Personalized motion generation system and method suitable for dialysis patient

Abstract

The invention discloses a personalized movement generation system and method suitable for a dialysis patient, and the method comprises the steps: integrating heterogeneous data of medical equipment, a physiological monitoring device and a user terminal through a data collection and standardization module, and constructing a structured medical data set through a standardization method; the feature extraction and state modeling module extracts key features from treatment cycle parameters, real-time physiological indexes and user behavior preferences by applying a multi-modal feature separation technology, and constructs a multi-dimensional patient state vector; and the intelligent exercise scheme generation module fuses the treatment cycle features, the dynamic physiological features and the user preference features based on a pre-trained neural network model to generate an exercise scheme adaptive to the individual pathological state. Intelligent generation of the movement scheme suitable for the dialysis patient is achieved, the individuation degree is improved through multi-dimensional feature fusion, and the scientificity and safety of the scheme are ensured in combination with medical data standardization processing.

Description

Technical Field

[0001] The present invention belongs to the field of medical health technology, and in particular relates to a personalized motion generation system and method suitable for dialysis patients. Background Art

[0002] With the development of smart healthcare and exercise rehabilitation technologies for chronic diseases, data-driven exercise interventions for dialysis patients are gaining increasing attention. These technologies aim to generate dynamic exercise plans tailored to individual pathological characteristics by integrating multidimensional physiological data with treatment cycle parameters. However, traditional exercise management for dialysis patients still relies on clinicians' empirical judgment and static program design. Clinicians typically standardize exercise intensity and duration based on single laboratory test results and subjective patient feedback, lacking continuous monitoring of real-time physiological indicators and failing to integrate treatment cycle characteristics and patient behavioral preferences. Traditional approaches impose a rigid execution process on exercise plans, preventing dynamic adjustments to accommodate sudden physiological fluctuations during exercise, leading to safety risks. Furthermore, the lack of standardization hinders collaborative analysis of heterogeneous data from multiple sources. Feature extraction is limited to a single time slice or modality, making it impossible to model the nonlinear relationships between treatment cycles, physiological states, and user preferences. Furthermore, traditional efficacy evaluation relies on periodic laboratory retests and lacks real-time tracking of exercise execution and high-frequency vital sign changes, resulting in delayed program optimization. This makes it difficult for existing technologies to achieve precise and safe exercise interventions for dialysis patients. Summary of the Invention

[0003] Based on this, it is necessary to provide a personalized motion generation system and method suitable for dialysis patients that can solve the above problems.

[0004] In a first aspect, the present application provides a personalized motion generation system suitable for dialysis patients, comprising:

[0005] The data acquisition and standardization module is used to obtain multi-source data sets from medical equipment, physiological monitoring devices and user terminals, and process them using data standardization methods to generate standardized medical data sets;

[0006] The feature extraction and state modeling module is used to separate treatment cycle features, physiological features, and user preference features based on standardized medical data sets using a multimodal feature extraction method to construct a patient state vector;

[0007] The intelligent exercise plan generation module is used to generate an exercise plan based on the patient's state vector and adopt a pre-trained neural network model to fuse treatment cycle characteristics, physiological characteristics and preference characteristics.

[0008] In one embodiment, the system further includes a dynamic exercise plan adjustment module for:

[0009] The physiological monitoring device collects the patient's real-time physiological sign data stream during the exercise execution;

[0010] Use dynamic fluctuation detection algorithm to analyze real-time physiological sign data stream in real time to identify abnormal physiological fluctuation signals;

[0011] When abnormal physiological fluctuation signals are detected, the safety constraint rule matching engine is triggered to perform risk assessment;

[0012] Based on the preset physiological feedback adjustment strategy, updated exercise instructions are generated according to risk assessment, and the updated exercise instructions are used to dynamically adjust the exercise intensity parameters or exercise duration parameters in the exercise plan.

[0013] In one embodiment, the motion scheme dynamic adjustment module is further configured to construct a dynamic fluctuation detection algorithm using the following formula:

[0014]

[0015] Among them, A(t) represents the abnormal fluctuation index, ω(τ) represents the time decay weight, represents the wavelet domain gradient operator, SWT(x(τ)) represents the stationary wavelet transform coefficient, KL(p||q) τ represents KL divergence, CUSUM k (t) represents the cumulative sum statistic, ζ represents the mutation component amplification coefficient, and Δt represents the dynamic detection window.

