Intelligent lupus condition monitoring system based on data fusion and application of intelligent lupus condition monitoring system in watch

Abstract

The invention discloses a lupus condition intelligent monitoring system based on data fusion and an application in a watch, and relates to the field of lupus condition monitoring, and the system comprises a data collection module, a preprocessing module, a fusion module, a condition assessment module, an early warning module, a storage and management module, a user interaction module, a model updating module, a personalized analysis module and a remote medical module. The method comprises the following steps: firstly, collecting data from a smart watch and medical equipment, denoising, normalizing and extracting features; fusing the data by using a deep belief network to obtain a comprehensive health index; evaluating an illness state by using an LSTM and an attention mechanism model; predicting and setting a threshold value for early warning according to an evaluation result and a trend; the model is updated by means of federal learning and transfer learning, lupus conditions are evaluated based on multiple data, and doctor-patient remote interaction is realized by adopting a 5G communication technology. According to the invention, illness conditions are monitored at any time, early warning is carried out in time when abnormity occurs, illness condition assessment is accurate, medical resource allocation is optimized, remote medical development is promoted, and lupus disease research is assisted.

Description

Technical Field

[0001] The present invention relates to the technical field of lupus condition monitoring, and in particular to a lupus condition intelligent monitoring system based on data fusion and its application in a watch. Background Art

[0002] Systemic lupus erythematosus (SLE) is a complex autoimmune disease characterized by recurring symptoms and significant individual variability. Accurate monitoring of lupus disease progression is crucial for developing appropriate treatment plans and improving patient outcomes. However, current monitoring of lupus disease faces numerous challenges.

[0003] In terms of data collection, traditional monitoring relies primarily on regular hospital checkups, which collect limited data types and at long intervals. While blood and urine tests can capture some key indicators, they cannot reflect real-time changes in the condition. Data on the patient's daily physiological state and living environment, which have a significant impact on their condition, is difficult to effectively collect. While wearable devices are becoming increasingly popular, they lack functionality for lupus monitoring and are unable to accurately collect physiological parameters closely related to lupus. Data from different devices is also poorly compatible, making integration and utilization difficult.

[0004] The assessment process is also plagued by numerous problems. Doctors often make subjective judgments based on experience and limited examination results, lacking objective, quantitative assessment criteria. The existing assessment indicator system is incomplete and fails to fully encompass the factors influencing lupus disease progression. Furthermore, the lack of long-term, continuous data makes it difficult to accurately assess disease trends, which can easily lead to misdiagnosis or delayed treatment adjustments, compromising treatment outcomes and quality of life for patients.

[0005] With the advancement of science and technology, technologies such as data fusion and artificial intelligence offer potential solutions to these problems. However, research on applying these technologies to lupus monitoring is still in its exploratory stages, lacking mature systems and solutions. There is an urgent need for an intelligent system and wristwatch-based application that can comprehensively, accurately monitor lupus conditions in real time to meet clinical needs and improve the quality of medical care for lupus patients. Summary of the Invention

[0006] The present invention proposes an intelligent lupus disease monitoring system based on data fusion to solve the problems mentioned in the above-mentioned prior art.

[0007] To achieve the above objectives, the present invention adopts the following technical solution: a lupus disease intelligent monitoring system based on data fusion, comprising:

[0008] Multi-source data acquisition module: This module collects physiological data using smartwatches, connects to external devices via Bluetooth or Wi-Fi to obtain key indicator data, and obtains symptom self-assessment data from patients' active input and electronic medical record systems simultaneously, using new micro-electromechanical technology;

[0009] Data preprocessing module: uses adaptive filtering algorithm to remove noise, adopts improved Z-score formula to normalize data of different dimensions, and uses convolutional neural network (CNN) feature extraction model to automatically extract physiological signal features;

[0010] Data fusion module: Build a data fusion model based on deep belief network (DBN), take preprocessed data as input, fuse the feature vector FV through RBM feature learning, calculate the comprehensive health index HI using the formula, and optimize the weight by back propagation algorithm;

[0011] Condition Assessment Module: This module uses a learning model based on a long short-term memory (LSTM) network combined with an attention mechanism to learn long-term dependencies between conditions, focusing on key data features, outputting a probability distribution of conditions and classifying it into three levels. It uses a cross-entropy loss function for training based on labeled data.

[0012] Early warning module: Dynamic early warning thresholds are set based on the probability distribution of disease assessment, data change trends, and individual characteristics. A gray prediction model is used to predict disease conditions. When the threshold is reached, early warnings are issued in various ways, along with disease conditions and treatment recommendations.

[0013] Data storage and management module: Build a distributed database using distributed hash table (DHT) technology, adopt homomorphic encryption algorithm for storage, provide a blockchain-based access control mechanism, and perform data access and operations based on the permission information recorded on the blockchain;

[0014] User interaction module: Develop a unified interface for smartwatches, mobile apps, and desktops, using adaptive layout technology to automatically adjust display content based on device screen size and resolution, and introduce emotional interaction design;

[0015] Model update module: Adopting a federated learning framework, the disease assessment model is trained with local data while protecting privacy. Model parameters are exchanged and updated in an aggregated manner through encryption technology, and the model is fine-tuned in combination with a transfer learning algorithm.

