provides the trend of drug effect changes over time based on historical data, helping doctors adjust treatment plans; The algorithm for exponential smoothing includes: Adaptive exponential smoothing dynamically adjusts the smoothing coefficient based on real-time prediction errors: ， ， in, Indicates time Actual efficacy observation values, Indicates time Smoothed or predicted values, This represents the smoothed value from the previous time step. Represents the time-varying smoothing coefficient. Represents the basic smoothing coefficient. This represents the error sensitivity parameter; The prediction and trend analysis unit, based on historical and real-time monitoring data, analyzes and predicts the changing trend of drug efficacy using the NARX model, and assesses the improvement of the patient's motor function. The NARX model combines historical disease data and external inputs, and uses a nonlinear function to capture the complex relationship between the two. The model formula is as follows: ， in, Indicates time The disease indicators, Indicates time Drug usage data, Indicates the preceding Real-time patient data, Indicates the preceding Real-time drug use data, Nonlinear mapping function, Indicates the error term; The anomaly detection and early warning unit monitors the patient's motor function improvement and changes in side effects in real time, and identifies potential abnormalities and risks through anomaly detection algorithms. The specific steps are as follows: Anomaly detection in SVDD involves constructing a hypersphere with a minimum radius in a high-dimensional space to enclose normal data; data outside this hypersphere is considered anomalous. ， in, Represents the radius of the hypersphere, and represents the boundary of normal data. The center of the hypersphere is represented by the center of the hypersphere. This represents a slack variable, which allows for a certain amount of outlier data. Represents the regularization parameter; When the anomaly detection module detects abnormal data, the system will automatically trigger an early warning notification to promptly notify doctors or relevant personnel for handling. The early warning information may include a detailed description of the abnormal data, a risk assessment, and recommended handling measures.

2. The home information management system for Parkinson's disease patients based on an internet hospital as described in claim 1, characterized in that, The data analysis and early warning module also includes: The report generation and decision support unit presents the analysis results to doctors or medical personnel in an easy-to-understand way, providing information to support decision-making and help adjust treatment plans. Report generation methods include: charts, key indicators, risk levels, and recommended actions. The feedback and optimization unit collects feedback from doctors and patients to continuously optimize models and algorithms, thereby improving the accuracy of prediction and analysis. Feedback methods include: the system collects feedback from doctors on the accuracy of early warnings, the effectiveness of decision-making suggestions, and the actual condition of patients through APP, SMS, online questionnaires, and direct input from medical staff; at the same time, it collects opinions from patients on treatment adjustments and monitoring experiences.

3. The home information management system for Parkinson's disease patients based on an internet hospital as described in claim 1, characterized in that, The system also includes: The multi-source disease data collection module, based on the Internet hospital, records the daily changes in the patient's condition and uses AI and video to demonstrate motor function and self-assessment scales to help patients record their condition in detail. The medication management module provides medication reminders, efficacy feedback, and side effect reporting functions based on daily records, helping patients to use medication in a standardized manner and report their condition in a timely manner. The data analysis and early warning module analyzes the correlation between changes in the patient's condition and medication based on the feedback data, and provides timely early warning of adverse reactions to exercise complications. The remote collaborative diagnosis and treatment module, based on the analyzed data, supports doctors in assessing the condition and adjusting the treatment plan, and quickly identifies high-risk patients and provides treatment suggestions.

4. A home-based information management system for Parkinson's disease patients based on an internet hospital, as described in claim 3, is characterized in that... The multi-source disease acquisition module includes: The patient data collection unit, based on the internet hospital, automatically collects objective motion data from patients using wearable devices, and simultaneously collects subjective self-reported data from patients via mobile terminals, including: Step 1.1 employs a data fusion algorithm to weight and integrate objective and subjective data, calculating a comprehensive disease score. The fusion algorithm formula is as follows: ， in, This indicates the overall condition score. This represents the objective score obtained from sensor data. This represents the scores obtained by patients through self-reporting. Indicates the weighting parameter; Step 1.2 Update the weight parameters online using the recursive least squares method. To adapt to changes in data characteristics, the gain vector is calculated as follows: , in, This represents the error covariance matrix from the previous step. Represents the input vector. This represents the input difference vector. The formula for the weight parameter update algorithm, representing the forgetting factor, is as follows: ， in, = , indicating the first The observed or target value of the step. This represents the updated weight parameters; The formula for the error covariance matrix update algorithm is as follows: .

