Comprehensive monitoring and evaluation system and method for typical expression of hepatolenticular degeneration

Abstract

The invention discloses a comprehensive monitoring and evaluation system and method for a hepatolenticular degeneration patient, relates to the technical field of medical data analysis, and solves the problems that in the prior art, single image or motion detection is often depended, multi-source data comprehensive analysis is ignored, and the detection accuracy is high. The accuracy and the efficiency of the comprehensive monitoring and evaluation method for the hepatolenticular degeneration patient are low. Sensor data passes through a single-modal evaluation unit to obtain a plurality of modal anomaly scores, weighted summation is carried out on the modal anomaly scores to calculate an original comprehensive anomaly score so as to determine a normalized comprehensive score, a structured comprehensive evaluation report and an alarm signal are generated, and a multi-modal anomaly result influencing the illness state is analyzed through multi-source data. And the influence degree of the multi-modal abnormal result on illness state analysis is dynamically adjusted, so that illness state analysis can fit the actual condition of the patient, the comprehensive monitoring result is more accurate, and the accuracy of the comprehensive monitoring evaluation method for the hepatolenticular degeneration patient is improved.

Description

Technical Field

[0001] This invention belongs to the field of medical data analysis technology, specifically a comprehensive monitoring and evaluation system and method for typical manifestations of Wilson's disease. Background Technology

[0002] Wilson's disease (WD) is an autosomal recessive genetic disorder that causes abnormal copper metabolism, leading to abnormal deposition of copper ions in the liver, eyes, and central nervous system. Clinical features of the disease include liver damage, neurological symptoms, and the characteristic KF rings in the eye. KF rings are one of the important diagnostic criteria for WD. In addition, WD patients often experience neuromotor symptoms such as gait abnormalities and balance disorders, which significantly impact their daily lives and functional abilities.

[0003] In the past, the detection of KF rings relied on manual observation by physicians using a slit-lamp microscope. However, due to subjective factors, the results were often inconsistent and lacked quantifiable evaluation standards. In recent years, automated KF ring detection technologies based on digital image processing and deep learning have made significant progress. However, these approaches are typically limited to processing ocular images and fail to integrate with other symptoms or disease manifestations, thus failing to provide a comprehensive disease assessment. Besides ocular manifestation detection, the assessment of movement disorders in WD patients is equally crucial. Traditional movement disorder assessment methods, such as clinical scoring systems, suffer from high equipment costs, operational complexity, and difficulty in achieving continuous, real-time monitoring. In recent years, wearable sensor-based motion detection technologies have gradually emerged as an effective solution. However, these technologies typically focus on the collection and analysis of motion data and fail to integrate with other clinical indicators such as ocular KF ring detection, thus failing to comprehensively assess the overall symptoms of WD.

[0004] Most existing technical solutions focus on single-aspect diagnosis or monitoring, neglecting comprehensive analysis of multi-source data. This results in low accuracy and efficiency of comprehensive monitoring and assessment methods for typical manifestations of Wilson's disease. Therefore, further improvements are needed for comprehensive monitoring and assessment methods for typical manifestations of Wilson's disease. Summary of the Invention

[0005] This application aims to address at least one of the technical problems existing in the prior art; to this end, this application proposes a comprehensive monitoring and evaluation system and method for typical manifestations of Wilson's disease, which is used to solve the technical problem that existing comprehensive monitoring and evaluation methods for typical manifestations of Wilson's disease often rely on single images or motion detection and ignore the comprehensive analysis of multi-source data, resulting in low accuracy and efficiency.

[0006] To achieve the above objectives, the first aspect of this application provides a comprehensive monitoring and assessment system for typical manifestations of Wilson's disease, including a user module, an information processing module, and a physician module;

[0007] The user module includes a user information management unit, an interactive data acquisition and control unit, a data integration and transmission unit, and a local display unit.

[0008] The user information management unit: Based on the application on the smart terminal device, it realizes user identity registration and information update, and uploads basic user information: The basic user information includes name, gender, age, duration of illness and stage of illness;

[0009] The interactive acquisition and control unit is used to acquire sensor data, including eye images, waist acceleration, waist angular velocity, foot acceleration, foot angular velocity, and time-series data of plantar pressure distribution.

[0010] The data integration and transmission unit encrypts the collected user basic information and sensor data, and then uploads them to the information processing module in real time via data transmission methods, including BLE, Wi-Fi, and cellular networks.

[0011] The local display unit displays the test progress, data acquisition status, and brief feedback in real time on the terminal device application.

[0012] The information processing module, located on a cloud server, is used for data processing and evaluation, and includes a data preprocessing unit, a feature extraction unit, a single-modal evaluation unit, a multimodal fusion unit, a report generation unit, an alarm unit, and a storage unit.

[0013] The data preprocessing unit is used to perform noise reduction, ROI cropping, and brightness / contrast adjustment on eye images, and to perform low-pass filtering, gravity culling, window slicing, and normalization on sensor data.

[0014] The feature extraction unit is used to extract KF ring image features from eye images and to extract waist movement features, foot movement features, and plantar pressure distribution features from sensor data.