[0016] Furthermore, it also includes:

[0017] Personalized analysis module: Based on the patient's genetic data GD, search for gene markers related to lupus through the gene-disease association database and construct the genetic risk score GRS formula w gi is the weight of gene marker i, g i is the genotype score of gene marker i; combined with the lifestyle data of diet frequency DF and exercise intensity EI, the hierarchical analysis method AHP is used to determine the weight of personalized evaluation indicators and revise the disease assessment results.

[0018] Furthermore, it also includes:

[0019] Telemedicine module: It uses 5G communication technology combined with video stream coding and compression technology to support real-time remote consultation functions. Through the system, various monitoring data and high-definition videos of patients can be viewed in real time. Virtual labeling technology is used to mark and annotate patient body parts or test images. Remote operation assistance technology is introduced to remotely guide medical operations.

[0020] Furthermore, the multi-source data acquisition module adds environmental data acquisition function, collects ultraviolet intensity UVI, allergen concentration AC, and noise level NL data through micro environmental sensors; integrates environmental and other physiological and medical data, and uses the improved Bayesian fusion formula H indicates the condition, E i Represents various types of data;

[0021] The calculation formula of the integrated health index is: w i is the weight of each dimension of the fused feature vector, and FV is the fused feature vector.

[0022] Furthermore, the improved Z-score normalization formula in the data preprocessing module μ s and σ s are the mean and standard deviation of the data in a specific time period, k and b are coefficients adjusted according to the characteristics of the data;

[0023] The generative adversarial network (GAN) is used to repair missing data. The generator learns the real data distribution to generate missing data values. The discriminator distinguishes between generated data and real data. The wavelet transform is used to perform multi-scale analysis on the data to extract the characteristics of different frequency components.

[0024] Furthermore, the loss function in the disease assessment module adopts the cross entropy loss function y j is the true disease level label, p j is the model predicted probability;

[0025] A multimodal fusion Transformer model is used to assess the condition. Image data is used to extract features through a convolutional neural network. Text data is converted into vector representation using a word embedding model. Numerical data is normalized and then input into the Transformer model. Data from different modalities are fused and analyzed through a multi-head attention mechanism.

[0026] Furthermore, the early warning module uses the grey prediction model GM(1,1) to predict the trend of disease indicators. The formula is: x (0) is the original data sequence, x (1) Generate a cumulative sequence for it, a and b are model parameters;

[0027] An algorithm combining fuzzy logic and reinforcement learning is used to determine the warning threshold. Fuzzy logic is used to process the condition assessment results, data change trends and individual patient characteristics to construct a fuzzy rule base. A reinforcement learning algorithm is used to optimize the fuzzy rule base and warning threshold based on the actual changes in the patient's condition.

[0028] Furthermore, the data storage and management module uses a combination of blockchain and Interstellar File System (IPFS) to store data. Blockchain stores data metadata and hash values, while IPFS stores actual data files. Data visualization tools are developed to display data statistical analysis results through interactive charts.

[0029] Furthermore, the user interaction module uses brain-computer interface (BCI) technology to achieve interaction, operates the smart watch through brain thinking activities to view data or issue commands, and uses voice emotion recognition technology to analyze voice emotions.

[0030] Furthermore, the watch integrates physiological and partial environmental data collection functions, uses Bluetooth technology to transmit data with external medical detection equipment and mobile phones, and has a built-in microprocessor to pre-process some data in real time, reminding patients through vibration, sound, and light signals, providing a simple interactive interface, and viewing medical data and receiving early warning information through touch operations.

[0031] Compared with the existing technology, the beneficial effects of the present invention are:

[0032] From the patient's perspective, this greatly improves the convenience of disease monitoring. The smartwatch integrates multiple functions, allowing patients to collect data anytime, anywhere, eliminating the need for frequent trips to the hospital. The system monitors the patient's condition in real time and provides immediate warnings upon detection of abnormalities, allowing patients to take immediate action. This enhances their ability to self-manage their disease and reduces the psychological burden. For example, patients can monitor their health status in their daily lives and respond promptly to sudden discomfort.

[0033] The system provides a more accurate basis for condition assessment and treatment. The fusion of multi-source data and advanced algorithms enables comprehensive and objective assessments, eliminating the limitations of relying solely on empirical judgment. Doctors can more accurately assess disease trends, adjust treatment plans promptly, and improve treatment outcomes. For patients with complex conditions, the system's comprehensive data integration helps doctors quickly make accurate diagnoses, avoiding misdiagnoses and delayed treatment.