5. A home-based information management system for Parkinson's disease patients based on an internet hospital, as described in claim 4, is characterized in that... The AI-assisted movement function display unit intelligently analyzes the collected patient movement video data, extracts movement features, and intuitively displays the movement status. The implementation steps are as follows: Step 1.3 involves segmenting the acquired video into frames, scaling the images, converting them to grayscale, and denoising them. A CNN is then used to extract features from the preprocessed video frames. The algorithm formula is as follows: ， in, Indicates the first The input feature map of the layer, Indicates the first Layer convolution kernel, Indicates the first Layer bias term, This represents the activation function. Indicates the upper level The input or feature map at location The value, This indicates the position index of the output feature map. A two-dimensional index representing the convolution kernel; Step 1.4 uses the features extracted by CNN and the results of motion state analysis to generate a motor function score, a motion trend graph, and abnormal detection prompts, which show the patient's motor ability and changes. The patient's medical record display unit uses a database or real-time data streaming platform to integrate data from multiple sources, forming a complete record of the patient's medical data. It generates a detailed and intuitive patient medical record for patients to manage themselves and doctors to monitor remotely. The system automatically generates a daily medical record based on the integrated data.

6. A home-based information management system for Parkinson's disease patients based on an internet hospital, as described in claim 3, is characterized in that... The medication management module includes: The medication reminder unit, through data collection and analysis, ensures that patients take their medications on time and in the correct dosage. Its implementation steps and key technologies are as follows: Step 2.1 The patient enters the drug name, dosage, and medication time information. The system automatically records the patient's medication plan and sends daily medication reminders through the reminder system. Step 2.11 Generate a detailed medication schedule based on the patient's medication plan and the doctor's advice. The system will automatically push medication reminder notifications at the set time points. Step 2.12 Remind patients to take their medication on time via SMS, APP notifications, and sound prompts. Monitor patients in real time to ensure they take their medication on time. If they fail to take their medication on time, send a reminder notification again.

7. A home-based information management system for Parkinson's disease patients based on an internet hospital, as described in claim 6, is characterized in that... The drug efficacy feedback unit monitors the patient's response to the drug in real time and helps doctors adjust the treatment plan. Its implementation steps and algorithm are as follows: Step 2.2 Acquire the patient's physiological data through wearable sensors, smart devices, or physiological monitoring devices; Step 2.21 Patients record feedback on the effects of the medication through mobile apps, online questionnaires, etc. Step 2.22 Monitor the improvement of patients' motor function through motion capture and posture assessment. This is especially important for patients with certain neurological disorders, where improved motor function is a key indicator of drug efficacy. The algorithms for motion capture and posture assessment include: The system tracks the patient's movements in real time using sensors, calculating parameters such as speed, acceleration, and angle. The acceleration calculation formula is as follows: , in, Represents the acceleration vector. This represents the patient's position vector in three-dimensional space. Indicates time, Operators for differentiation; Formula for calculating angular velocity: in, Indicates angular velocity, Indicates angle; Step 2.23 Use psychological assessment scales and emotion recognition tools to monitor the patient's psychological state and emotional changes, and assess the effect of medication on psychological state; Step 2.24 uses a Bayesian inference algorithm to update the probability estimate of drug efficacy by combining the patient's historical data and real-time monitoring information. The Bayesian inference algorithm formula is as follows: ， in, Represents given data Post-drug effects The posterior probability, Indicates the effect of a given drug In this case, the observed data The likelihood function, Indicates drug efficacy The prior probability, Indicates observed data The marginal probability.