[0015] The single-modal evaluation unit includes a KF loop detection model, a balance evaluation model, and a gait evaluation model, used to generate modal anomaly scores, which include KF loop anomaly scores, balance anomaly scores, and gait anomaly scores.

[0016] The multimodal fusion unit employs a decision-level fusion strategy to weight and fuse several modal anomaly scores through a logistic regression model, learns the importance of different modalities, and finally outputs a normalized comprehensive score.

[0017] The report generation unit generates a structured comprehensive evaluation report by combining the user's basic information.

[0018] The alarm unit outputs an alarm signal based on a comparison between a normalized comprehensive score and a personalized threshold; the alarm signal includes a mild anomaly alarm signal and a severe anomaly alarm signal.

[0019] The storage unit is used to persistently store user basic data, modal anomaly scores, comprehensive anomaly scores, and structured comprehensive evaluation reports, and supports historical data queries.

[0020] The physician module, used for report reception and clinical decision-making, includes a physician information management unit, a report display unit, a trend analysis unit, and an intervention suggestion unit.

[0021] The doctor information management unit enables doctor account registration, permission configuration, and information updates.

[0022] The report display unit retrieves and visualizes the user's structured comprehensive assessment report, modal anomaly score, and alarm signals in real time via the web interface.

[0023] The trend analysis unit displays the changing trends of the user's comprehensive anomaly score and modal anomaly score at different points in time, arranged chronologically along a time axis.

[0024] The intervention suggestion unit generates personalized rehabilitation suggestions based on the structured comprehensive assessment report, the trend of changes in the comprehensive abnormality score and the modal abnormality score, and pushes them to the user module.

[0025] Furthermore, the user module is implemented using a smart terminal device, a portable inertial sensor, and a foot pressure insole; the information processing module is implemented using a server deployed in the cloud; and the doctor module is implemented using a web management backend.

[0026] Furthermore, the interactive acquisition and control unit is implemented in the form of a terminal application, including: acquiring eye images in the KF ring shooting task; acquiring waist acceleration and waist angular velocity in the balance posture test task; and acquiring foot acceleration, foot angular velocity, and plantar pressure distribution time-series data in the gait test task.

[0027] Furthermore, the single-modal evaluation unit includes:

[0028] Extract KF ring image features; the image features include ring width features and gray-level distribution features;

[0029] The KF ring image features are input into the KF ring detection model to obtain the KF ring anomaly score; the KF ring detection model is constructed based on the YOLOv8 model;

[0030] Extract the time-domain and frequency-domain features corresponding to waist acceleration and waist angular velocity; the time-domain features include mean, variance, and skewness; the frequency-domain features include power spectrum and peak frequency;

[0031] The time-domain and frequency-domain features are input into the balance assessment model to obtain the balance anomaly score;

[0032] Extract foot motion features and plantar pressure distribution features corresponding to time-series data of foot acceleration, foot angular velocity, and plantar pressure distribution; the foot motion features and plantar pressure distribution features include gait deviation, pressure center trajectory, and inertial features;

[0033] Foot movement characteristics and plantar pressure distribution characteristics are input into the gait assessment model to obtain a gait abnormality score.

[0034] Furthermore, the multimodal fusion unit includes a fusion input subunit, a modal influence factor calculation subunit, an initial weight acquisition subunit, a decision weight generation subunit, a real-time fusion subunit, a normalization subunit, a missing compensation subunit, and an online update subunit;

[0035] The fusion input subunit is used to receive the modal anomaly scores and basic user information output from the single-modal evaluation unit.

[0036] The modal influence factor calculation subunit: obtains the most recent N modal anomaly scores and calculates the standard deviation (MBC) of the i-th modal anomaly score. i With instantaneous rate of change MSL i The modal influence factor was obtained through the nonlinear function MYF. Where β>0 is the sensitivity adjustment parameter, and ε is a small constant to prevent division by zero;

[0037] The initial weight acquisition subunit: acquires the user's disease category BL, and retrieves the initial weight vector CQ = [CQ1, CQ2, CQ3] from the weight analysis table set by the expert based on the user's disease category BL and disease stage PS. This table is set by the clinical expert based on the disease classification and severity.

[0038] The decision weight generation subunit constructs a decision weight function JQF(BS, CQ, MYY) based on the nonlinear relationship between the disease duration BS, initial weight CQ, modal influence factor MYY, and multimodal decision weights. The decision weight function satisfies... The decision weight W is calculated. i =JQF i (BS,CQ i ,MYY i), where i and j represent the numbers corresponding to the modal anomaly scores, and τ and μ represent the initial weight decay time constant and modal factor enhancement coefficient, respectively; both τ and μ are greater than 0;

[0039] The real-time fusion subunit calculates the original comprehensive anomaly score by weighting the KF loop anomaly score S1, balance anomaly score S2, and gait anomaly score S3 according to decision weights and then applying an activation function. in, Here, b is the Sigmoid function, and b is the bias constant.

[0040] The normalization subunit: linearly maps S to a preset clinical score interval [S] according to the Min-Max algorithm. min ,S max The final normalized comprehensive score was obtained. Where σ min and σ max These are the minimum and maximum values ​​of the Sigmoid output, respectively.