[0034] From the perspective of medical resource utilization, the system optimizes resource allocation. The telemedicine module enables remote communication between patients and doctors, reducing unnecessary medical visits and alleviating pressure on hospitals. Data storage and management functions facilitate the sharing of medical data, promote collaboration among medical institutions, and enhance overall medical services. Furthermore, the extensive medical data continuously collected by the system provides a rich source of material for lupus research, enabling a deeper understanding of disease mechanisms and driving the development of new diagnostic methods and therapeutic drugs. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a schematic block diagram of an intelligent lupus disease monitoring system based on data fusion proposed by the present invention;

[0036] Figure 2 This is a schematic diagram comparing the timeliness of patient condition monitoring;

[0037] Figure 3 This is a diagram comparing the accuracy of doctors' judgments on the condition;

[0038] Figure 4 Schematic diagram of the distribution of patients' scores on the convenience of seeking medical treatment. DETAILED DESCRIPTION

[0039] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0040] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise" and the like to indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as limiting the present invention.

[0041] In addition, the terms "first" and "second" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the said features. In the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined. In addition, the terms "installed", "connected" and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. The present invention will be further described in detail below with reference to the accompanying drawings.

[0042] Reference Figures 1 to 4 : A lupus disease intelligent monitoring system based on data fusion, comprising:

[0043] Multi-source data acquisition module: For physiological data collection, the core device is a smartwatch integrated with multiple high-precision sensors. The watch's built-in heart rate sensor uses a photoelectric transmission measurement method. By illuminating the skin with light of a specific wavelength, the watch accurately captures changes in blood volume caused by the heartbeat based on the periodic changes in blood absorption of light, thereby obtaining real-time and accurate heart rate (HR) data. Blood pressure measurement utilizes a new oscillometric sensor that can sensitively sense the vibration waves of the arterial walls. Combined with a complex algorithm, it dynamically monitors blood pressure (BP) values. The skin temperature (ST) sensor uses infrared sensing technology to quickly and accurately measure body surface temperature without contacting the skin, capturing subtle temperature fluctuations. The step count (SS) relies on a high-precision accelerometer that accurately counts each step by identifying wrist acceleration, direction, and other information. When connecting to external medical testing equipment, Bluetooth 5.0 low power consumption or high-speed Wi-Fi 6 technology is used to achieve stable and fast data transmission. By connecting to a device that detects the concentration of anti-double-stranded DNA antibodies (dsDNA) in the blood, this key autoimmune indicator can be obtained in real time. By communicating with an instrument that detects the erythrocyte sedimentation rate (ESR), relevant information such as inflammatory responses can be obtained. In addition, the module allows patients to actively input self-assessment data on symptoms, such as fatigue level (FS) and joint pain level (JPS). Through a simple and easy-to-use interactive interface, patients can enter their feelings at any time. At the same time, through standardized interfaces, deep integration with the electronic medical record system automatically synchronizes information such as past medical history (DHS) and medication record (MR). The sensor on the smartwatch uses new micro-electromechanical technology. By optimizing the sensor's microstructure design, such as reducing the size of the sensitive element and increasing the integration level, the accuracy and stability of data collection have been greatly improved, providing a solid data foundation for subsequent medical analysis and diagnosis.

[0044] Data preprocessing module: In the denoising process, an adaptive filtering algorithm is used. This algorithm is highly intelligent and can dynamically adjust the filtering parameters based on the real-time characteristics of the collected physiological signals (such as heart rate, blood pressure, etc.). When motion artifacts or electromagnetic interference generate noise, the adaptive filter will keenly capture abnormal fluctuations in the signal. By analyzing the noise frequency, amplitude and other characteristics, a matching filtering strategy is adopted to accurately remove the noise from the original physiological signal to ensure the purity of the signal and provide reliable basic data for subsequent analysis. For data normalization of different dimensions, an improved Z-score normalization formula is used. In this formula, μ s and σ s Representing the mean and standard deviation of data within a specific time period, they are key statistics for measuring trends and dispersion in a dataset. The coefficients k and b are not fixed values ​​but are fine-tuned based on factors such as the distribution characteristics and range of variation of different types of physiological data. For example, for step count data, which fluctuates widely, and skin temperature data, the values ​​of k and b may differ. This ensures comparability after normalization and facilitates subsequent unified analysis. For feature extraction, a feature extraction model based on a convolutional neural network (CNN) is employed. The convolutional layers in the CNN model use sliding convolutions on the physiological signal data using different convolution kernels, automatically mining the characteristics of heart rate variability (HRV) in the time domain, such as the intervals between adjacent heartbeats, and the distribution characteristics of different frequency components in the frequency domain. For blood pressure fluctuation data, the model accurately captures its periodic variation patterns, identifying characteristics such as the rise and fall phases of blood pressure within a cycle, as well as the amplitude of fluctuations. This deep learning-based feature extraction approach is more efficient and accurate than traditional methods, capable of mining subtle features that are difficult to detect manually, providing rich and valuable information for medical applications such as disease diagnosis and health assessment.