8. A home-based information management system for Parkinson's disease patients based on an internet hospital, as described in claim 7, is characterized in that... The side effect reporting unit is used to collect and analyze information on possible side effects of patients to help doctors intervene in a timely manner. The implementation steps and algorithms are as follows: Step 2.3 Patients report side effects of the medication through self-reporting and sensor devices. The reported data includes adverse reactions, their frequency, and severity. Step 2.31: Based on the set side effect threshold, the system determines whether serious side effects exist, analyzes the side effect pattern using a decision tree algorithm, and automatically alerts the doctor. The decision tree algorithm includes: Each node of the decision tree splits the data based on a certain feature value, and the quality of the split is measured by information gain. , in, Represents the dataset entropy, Indicates that feature A takes the value of a subset of Representing a subset Entropy; Step 2.32 uses a multi-level weighted system to score the frequency and severity of side effects, and comprehensively derives a side effect risk score. The formula for calculating the side effect score is as follows: ， in, Score the risk of side effects. A score indicating the severity of side effects. A score indicating the frequency of side effects. , This represents the weighting coefficient.

9. A home-based information management system for Parkinson's disease patients based on an internet hospital, as described in claim 8, is characterized in that... The remote collaborative diagnosis and treatment module includes: The real-time communication and data transmission unit is used for the real-time and secure transmission of audio, video, text, and medical data between the doctor's and patient's terminals, ensuring low latency and high reliability during the interaction process. Low latency can be achieved using a dynamic latency compensation algorithm, the formula of which is as follows: ， in, This indicates the currently measured network latency. Indicates the magnitude of the delay; Indicates the adjustment coefficient; The multimodal interaction fusion unit preprocesses and extracts features from audio, video, and text respectively. It then employs an attention-based algorithm to dynamically weight and fuse features from different modalities, generating a fused interaction evaluation index. The attention calculation formula is as follows: ， in, These represent the query, key, and value vectors, respectively. The dimension of the key vector Fusion formula: ， in, These represent features extracted from audio, video, and text, respectively. This represents the weights of each mode. This represents the similarity scoring function; The remote diagnosis and intelligent decision-making unit, based on real-time data, historical cases, and abnormal detection results, calls the algorithm in step 4.1 to conduct risk assessment, automatically generates diagnostic reports and treatment suggestions, and supports online doctor consultations.

10. A home-based information management system for Parkinson's disease patients based on an internet hospital, as described in claim 9, is characterized in that... The feedback and optimization unit collects feedback from doctors and patients during the interaction process, dynamically adjusts system parameters and models, and continuously optimizes the accuracy of remote interaction and diagnosis, as well as the user experience. Its dynamic adjustment methods include: Step 4.2 Continuously update the existing model through incremental learning, using small batches of data for online updates, employing the following update formula: , in, Indicates the model parameters at the current time. Indicates the learning rate. Indicates the current data The corresponding loss function, This represents the gradient of the loss function with respect to the parameters; To prevent the model from forgetting previous knowledge during the learning process on new data, EWC introduces a penalty term: ， in, Overall objective function, The loss function representing the new data, This represents the parameter values ​​learned from previous tasks. Represents the information matrix. Indicates the regularization strength; Step 4.3 Based on feedback data from doctors and patients, fine-grained adjustments are made to the model's hyperparameters and weights. When the system detects changes in model error, the learning rate can be dynamically adjusted. ， in, Represents the original learning rate. This indicates the amount of error change detected in the feedback. Indicates the adjustment factor; For regularization or threshold parameters, the exponential decay method can also be used: ， in, Indicates the current regularization parameter. Indicates the adjustment coefficient; Step 4.4 Integrate the prediction results of multiple models in the same task, make full use of the advantages of each model, and perform a weighted average of the prediction results of multiple models, as shown in the following formula: ， in, Indicates the first Each model is based on the input data. The prediction This indicates the number of models participating in the fusion. This represents the corresponding fusion weights. It can be dynamically adjusted based on the performance and feedback of each model on historical data; The prediction results of each model are dynamically weighted using an attention mechanism: ， in, The scoring function for the prediction result of the i-th model.

Images