[0041] The missing compensation subunit: when there is a score S among KF ring abnormality score S1, balance abnormality score S2, and gait abnormality score S3. j If missing, then the remaining weights {W} are... i} i≠j Normalization readjustment The original comprehensive anomaly score is then calculated using the readjusted decision weights; where k represents the number of the anomaly score in the KF loop anomaly score S1, balance anomaly score S2, and gait anomaly score S3 after the score is missing; and i and j represent the numbers of the modal anomaly scores.

[0042] The online update subunit uses an online gradient descent algorithm to fine-tune the parameters {τ,μ,b} in the decision weight function.

[0043] Furthermore, the report generation unit includes: a report formatting subunit, used to organize the normalized comprehensive score, several modal anomaly scores, and user basic information into a structured text report according to a preset template; a visualization subunit, used to generate line charts from the normalized comprehensive score and several modal anomaly scores, generate pie charts from the decision weight coefficients corresponding to the several modal anomaly scores, and generate plantar pressure heatmaps from the time-series data of plantar pressure distribution; an export subunit, used to generate several export formats from the structured text report, line charts, pie charts, and plantar pressure heatmaps; the export formats include PDF, HTML, and CSV formats; a push subunit, used to distribute the structured text report to the doctor module and user module in real time via a RESTful API, the distribution including email push and message center push; and a personalization subunit, used to generate personalized report data from the structured text report, line charts, pie charts, and plantar pressure heatmaps according to roles; the personalized report data includes report content and layout templates; the roles include users and doctors.

[0044] Furthermore, the information processing module also includes an alarm unit, which outputs an alarm signal based on a comparison between a normalized comprehensive score and a personalized threshold; the alarm signal includes a mild abnormality alarm signal and a severe abnormality alarm signal; and is sent to users and doctors in real time through multiple channels.

[0045] The alarm unit includes: a personnel data acquisition subunit, which acquires the user's age (RN), disease category (BL), treatment duration (ZS), and disease stage (PS) in real time; and an influence factor calculation subunit, which constructs an age influence function NYF(RN) based on the user's age, satisfying the formula... Substituting the individuals' ages into the age influence function yields the age influence factor NYY, where a and b are influence factors, and NY... a and NY b This is represented by an age threshold; a disease impact function BYF(BL) is constructed based on the disease category, satisfying the formula... Substituting the disease category into the disease influence function yields the disease influence factor BYY, where c and d are influence factors. The personalized threshold calculation subunit constructs a threshold function T = YF(NYY, BYY, ZS) based on the nonlinear relationship between NYY, BYY, ZS, and the personalized threshold, satisfying the formula... The personalized threshold is calculated by substituting NYY, BYY, and ZS into the threshold function, where BY represents the standard threshold and DS represents the unit duration. In the first-level alarm judgment subunit, if S'≤T, no alarm is generated; if S'>T, the second-level alarm judgment is initiated. In the second-level alarm judgment subunit, a severe threshold T is set. sev = D×T, where the proportionality coefficient D>1, if T<S′≤T sevA mild anomaly alarm signal is generated; if S′>T sev The system generates severe abnormality alarm signals; the alarm message generation subunit constructs alarm messages, including basic user information, normalized comprehensive score and personalized threshold, severe threshold, alarm signal and personalized rehabilitation suggestions; the multi-channel push subunit distributes alarm messages in real time through the following methods: terminal App push notification, doctor's web message center highlight display, email and SMS alarms.

[0046] Another aspect of the present invention provides a comprehensive monitoring and assessment method for typical manifestations of Wilson's disease, comprising the following steps:

[0047] Obtain basic user information; the basic user information includes name, gender, age, duration of illness, and stage of illness;

[0048] Acquire sensor data; the sensor data includes eye images, waist acceleration, waist angular velocity, foot acceleration, foot angular velocity, and time-series data of plantar pressure distribution;

[0049] The sensor data is preprocessed to obtain preprocessed sensor data; the preprocessing operations include noise reduction, ROI cropping, low-pass filtering, gravity culling, sliding window slicing, and normalization.

[0050] Based on the preprocessed sensor data, several modal anomaly scores are generated through the KF loop detection model, balance evaluation model, and gait evaluation model; the modal anomaly scores include KF loop anomaly scores, balance anomaly scores, and gait anomaly scores;

[0051] The original comprehensive anomaly score is calculated by weighted summation of several modal anomaly scores, and the normalized comprehensive score is output by Min-Max linear mapping.

[0052] A structured comprehensive evaluation report and alarm signals are generated based on the normalized comprehensive score and basic user information.

[0053] Compared with the prior art, the beneficial effects of this application are:

[0054] 1. By simultaneously collecting eye images, waist data, foot data, and plantar pressure distribution data, and employing a decision-level fusion strategy to dynamically adjust the decision weight coefficients of each modality abnormality score, the original comprehensive abnormality score is obtained by combining the abnormality scores of each modality. This significantly improves the sensitivity and specificity for motor and functional abnormalities in patients with Wilson's disease, and overcomes the false positive or false negative problems that are easily generated by single modality detection.

[0055] 2. By constructing a nonlinear threshold model based on age and disease course influence functions, a personalized threshold is dynamically generated for each user, and mild and severe abnormality alarm signals are pushed accordingly, which significantly improves the accuracy and timeliness of early warning and helps with early intervention.