[0045] Data Fusion Module: Leveraging advanced Deep Belief Network (DBN) technology, various physiological and medical data are organically integrated. First, a data fusion model based on a DBN is constructed. The DBN consists of multiple layers of stacked Restricted Boltzmann Machines (RBMs). Restricted Boltzmann Machines are energy-based models with a unique architecture where neurons within a layer are disconnected and neurons between layers are fully connected. During the data fusion process, preprocessed data, such as heart rate, blood pressure, blood pressure indicators, and self-assessment of symptoms, are fed into the DBN as input. Each RBM layer performs the important task of feature learning. During learning, the RBM adjusts the connection weights between neurons to capture underlying patterns and features in the data. For example, for heart rate variability data, the RBM exploits features such as periodicity and trend in its time series; for blood pressure data, it learns features such as fluctuation amplitude and frequency of change. The features learned by one RBM layer serve as input to the next layer. As the number of layers increases, the model abstracts and refines the data from different levels and perspectives, gradually learning from low-level features to high-level, comprehensive features. After layer-by-layer processing of multi-layer RBM, the fused feature vector FV is finally obtained. In order to quantify the comprehensive health status of an individual, the calculation formula of the fused comprehensive health index is introduced Among them, FV i It is the dimensions of the fusion feature vector FV, representing the key health features extracted from different data; i These weights are assigned to corresponding dimensions. These weights are not set arbitrarily but are optimized through the backpropagation algorithm. Based on the principle of gradient descent, the backpropagation algorithm uses the error between the model's predicted value and the true value as a guide to continuously adjust the weight parameters, ensuring that the comprehensive health indicators more accurately reflect the individual's true health status. Through this complex and sophisticated series of operations, the data fusion module achieves deep integration and value enhancement of multivariate data, providing a more comprehensive and accurate basis for medical decision-making.

[0046] The disease assessment module utilizes a deep learning model based on a long short-term memory (LSTM) network combined with an attention mechanism. The LSTM network is a special type of recurrent neural network (RNN), unique in its memory cells and gating structure. Memory cells store information, acting like an "information warehouse," storing key information from medical data for extended periods. The input gate, forget gate, and output gate act like "intelligent valves," controlling the inflow, outflow, and retention of information. When processing medical data with time-series characteristics, such as changes in a patient's symptoms over time or fluctuations in vital signs, the LSTM network effectively learns long-term dependencies through these gating mechanisms, avoiding the vanishing or exploding gradient problems common in traditional RNNs. This allows the model to accurately capture the evolution of the disease over extended periods of time. The integration of the attention mechanism makes the model more intelligent. In a vast amount of medical data, not all data is equally important. The attention mechanism enables the model to automatically focus on key data features. For example, when assessing a heart disease patient, the model will prioritize critical information such as recent abnormal electrocardiogram fluctuations and rapid changes in blood pressure, while deemphasizing more common and stable physiological indicators. In this way, the model can extract the most valuable parts from the complex data and improve the accuracy of the assessment. The output of the model is the probability distribution of the severity of the disease, which divides the disease into three levels: remission period (RP), stable period (SP), and active period (AP). In order for the model to make accurate predictions, it needs to be trained with a large amount of labeled historical data. During the training process, the loss function uses the cross entropy loss function. Among them, y j It is the label of the real disease level, representing the actual disease status; j is the probability predicted by the model. By continuously adjusting its parameters to minimize the loss function, the model keeps the predicted probability as close to the true label as possible, thereby improving the model's performance in disease assessment tasks and providing a reliable reference for doctors to formulate treatment plans and determine patient prognosis.

[0047] Early warning module: In terms of early warning threshold setting, the traditional fixed threshold mode is abandoned and a dynamic early warning threshold is adopted. First, based on the probability distribution of the severity of the disease output by the disease assessment module, the possibility of the disease being at different levels and the changing trend are analyzed. For example, if the probability of the disease being in the active phase (AP) gradually increases, the early warning threshold needs to be adjusted accordingly. At the same time, combined with the changing trends of various physiological data, such as the fluctuation amplitude and rate of indicators such as heart rate and blood pressure over a period of time. In addition, the individual characteristics of the patient cannot be ignored. Age, underlying medical history, genetic factors, etc. will affect the setting of the early warning threshold. There must be differences in the early warning thresholds for young patients in good physical condition and elderly patients with multiple chronic diseases. For the prediction of the changing trend of disease indicators, the gray prediction model GM(1,1) is adopted, and the formula is This model has the unique advantage of being able to make effective predictions based on a small amount of data. (0) Records the historical data of disease-related indicators, and generates x through accumulation operation (1) Sequence analysis mitigates the randomness of the raw data and reveals underlying regularities. The model's parameters a and b are determined by a specific algorithm and determine the shape and trend of the prediction curve. This model can predict changes in medical indicators over the next few days, such as whether blood pressure will rise or heart rate will fluctuate abnormally. When the gray prediction model identifies a worsening trend and relevant indicators reach pre-set dynamic warning thresholds, the warning module takes swift action. Warnings are issued through various channels. Smartwatches vibrate and provide voice prompts, ensuring immediate notification. Detailed warning information is also pushed via a mobile app. This information not only includes potential changes in the patient's condition, such as blood pressure potentially exceeding normal ranges or worsening symptoms, but also provides emergency response recommendations, such as prompting patients to rest and informing them of simple first aid measures. This ensures that patients can take timely action before their condition worsens, buying valuable time for medical intervention.

[0048] The data storage and management module utilizes distributed hash table (DHT) technology to build a distributed database. DHT technology evenly distributes data across multiple nodes, avoiding the performance bottlenecks and single points of failure associated with centralized data storage. Each node is responsible for storing and managing only a portion of the data, and nodes communicate and collaborate with each other via a specific network protocol. When data needs to be stored or retrieved, the system calculates the data's storage location using a hash algorithm and quickly locates the corresponding node, enabling efficient data access. This distributed storage approach significantly improves data availability. Even if some nodes fail, the system remains operational, ensuring continuous data accessibility. Furthermore, data redundancy and replica management mechanisms ensure fault tolerance and prevent data loss. Homomorphic encryption is employed during data storage. Homomorphic encryption is a specialized encryption method that allows specific computational operations to be performed directly on encrypted data without first decrypting it. For example, in a medical data statistical analysis scenario, analysts can perform statistical calculations such as summation and averaging on encrypted patient physiological data, and the resulting results are consistent with those obtained on the plaintext data. This feature protects data privacy while maintaining normal data analysis and processing, meeting the dual needs of data security and data utilization in the medical field. Regarding data access control, a blockchain-based access control mechanism is provided. Blockchain is decentralized and tamper-proof. Patients and authorized doctors each have different private keys, which act like personal digital identities. The system manages data access and operations based on the permissions recorded on the blockchain. When a doctor needs to access a patient's medical data, their request is compared and verified against the permissions recorded on the blockchain. Only when the permissions match can the doctor decrypt and access the corresponding data using the private key. This mechanism fundamentally eliminates unauthorized data access, ensuring the security and privacy of medical data and comprehensively protecting patients' health information.

[0049] User Interaction Module: The interactive interface is developed to create a unified yet distinctive interface tailored to the specific characteristics of different devices. As a convenient, portable device, the smartwatch's interface design strives for simplicity and intuitiveness. Key medical data, such as real-time heart rate and blood pressure readings, is displayed in large fonts and high-contrast colors, allowing patients to access important health information at a glance. Warning messages are presented with eye-catching icons and concise text to ensure immediate attention. Furthermore, the watch's vibration and sound alerts provide timely notifications, ensuring no important health signals are missed. The mobile app further expands its functionality, offering more detailed data charts and analytical reports. Collected physiological data, such as steps and sleep quality, is presented in intuitive line and bar charts, allowing patients to easily understand trends in their health data. Analytical reports, powered by specialized algorithms, provide in-depth analysis of medical data, such as assessing autonomic nervous system function based on heart rate variability, providing patients with a comprehensive understanding of their health. Furthermore, the app supports detailed data queries and historical record tracking, meeting patients' needs for comprehensive health information. The desktop version is primarily intended for physicians, allowing them to comprehensively review patient data and perform diagnostics. Designed according to medical workflows, the interface integrates all patient medical information, including physiological data, medical assessment results, and historical records. Physicians can use the desktop version to accurately screen and compare data, drawing on professional diagnostic tools to diagnose and develop treatment plans. It also supports the editing and storage of electronic medical records, enabling efficient management of medical information. The interface utilizes adaptive layout technology. By detecting device parameters such as screen size and resolution, it automatically adjusts the size, position, and display of interface elements. For example, on a large-screen desktop, multiple data windows and detailed diagnostic tools can be displayed simultaneously; on a small-screen smartwatch, only key information is highlighted, ensuring optimal visual quality and user experience on any device. Furthermore, emotional interaction is incorporated. The system uses voice and intonation analysis to capture patient emotional data. If negative emotions such as anxiety or frustration related to the patient's condition are detected, the interface provides personalized encouragement and reassurance. For example, when patients are checking their medical reports, warm reminders will pop up at the right time, such as "Don't worry, your condition is under control, and you will gradually get better if you persist in treatment." This allows patients to feel humanistic care during the medical process, improving their medical experience and treatment compliance.