[0056] 3. The user end integrates smart terminal device applications, portable inertial sensors, and foot pressure insoles to achieve local data collection, encryption, and real-time uploading throughout the process; after centralized processing and analysis in the cloud, assessment reports and alarm signals are seamlessly transmitted back via Web / App, breaking through geographical limitations and greatly expanding the coverage of high-quality neurological disease monitoring resources.

[0057] 4. The report generation unit supports line charts, pie charts, and heatmaps, providing multi-dimensional visualization of structured comprehensive assessment reports and modal anomaly scores; it also provides PDF / HTML format export and API interfaces to meet the needs of various scenarios such as clinical consultation, case archiving, and scientific research analysis.

[0058] 5. By leveraging the real-time progress visualization, phased feedback prompts, and personalized strategy push mechanisms provided by the terminal application, the smoothness of user operation and task completion rate can be improved; the doctor's end can realize two-way information interaction through trend data analysis and remote collaboration functions, supporting individualized decision-making and long-term monitoring and management. Attached Figure Description

[0059] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0060] Figure 1 This is a schematic diagram illustrating the principle of a comprehensive monitoring and assessment system for typical manifestations of Wilson's disease according to the present invention.

[0061] Figure 2 This is a schematic diagram of the overall architecture of a comprehensive monitoring and evaluation system for typical manifestations of Wilson's disease according to the present invention;

[0062] Figure 3 This is a flowchart of the alarm signal generation process of the present invention;

[0063] Figure 4 This is a flowchart of a comprehensive monitoring and evaluation method for typical manifestations of Wilson's disease according to the present invention. Detailed Implementation

[0064] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0065] Please see Figure 1 The first aspect of this application provides a comprehensive monitoring and assessment system for typical manifestations of Wilson's disease, including a user module (101), an information processing module (102), and a doctor module (103).

[0066] The user module (101) is used for data acquisition and interaction, including a user information management unit, an interactive acquisition control unit, a data integration and transmission unit, and a local display unit. The user information management unit realizes user identity registration and information updates based on the application on the smart terminal device, and uploads basic user information such as name, gender, age, duration of illness, and stage of illness. The interactive acquisition control unit is used to synchronously acquire sensor data, including: acquiring eye images in the KF ring shooting task; acquiring time-series data such as waist acceleration and waist angular velocity from the waist inertial sensor in the balance posture test task; and acquiring time-series data of foot acceleration, foot angular velocity, and foot pressure distribution from the foot inertial sensor and the foot pressure insole in the gait test task. The data integration and transmission unit encrypts the acquired basic user information and sensor data and uploads them to the cloud in real time via BLE, Wi-Fi, or cellular network. The local display unit displays the test progress, acquisition status, and brief feedback in real time on the device application.

[0067] The information processing module (102), located on the cloud server, is used for data processing and evaluation. It includes a data preprocessing unit, a feature extraction unit, a single-modal evaluation unit, a multimodal fusion unit, a report generation unit, an alarm unit, and a storage unit. The data preprocessing unit performs denoising, ROI cropping, and brightness / contrast adjustment on eye images, low-pass filtering, gravity removal, window slicing, and normalization on sensor data, and filtering and normalization on plantar pressure data. The feature extraction unit extracts KF ring image features, waist movement features, foot movement features, and plantar pressure distribution features from sensor data. The single-modal evaluation unit calls the KF ring detection model, balance evaluation model, and gait evaluation model to generate anomalies in each modality. The scoring system comprises KF loop anomaly score, balance anomaly score, and gait anomaly score. The multimodal fusion unit employs a decision-level fusion strategy to weight and fuse the anomaly scores of each modality through a logistic regression model, learning the importance of different modalities and ultimately outputting a normalized comprehensive score. The report generation unit combines user basic information, including age, disease duration, and disease stage, to generate a formatted structured comprehensive assessment report. The alarm unit compares the normalized comprehensive score with personalized thresholds and outputs alarm signals, including mild and severe anomaly alarm signals. The storage unit persistently stores user basic data, modal anomaly scores, comprehensive anomaly scores, and structured comprehensive assessment reports, and supports historical data queries.

[0068] The physician module (103) is used for report reception and clinical decision-making, including a physician information management unit, a report display unit, a trend analysis unit, and an intervention suggestion unit. The physician information management unit realizes physician account registration, permission configuration, and information updates. The report display unit pulls and visualizes the user's comprehensive assessment report, modal abnormality scores, and alarm signals in real time through the web. The trend analysis unit is used to display the changing trends of the user's comprehensive abnormality score and modal abnormality score at different time points in chronological order. The intervention suggestion unit is used to generate personalized rehabilitation suggestions based on the changing trends of the structured comprehensive assessment report, comprehensive abnormality score, and modal abnormality score, and push them to the user module.

[0069] The user module (101) in this embodiment is implemented by a smart terminal device (1011), a portable inertial sensor (1012), and a foot pressure insole (1013); the information processing module (102) in this embodiment is implemented by a server deployed in the cloud; and the doctor module (103) in this embodiment is implemented by a Web management backend.