[0050] Model Update Module: Utilizing a federated learning framework, each medical institution trains the disease assessment model using local data while protecting patient data privacy. This framework breaks with the traditional centralized data training model. Instead of uploading patient data to a central server, each medical institution trains the disease assessment model locally using its own data. During local training, medical institutions utilize encryption techniques to process model parameters to ensure data privacy. For example, techniques such as homomorphic encryption or differential privacy are used to securely encrypt model parameters without leaking the original data. Subsequently, medical institutions exchange the encrypted model parameters via a secure communication channel. These parameters are then integrated at the aggregation center and updated globally using specific aggregation algorithms, such as weighted averaging. This approach prevents centralized exposure of patient data while fully leveraging the diverse data across medical institutions, improving the model's generalization capabilities. To further enhance the model's training efficiency and adaptability, a transfer learning algorithm is introduced. First, pre-training is performed on large-scale, generalized data covering a wide range of patient populations and disease types. Deep neural network training is then used to generate an initial model with reasonable generalization capabilities. When a specific medical institution needs to optimize a model for its own patient population, this initialization model serves as a foundation. Because patient populations vary across institutions, fine-tuning the model based on the institution's local data allows the model to quickly adapt to the characteristics of that specific patient population. During fine-tuning, adjustments are primarily made to certain model parameters, particularly those for layers closely related to local data characteristics. For example, for institutions with a preponderance of elderly patients, adjustments are focused on model parameters related to common geriatric diseases. This accelerates model convergence, reduces training time and computing resources, and improves the accuracy and adaptability of the model's assessment of conditions across diverse patient populations, enabling more precise service for diverse patient groups.

[0051] The present invention also includes the following modules:

[0052] Personalized analysis module: Focusing on the patient's genetic data (GD), it conducts a comprehensive search through an in-depth gene-disease association database, which stores a vast amount of scientifically verified gene-disease relationship information. Professional algorithms carefully screen for genetic markers that are closely related to lupus. These genetic markers are like "secret codes" hidden in the human genetic code, suggesting an individual's susceptibility to lupus and the tendency of the disease to develop. Once the relevant genetic markers are determined, the genetic risk score (GRS) is constructed. Formula is the core of the calculation, where w gi is the weight of gene marker i. It is not set arbitrarily, but is determined based on a large amount of medical research and clinical data, taking into account the importance and impact of gene markers in the pathogenesis of lupus. iThe genotype score for gene marker i is quantified based on the type and frequency of different gene variants. Ultimately, a precise calculation results in a unique genetic risk score for each patient, quantifying the potential impact of genetic factors on lupus disease. In addition to genetic data, lifestyle data is also a key component of personalized analysis. Taking dietary frequency (DF) as an example, the system records detailed information such as the patient's mealtimes, food types, and amounts consumed at each meal. Frequent consumption of high-sugar and high-fat foods may exacerbate inflammatory responses and affect lupus disease progression. Regarding exercise intensity (EI), sensors in devices such as smartwatches accurately monitor heart rate, acceleration, and other data during exercise, which are then converted into exercise intensity. Patients who are sedentary and engage in very low-intensity exercise may have compromised immunity and metabolic function, increasing the risk of lupus exacerbation. The Analytic Hierarchy Process (AHP) is used to systematically and scientifically determine the weights of personalized assessment indicators. The AHP classifies multiple factors, such as genetic risk score, dietary frequency, and exercise intensity, according to their importance, and determines the relative weights of each factor through pairwise comparisons. For example, for patients with a family history of lupus, the weight of the genetic risk score will be relatively increased; and for patients who have recently changed their lifestyle suddenly and have obvious bad habits, the weight of lifestyle data will be increased. The weight of each factor is combined to revise the disease assessment results. For example, for patients with specific high-risk genes and low physical activity, the system will appropriately increase the assessment level of disease severity based on the weight calculation, providing doctors with a more tailored diagnosis of the disease to the individual patient's situation and helping to develop more targeted treatment and management plans.

[0053] The present invention also includes the following modules:

[0054] The telemedicine module utilizes 5G communication technology, featuring high speeds, low latency, and large capacity, providing a solid foundation for high-speed data transmission between patients and doctors. Physiological data collected by patients' smart devices, such as heart rate, blood pressure, and blood pressure indicators, can be transmitted to doctors at extremely fast and stable speeds. This effectively addresses issues previously associated with delayed data transmission and lag caused by network latency, ensuring that doctors receive the latest patient health information in real time. The module utilizes video stream encoding and compression technologies, such as H.265. The H.265 encoding standard uses advanced algorithms to significantly reduce data transmission while maintaining video quality. It efficiently encodes images in videos and intelligently identifies and compresses redundant information. Even under limited network conditions, doctors can receive clear and smooth video footage of patients, allowing them to observe their mental state, facial features, and other characteristics in detail. This intuitive information is crucial for diagnosis. The core function of this module is real-time remote consultation. Doctors can view various patient monitoring data and high-definition video in real time through the system. The system provides an integrated interface that integrates data and video, facilitating comprehensive evaluation. Virtual annotation technology plays a crucial role in this process. Doctors can mark and annotate patient body parts or test images (such as X-rays and ultrasound images). For example, abnormal areas on joint images can be marked and captioned. This facilitates communication with patients and other medical staff, accurately conveying diagnostic ideas and key points, and improving consultation efficiency and accuracy. Furthermore, remote operation assistance technologies, exemplified by force feedback devices, are being introduced. These devices provide real-time feedback to nurses regarding the doctor's force and direction of manipulation. For simple medical procedures, such as venipuncture positioning and wound dressing changes, doctors can remotely operate force feedback devices to simulate the feel of actual manipulation and provide remote guidance to nurses. Nurses can sense the doctor's intended manipulations through force feedback, improving the standardization and accuracy of their procedures. This overcomes spatial limitations and enables precise remote medical guidance, allowing patients to enjoy the guidance of highly skilled medical experts in their local area.

[0055] In the present invention, the multi-source data acquisition module adds the function of environmental data acquisition. Through the micro environmental sensors integrated in the smart watch or peripheral devices, the ultraviolet intensity (UVI), allergen concentration (AC), noise level (NL) and other data in the environment are collected. In the data fusion module, the environmental data is fused with other physiological and medical data, and the improved Bayesian fusion formula is used. (where H represents the disease state, E i Representing various types of data, including environmental data), to more accurately calculate comprehensive health indicators.

[0056] In this paper, the data preprocessing module uses a generative adversarial network (GAN) to repair small amounts of missing data. The generator learns the distribution of real data and generates reasonable missing data values. The discriminator distinguishes between generated data and real data. Adversarial training improves the accuracy of missing data repair. Simultaneously, a wavelet transform is used to perform multi-scale analysis on the data, extracting features of different frequency components, providing richer information for subsequent disease assessment.

[0057] In the present invention, the condition assessment module uses a multimodal fusion Transformer model to perform condition assessment. This model extracts features from image data (such as high-definition images of skin erythema) through a convolutional neural network, converts text data (such as patient symptom descriptions) into vector representations through a word vector model, and normalizes numerical data (such as various physiological indicators) before inputting them into the Transformer model. The Transformer model fuses and analyzes data from different modalities through a multi-head attention mechanism, improving the accuracy and reliability of condition assessment.

[0058] In this invention, the early warning module uses an algorithm that combines fuzzy logic and reinforcement learning to determine the warning threshold. Fuzzy logic is used to fuzzify factors such as condition assessment results, data trends, and individual patient characteristics to construct a fuzzy rule base. Using a reinforcement learning algorithm, the fuzzy rule base and warning threshold are continuously optimized based on actual changes in the patient's condition. For example, if a patient's condition worsens, the warning threshold is adjusted to ensure more timely warnings and improve their effectiveness.

[0059] In this invention, the data storage and management module combines blockchain and IPFS (InterPlanetary File System) to store data. Blockchain stores metadata and hash values, ensuring data immutability and traceability; IPFS stores actual data files, utilizing content-addressable technology to improve data storage and retrieval efficiency. Furthermore, a data visualization tool has been developed to present statistical analysis results through interactive charts, enabling doctors and patients to intuitively understand disease trends.

[0060] In this invention, the user interaction module uses brain-computer interface (BCI) technology to achieve more natural interaction. Patients can operate the smartwatch to view data or issue commands through simple brain activities (such as imagining hand movements). Using speech emotion recognition technology, the patient's voice emotion when interacting with the system is analyzed to provide patients with more personalized services. For example, if the patient is detected to be anxious, the system will automatically push relaxation training audio or video.

[0061] In the present invention, the watch serves as the core of the wearable device, integrating the physiological data acquisition function and some environmental data acquisition functions of the multi-source data acquisition module. The watch uses low-power Bluetooth technology to transmit data with external medical detection equipment and mobile phones, and sends the collected data to the data preprocessing module for processing. The watch has a built-in high-performance microprocessor that can preprocess some data in real time, reducing the burden on the system server. At the same time, the watch has the reminder function of the early warning module, which can remind patients through various methods such as vibration, sound, and light signals. In addition, the watch provides a simple and easy-to-use user interaction interface. Patients can view medical data and receive early warning information through touch operations, making it convenient for patients to monitor and manage their condition anytime and anywhere.

[0062] The above are only preferred specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with this technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solutions and inventive concepts of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. An intelligent lupus disease monitoring system based on data fusion, characterized in that: Includes the following modules: Multi-source data acquisition module: This module collects physiological data using smartwatches, connects to external devices via Bluetooth or Wi-Fi to obtain key indicator data, and obtains symptom self-assessment data from patients' active input and electronic medical record systems simultaneously, using new micro-electromechanical technology; Data preprocessing module: uses adaptive filtering algorithm to remove noise, adopts improved Z-score formula to normalize data of different dimensions, and uses convolutional neural network (CNN) feature extraction model to automatically extract physiological signal features; Data fusion module: Build a data fusion model based on deep belief network (DBN), take preprocessed data as input, fuse the feature vector FV through RBM feature learning, calculate the comprehensive health index HI using the formula, and optimize the weight by back propagation algorithm; Condition Assessment Module: This module uses a learning model based on a long short-term memory (LSTM) network combined with an attention mechanism to learn long-term dependencies between conditions, focusing on key data features, outputting a probability distribution of conditions and classifying it into three levels. It uses a cross-entropy loss function for training based on labeled data. Early warning module: Dynamic early warning thresholds are set based on the probability distribution of disease assessment, data change trends, and individual characteristics. A gray prediction model is used to predict disease conditions. When the threshold is reached, early warnings are issued in various ways, along with disease conditions and treatment recommendations. Data storage and management module: Build a distributed database using distributed hash table (DHT) technology, adopt homomorphic encryption algorithm for storage, provide a blockchain-based access control mechanism, and perform data access and operations based on the permission information recorded on the blockchain; User interaction module: Develop a unified interface for smartwatches, mobile apps, and desktops, using adaptive layout technology to automatically adjust display content based on device screen size and resolution, and introduce emotional interaction design; Model update module: Adopting a federated learning framework, the disease assessment model is trained with local data while protecting privacy. Model parameters are exchanged and updated in an aggregated manner through encryption technology, and the model is fine-tuned in combination with a transfer learning algorithm.