[0070] In this embodiment, the interactive acquisition and control unit is implemented in the form of a terminal application, including: acquiring eye images in the KF ring shooting task; acquiring waist acceleration and waist angular velocity in the balance posture test task; and acquiring foot acceleration, foot angular velocity, and plantar pressure distribution time-series data in the gait test task.

[0071] In this embodiment, the multimodal fusion unit is used to convert the modal anomaly scores output by the single-modal assessment unit, including the KF loop anomaly score S1, balance anomaly score S2, and gait anomaly score S3, along with the user's basic information such as age RN, disease category BL, disease duration BS, and disease stage PS, into a final normalized comprehensive score, and to adaptively update the decision weight coefficients corresponding to several modal anomaly scores. The multimodal fusion unit is subdivided into a fusion input subunit, a modal influence factor calculation subunit, an initial weight acquisition subunit, a decision weight generation subunit, a real-time fusion subunit, a normalization subunit, a missing compensation subunit, and an online update subunit. Its specific structure and functions are as follows: The fusion input subunit is used to receive the modal anomaly scores output by the single-modal assessment unit and the user's basic information; the modal influence factor calculation subunit obtains the most recent N modal anomaly scores and calculates the standard deviation (MBC) of the i-th modal anomaly score. i With instantaneous rate of change MSL i The modal influence factor was obtained through the nonlinear function MYF. Where β>0 is the sensitivity adjustment parameter, and in this embodiment, β=1.2 is taken, and ε is a small constant to prevent division by zero; Initial weight acquisition subunit: acquire the user's disease category BL, and according to the user's disease category BL and disease stage PS, retrieve the initial weight vector CQ=[CQ1,CQ2,CQ3] from the weight analysis table set by the expert. This table is set by clinical experts based on disease classification and severity; Decision weight generation subunit: construct the decision weight function JQF(BS,CQ,MYY) based on the nonlinear relationship between disease duration BS, initial weight CQ, modal influence factor MYY and multimodal decision weights. The decision weight function satisfies The decision weight W is calculated. i =JQF i (BS,CQ i ,MYY i ), where i and j represent the numbers corresponding to the modal anomaly scores, and τ and μ represent the initial weight decay time constant and modal factor enhancement coefficient, respectively. In this embodiment, τ and μ are set to 12 and 0.1, respectively. This is represented as an initial weight decay term, respecting expert experience but avoiding long-term fixation. For example, the importance of the KF loop decreases in the later stages of treatment. i×tanh(μ×BS) represents the dynamic sensitivity enhancement term. The longer the disease course, the higher the weight of real-time indicator changes, making it more sensitive to sudden deterioration in long-term patients. The real-time fusion subunit calculates the weighted sum of the KF ring abnormality score S1, balance abnormality score S2, and gait abnormality score S3 according to decision weights and then passes it through an activation function to obtain the original comprehensive abnormality score. in The function is the Sigmoid function, and b is the bias constant; the normalized subunit maps S linearly to the preset clinical score interval [S] according to the Min-Max method. min ,S max The final normalized comprehensive score was obtained. Where σ min and σ max These are the minimum and maximum values ​​of the Sigmoid output, respectively; a missing compensation subunit occurs when any of the following scores exist: KF loop anomaly score S1, balance anomaly score S2, and gait anomaly score S3. j If missing, then the remaining weights {W} are... i} i≠j Normalization readjustment The original comprehensive anomaly score is then calculated using the readjusted decision weights; where k represents the number of the anomaly score in the KF ring anomaly score S1, balance anomaly score S2, and gait anomaly score S3 after score loss; i and j represent the numbers of the modal anomaly scores; the online update subunit, after receiving expert diagnosis or real clinical labels, uses an online gradient descent algorithm to fine-tune the parameters {τ,μ,b} in the decision weight function to optimize subsequent fusion performance.

[0072] The report generation unit in this embodiment includes: a report formatting subunit, used to organize the normalized comprehensive score, several modal abnormality scores, and user basic information into a structured text report according to a preset template; a visualization subunit, used to generate line charts of the normalized comprehensive score and several modal abnormality scores, generate pie charts of the decision weight coefficients corresponding to the several modal abnormality scores, and generate plantar pressure heatmaps from the plantar pressure distribution time-series data; an export subunit, used to package and export the structured report, line charts, pie charts, and plantar pressure heatmaps into multiple formats such as PDF, HTML, or CSV; a push subunit, used to send the structured text report to the doctor module and user module in real time via RESTful API, including email push and message center push; and a personalization subunit, used to dynamically generate personalized report data based on different roles, including doctors and users, or clinical scenarios. Personalized report data includes report content and layout templates, etc.

[0073] Please see Figure 3In this embodiment, the alarm unit is used to generate alarm signals based on the normalized comprehensive score and personalized thresholds, and to send them to users and doctors in real time through multiple channels. The alarm unit includes: a personnel data acquisition subunit, which periodically or in real time acquires basic user information such as the user's age (RN), disease category (BL), treatment duration (ZS), and disease stage (PS); and an influence factor calculation subunit, which constructs an age influence function NYF(RN) based on the user's age, satisfying the formula... Substituting the individuals' ages into the age influence function yields the age influence factor NYY, where a and b are influence factors. In this embodiment, a is set to 0.15 and b to 0.05. a and NY b Represented as an age threshold, in this embodiment, NY is... a and NY b Ages are set to 12 and 18 years respectively; a disease impact function BYF(BL) is constructed based on the disease category, satisfying the formula... Substituting the disease category into the disease influence function yields the disease influence factor BYY, where c and d are influence factors. In this embodiment, c is set to 0.25 and b to 0.1. The personalized threshold calculation subunit constructs a threshold function T = YF(NYY, BYY, ZS) based on the nonlinear relationship between NYY, BYY, ZS, and the personalized threshold, satisfying the formula... The personalized threshold is calculated by substituting NYY, BYY, and ZS into the threshold function, where BY represents the standard threshold and DS represents the unit duration. In this embodiment, DS is set to 1 month. In the first-level alarm judgment subunit, if S'≤T, no alarm is generated; if S'>T, the second-level alarm judgment is initiated. In the second-level alarm judgment subunit, a severe threshold T is set. sev = D × T, where the proportionality coefficient D > 1; in this embodiment, D = 1.3, if T < S′ ≤ T sev A mild anomaly alarm signal is generated; if S′>T sev The system generates a severe abnormality alarm signal; the alarm message generation subunit constructs an alarm message, which includes basic user information, normalized comprehensive score and personalized threshold, severe threshold, alarm signal and personalized rehabilitation suggestions. In this embodiment, the personalized rehabilitation suggestions include retesting, medical treatment or adjustment of treatment plan, etc.; the multi-channel push subunit sends the alarm message in real time through the following methods: terminal App push notification, doctor's web message center highlight display, email and SMS alarm.

[0074] Please see Figure 4 Another embodiment of this application provides a comprehensive monitoring and assessment method for typical manifestations of Wilson's disease, including the following steps:

[0075] Obtain basic user information; the basic user information includes name, gender, age, duration of illness, and stage of illness;

[0076] Acquire sensor data; the sensor data includes eye images, waist acceleration, waist angular velocity, foot acceleration, foot angular velocity, and time-series data of plantar pressure distribution;

[0077] The sensor data is preprocessed to obtain preprocessed sensor data; the preprocessing operations include noise reduction, ROI cropping, low-pass filtering, gravity culling, sliding window slicing, and normalization.

[0078] Based on the preprocessed sensor data, several modal anomaly scores are generated through the KF loop detection model, balance evaluation model, and gait evaluation model; the modal anomaly scores include KF loop anomaly scores, balance anomaly scores, and gait anomaly scores;

[0079] The original comprehensive anomaly score is calculated by weighted summation of several modal anomaly scores, and the normalized comprehensive score is output by Min-Max linear mapping.

[0080] A structured comprehensive evaluation report and alarm signals are generated based on the normalized comprehensive score and basic user information.

[0081] Some of the data in the above formula are calculated by removing dimensions and taking their numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0082] The working principle of this application is as follows: Sensor data is acquired; the sensor data is processed by a single-modal evaluation unit to obtain several modal anomaly scores, which are then weighted and summed to calculate the original comprehensive anomaly score, thereby determining the normalized comprehensive score. This leads to the generation of a structured comprehensive evaluation report and alarm signals. Multi-source data analysis is used to identify multi-modal anomaly results affecting the patient's condition, and the influence of these results on the disease analysis is dynamically adjusted. This ensures that the disease analysis closely matches the patient's actual situation, making the comprehensive monitoring results more accurate. This improves the accuracy of the comprehensive monitoring and evaluation method for Wilson's disease patients, avoiding the problem in existing technologies that often rely on single images or motion detection and neglect multi-source data analysis, resulting in low accuracy and efficiency in the comprehensive monitoring and evaluation methods for Wilson's disease patients.

[0083] The above embodiments are only used to illustrate the technical methods of this application and are not intended to limit it. Although this application has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of this application without departing from the spirit and scope of the technical methods of this application.

Claims

1. A comprehensive monitoring and assessment system for typical manifestations of Wilson's disease, characterized in that, It includes a user module, an information processing module, and a doctor module; The user module includes a user information management unit, an interactive acquisition and control unit, and a data integration and transmission unit. The user information management unit: enables user identity registration and information updates, and uploads basic user information, including name, gender, age, duration of illness, and stage of illness; The interactive acquisition and control unit is used to acquire sensor data, including eye images, waist acceleration, waist angular velocity, foot acceleration, foot angular velocity, and time-series data of plantar pressure distribution. The data integration and transmission unit encrypts the collected user basic information and sensor data and uploads them to the information processing module in real time via data transmission. The information processing module includes a single-modal evaluation unit, a multimodal fusion unit, and a report generation unit; The single-modal evaluation unit includes a KF loop detection model, a balance evaluation model, and a gait evaluation model, which are used to generate modal anomaly scores, including KF loop anomaly scores, balance anomaly scores, and gait anomaly scores. The multimodal fusion unit employs a decision-level fusion strategy to weight and fuse several modal anomaly scores through a logistic regression model, ultimately outputting a normalized comprehensive score. The report generation unit generates a structured comprehensive evaluation report by combining the user's basic information. The doctor module includes a doctor information management unit, a report display unit, a trend analysis unit, and an intervention suggestion unit. The doctor information management unit enables doctor account registration, permission configuration, and information updates. The report display unit retrieves and visualizes the user's structured comprehensive assessment report, modal anomaly score, and alarm signals in real time via the web interface; The trend analysis unit displays the changing trends of the user's comprehensive anomaly score and modal anomaly score at different points in time, arranged chronologically along a time axis. The intervention suggestion unit generates personalized rehabilitation suggestions based on the structured comprehensive assessment report, the trend of changes in the comprehensive abnormality score and the modal abnormality score, and pushes them to the user module.