2. The intelligent lupus disease monitoring system based on data fusion according to claim 1 is characterized in that: Also includes: Personalized analysis module: Based on the patient's genetic data GD, search for gene markers related to lupus through the gene-disease association database and construct the genetic risk score GRS formula w gi is the weight of gene marker i, g i is the genotype score of gene marker i; combined with the lifestyle data of diet frequency DF and exercise intensity EI, the hierarchical analysis method AHP is used to determine the weight of personalized evaluation indicators and revise the disease assessment results.

3. The intelligent lupus disease monitoring system based on data fusion according to claim 1 is characterized in that: Also includes: Telemedicine module: It uses 5G communication technology combined with video stream coding and compression technology to support real-time remote consultation functions. Through the system, various monitoring data and high-definition videos of patients can be viewed in real time. Virtual labeling technology is used to mark and annotate patient body parts or test images. Remote operation assistance technology is introduced to remotely guide medical operations.

4. The intelligent lupus disease monitoring system based on data fusion according to claim 1, characterized in that: The multi-source data acquisition module adds environmental data acquisition function, collects ultraviolet intensity UVI, allergen concentration AC, and noise level NL data through micro environmental sensors; integrates environmental and other physiological and medical data, and uses the improved Bayesian fusion formula H indicates the condition, E i Represents various types of data; The calculation formula of the integrated health index is: w i is the weight of each dimension of the fused feature vector, and FV is the fused feature vector.

5. The intelligent lupus disease monitoring system based on data fusion according to claim 1 is characterized in that: Improved Z-score normalization formula in the data preprocessing module μ s and σ s are the mean and standard deviation of the data in a specific time period, k and b are coefficients adjusted according to the characteristics of the data; The generative adversarial network (GAN) is used to repair missing data. The generator learns the real data distribution to generate missing data values. The discriminator distinguishes between generated data and real data. The wavelet transform is used to perform multi-scale analysis on the data to extract the characteristics of different frequency components.

6. The intelligent lupus disease monitoring system based on data fusion according to claim 1 is characterized in that: The loss function in the disease assessment module adopts the cross entropy loss function y j is the true disease level label, p j is the model predicted probability; A multimodal fusion Transformer model is used to assess the condition. Image data is used to extract features through a convolutional neural network. Text data is converted into vector representation using a word embedding model. Numerical data is normalized and then input into the Transformer model. Data from different modalities are fused and analyzed through a multi-head attention mechanism.

7. The intelligent lupus disease monitoring system based on data fusion according to claim 1 is characterized in that: The grey prediction model GM(1,1) is used in the early warning module to predict the trend of disease indicators. The formula is: x (0) is the original data sequence, x (1) Generate a cumulative sequence for it, a and b are model parameters; An algorithm combining fuzzy logic and reinforcement learning is used to determine the warning threshold. Fuzzy logic is used to process the condition assessment results, data change trends, and individual patient characteristics to build a fuzzy rule library. Using reinforcement learning algorithms, the fuzzy rule base and warning thresholds are optimized through feedback from actual changes in the patient's condition.

8. The intelligent lupus disease monitoring system based on data fusion according to claim 1 is characterized in that: The data storage and management module uses a combination of blockchain and the Interstellar File System (IPFS) to store data. Blockchain stores data metadata and hash values, while IPFS stores actual data files. Data visualization tools are developed to display data statistical analysis results through interactive charts.

9. The intelligent lupus disease monitoring system based on data fusion according to claim 1, characterized in that: The user interaction module uses brain-computer interface (BCI) technology to achieve interaction, operates the smart watch through brain thinking activities to view data or issue commands, and uses voice emotion recognition technology to analyze voice emotions.

10. An application of a smart lupus disease monitoring system based on data fusion according to any one of claims 1 to 9 in a watch, characterized in that: The watch integrates physiological and partial environmental data collection functions, uses Bluetooth technology to transmit data with external medical testing equipment and mobile phones, and has a built-in microprocessor that pre-processes some data in real time. It reminds patients through vibration, sound, and light signals, and provides a simple interactive interface to view medical data and receive early warning information through touch operations.

Images