2. The comprehensive monitoring and evaluation system for typical manifestations of Wilson's disease according to claim 1, characterized in that, The user module is implemented using a smart terminal device, a portable inertial sensor, and a foot pressure insole; the information processing module is implemented using a server deployed in the cloud; and the doctor module is implemented using a web management backend.

3. The comprehensive monitoring and evaluation system for typical manifestations of Wilson's disease according to claim 1, characterized in that, The interactive acquisition and control unit is implemented in the form of a terminal application, including: acquiring eye images in the KF ring shooting task; acquiring waist acceleration and waist angular velocity in the balance posture test task; and acquiring time-series data of foot acceleration, foot angular velocity, and plantar pressure distribution in the gait test task.

4. The comprehensive monitoring and evaluation system for typical manifestations of Wilson's disease according to claim 1, characterized in that, The single-modal evaluation unit includes: Extract KF ring image features; the image features include ring width features and gray-level distribution features; The KF ring image features are input into the KF ring detection model to obtain the KF ring anomaly score; the KF ring detection model is constructed based on the YOLOv8 model; Extract the time-domain and frequency-domain features corresponding to waist acceleration and waist angular velocity; the time-domain features include mean, variance, and skewness; the frequency-domain features include power spectrum and peak frequency; The time-domain and frequency-domain features are input into the balance assessment model to obtain the balance anomaly score; Extract foot motion features and plantar pressure distribution features corresponding to time-series data of foot acceleration, foot angular velocity, and plantar pressure distribution; the foot motion features and plantar pressure distribution features include gait deviation, pressure center trajectory, and inertial features; Foot movement characteristics and plantar pressure distribution characteristics are input into the gait assessment model to obtain a gait abnormality score.

5. A comprehensive monitoring and evaluation system for typical manifestations of Wilson's disease according to claim 1, characterized in that, The multimodal fusion unit includes a fusion input subunit, a modal influence factor calculation subunit, an initial weight acquisition subunit, a decision weight generation subunit, a real-time fusion subunit, and a normalization subunit; The fusion input subunit is used to receive the modal anomaly scores and basic user information output from the single-modal evaluation unit. The modal influence factor calculation subunit: obtains the most recent N modal anomaly scores and calculates the standard deviation (MBC) of the i-th modal anomaly score. i With instantaneous rate of change MSL i The modal influence factor was obtained through the nonlinear function MYF. Where β represents the sensitivity adjustment parameter, β>0; ε is a small constant to prevent division by zero, ε>0; N is an integer, N>0; The initial weight acquisition subunit: acquires the user's disease category BL; based on the user's disease category BL and disease stage PS, retrieves the initial weight vector CQ = [CQ1, CQ2, CQ3] from the weight analysis table set by the expert. The weight analysis table is set by the clinical expert based on disease classification and severity. The decision weight generation subunit constructs a decision weight function JQF(BS, CQ, MYY) based on the nonlinear relationship between the disease duration BS, initial weight CQ, modal influence factor MYY, and multimodal decision weights. The decision weight function satisfies... Where i and j represent the numbers corresponding to the modal abnormality scores, τ and μ represent the initial weight decay time constant and modal factor enhancement coefficient, respectively, and both τ and μ are greater than 0; the decision weight W is calculated by substituting the disease duration, initial weight, and modal influence factor into the decision weight function. i ; The real-time fusion subunit calculates the original comprehensive anomaly score by weighting the KF loop anomaly score S1, balance anomaly score S2, and gait anomaly score S3 according to decision weights and then applying an activation function. in, Here, b is the Sigmoid function, and b is the bias constant. The normalization subunit: linearly maps S to a preset clinical score interval [S] according to the Min-Max algorithm. min ,S max The final normalized comprehensive score was obtained. Where σ min and σ max These are the minimum and maximum values ​​output by the Sigmoid function, respectively.

6. The comprehensive monitoring and evaluation system for typical manifestations of Wilson's disease according to claim 1, characterized in that, The report generation unit includes a report formatting subunit, a visualization subunit, an export subunit, a push subunit, and a personalization subunit; The report formatting subunit is used to organize the normalized comprehensive score, several modal anomaly scores, and user basic information into a structured text report according to a preset template. The visualization sub-unit is used to generate a line chart from the normalized comprehensive score and several modal anomaly scores, generate a pie chart from the decision weight coefficients corresponding to several modal anomaly scores, and generate a plantar pressure heat map from the time series data of plantar pressure distribution. The export subunit is used to generate several export formats from structured text reports, line charts, pie charts, and plantar pressure heatmaps. The push sub-unit is used to send structured text reports to the doctor module and user module in real time via a RESTful API. The sending includes email push and message center push. The personalized customization sub-unit is used to generate personalized report data based on roles for structured text reports, line charts, pie charts, and plantar pressure heatmaps; the personalized report data includes report content and layout templates; the roles include users and doctors.

7. A comprehensive monitoring and evaluation system for typical manifestations of Wilson's disease according to claim 1, characterized in that, The information processing module also includes an alarm unit; The alarm unit outputs an alarm signal based on a comparison between a normalized comprehensive score and a personalized threshold; the alarm signal includes a mild anomaly alarm signal and a severe anomaly alarm signal. The alarm unit includes: a personnel data acquisition subunit, an impact factor calculation subunit, a personalized threshold calculation subunit, a first-level alarm judgment subunit, a second-level alarm judgment subunit, an alarm message generation subunit, and a multi-channel push subunit; The personnel data acquisition subunit: acquires personnel age (RN), disease category (BL), treatment duration (ZS), and disease stage (PS) in real time; The impact factor calculation subunit constructs an age impact function NYF(RN) based on the age of the personnel. The age impact function satisfies the formula... Where a and b are the impact factors, NY a and NY b This is represented as an age threshold; the age influence factor NYY is calculated by substituting the individuals' ages into the age influence function. Based on the disease category, a disease impact function BYF(BL) is constructed, which satisfies the formula... Where c and d are influencing factors; the disease impact factor BYY is obtained by substituting the disease category into the disease impact function; The personalized threshold calculation subunit constructs a threshold function T = YF(NYY, BYY, ZS) based on the nonlinear relationships between NYY, BYY, and ZS and the personalized threshold, and the threshold function satisfies the formula... Where BY represents the standard threshold and DS represents the unit duration; the personalized threshold is obtained by substituting NYY, BYY, and ZS into the threshold function. The first-level alarm judgment subunit: if S′≤T, then no alarm is generated; if S′>T, then proceed to the second-level alarm judgment subunit. The secondary alarm judgment subunit: sets the severity threshold T. sev = D×T, where the proportionality coefficient D>1, if T<S′≤T sev A mild anomaly alarm signal is generated; if S′>T sev This generates a severe anomaly alarm signal; The alarm message generation subunit: constructs an alarm message, which includes basic user information, normalized comprehensive score and personalized threshold, severe threshold, alarm signal and personalized rehabilitation suggestions; The multi-channel push subunit will send alarm messages in real time through the following methods: push notifications from the terminal app, highlighted display in the doctor's web message center, and alarms via email and SMS.

8. A comprehensive monitoring and assessment system for typical manifestations of Wilson's disease according to claim 1, characterized in that, The information processing module further includes a data preprocessing unit, a feature extraction unit, and a storage unit; The data preprocessing unit is used to perform noise reduction, ROI cropping, and brightness / contrast adjustment on eye images, and to perform low-pass filtering, gravity culling, window slicing, and normalization on sensor data. The feature extraction unit is used to extract KF ring image features from eye images and to extract waist movement features, foot movement features, and plantar pressure distribution features from sensor data. The storage unit is used to persistently store user basic data, modal anomaly scores, comprehensive anomaly scores, and structured comprehensive evaluation reports, and supports historical data queries.

9. A comprehensive monitoring and evaluation system for typical manifestations of Wilson's disease according to claim 5, characterized in that, The multimodal fusion unit further includes a missing compensation subunit and an online update subunit; The missing compensation subunit: when there is a score S among KF ring abnormality score S1, balance abnormality score S2, and gait abnormality score S3. j If missing, then the remaining weights {W} are... i } i≠j Normalization readjustment The original comprehensive anomaly score is then calculated using the readjusted decision weights; where k represents the number of the anomaly score in the KF loop anomaly score S1, balance anomaly score S2, and gait anomaly score S3 after the score is missing; and i and j represent the numbers of the modal anomaly scores. The online update subunit uses an online gradient descent algorithm to fine-tune the parameters {τ,μ,b} in the decision weight function.

10. A comprehensive monitoring and assessment method for typical manifestations of Wilson's disease, applied to the comprehensive monitoring and assessment system for typical manifestations of Wilson's disease as described in any one of claims 1-9, characterized in that, include: Obtain basic user information; the basic user information includes name, gender, age, duration of illness, and stage of illness; Acquire sensor data; the sensor data includes eye images, waist acceleration, waist angular velocity, foot acceleration, foot angular velocity, and time-series data of plantar pressure distribution; The sensor data is preprocessed to obtain preprocessed sensor data; the preprocessing operations include noise reduction, ROI cropping, low-pass filtering, gravity culling, sliding window slicing, and normalization. Based on the preprocessed sensor data, several modal anomaly scores are generated through the KF loop detection model, balance evaluation model, and gait evaluation model; the modal anomaly scores include KF loop anomaly scores, balance anomaly scores, and gait anomaly scores; The original comprehensive anomaly score is calculated by weighted summation of several modal anomaly scores, and the normalized comprehensive score is output by Min-Max linear mapping. A structured comprehensive evaluation report and alarm signals are generated based on the normalized comprehensive score and basic user information.

Images