[0016] In one embodiment, the system further includes an effect evaluation module for:

[0017] Align the time stamps of the updated motion instructions and real-time physiological sign data streams and extract motion execution record data;

[0018] Statistical modeling methods are used to process exercise execution record data to generate exercise execution indexes and physical sign change curves;

[0019] The exercise execution index, physical sign change curve and exercise plan are correlated and mapped to generate an effect evaluation dataset containing multi-dimensional features.

[0020] In one embodiment, the effect evaluation module is further configured to:

[0021] Based on the parameter association matrix, a dynamic association analysis algorithm is used to calculate the association strength between each treatment cycle feature and the physical sign change curve, and the features with association strength exceeding the preset threshold are screened to form an optimized decision parameter set;

[0022] If the key indicators in the optimized decision parameter set are lower than the preset threshold, the reinforcement learning model is triggered to update the weight parameters of the neural network model.

[0023] In one embodiment, the system further includes a visualization module for:

[0024] The multidimensional feature parameters in the effect evaluation dataset are integrated with the historical medical record data, and a time series correlation matrix is ​​constructed using a sliding time window algorithm.

[0025] The time series correlation matrix is ​​processed using the t-SNE dimensionality reduction algorithm and mapped to a three-dimensional space coordinate system;

[0026] Based on the three-dimensional spatial coordinates mapped by the time series correlation matrix, a visualization rendering engine is used to generate visualization images, which include dynamic heat maps, three-dimensional scatter matrix and time axis synchronization comparison chart.

[0027] In one embodiment, the system further includes a remote monitoring module for:

[0028] Synchronously collect the patient's physiological characteristic data and motion trajectory spatial coordinates during exercise to generate a monitoring data stream;

[0029] De-noise the monitoring data stream and identify abnormal physiological fluctuations and movement trajectory deviations;

[0030] When physiological fluctuations or movement trajectory deviations exceed the preset threshold, a graded alarm is triggered according to the preset alarm mechanism.

[0031] In a second aspect, the present application further provides a personalized motion generation method applicable to dialysis patients, comprising:

[0032] Acquire multi-source data sets from medical equipment, physiological monitoring devices, and user terminals, and process them using data standardization methods to generate standardized medical data sets;

[0033] Based on the standardized medical dataset, a multimodal feature extraction method is used to separate treatment cycle features, physiological features, and user preference features to construct a patient state vector.

[0034] According to the patient's state vector, a pre-trained neural network model is used to fuse treatment cycle characteristics, physiological characteristics and preference characteristics to generate an exercise plan.

[0035] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of implementing the above-mentioned personalized motion generation system for dialysis patients are as follows:

[0036] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned personalized motion generation system for dialysis patients:

[0037] The above-mentioned personalized exercise generation system and method for dialysis patients integrates heterogeneous data from medical equipment, physiological monitoring devices and user terminals, adopts standardized methods to eliminate differences in data formats and time scales, constructs a structured medical data set, and solves the problem of one-sided feature extraction caused by the fragmentation of multi-source data in traditional technologies; based on the feature extraction and state modeling module, multimodal separation and fusion of treatment cycle characteristics, physiological characteristics and user preference characteristics are carried out to construct a multidimensional patient state vector, breaking through the limitations of traditional methods that rely on a single data dimension such as static biochemical indicators, and realizing refined modeling of individual pathological states and behavioral preferences; through the pre-trained neural network model in the intelligent exercise plan generation module, the multi-dimensional features are nonlinearly mapped to the decision space of exercise intensity, duration and type, and a personalized exercise plan that dynamically adapts to the patient's real-time physiological state and changes in treatment stages is generated, overcoming the defect of traditional standardized plans that ignore individual differences and the dynamic relationship between treatment cycles, and significantly improving the scientificity and safety of exercise intervention. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0039] Figure 1 This is a structural diagram of a personalized motion generation system suitable for dialysis patients according to the present invention;

[0040] Figure 2 This is a structural diagram of an embodiment of a personalized motion generation system for dialysis patients according to the present invention;

[0041] Figure 3 This is a flow chart of a personalized motion generation method suitable for dialysis patients according to the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0043] The relevant hardware architecture of this application includes data acquisition equipment, user terminals and edge computing nodes. In scenarios where personalized exercise plans need to be generated, the data acquisition equipment monitors the patient's physiological indicators and treatment data in real time and transmits them to the edge computing node. The edge computing node integrates and analyzes multi-source information and generates an exercise plan and sends it to the user terminal. The user terminal receives instructions and feedbacks the execution status, realizing dynamic adjustment and safety control of the exercise plan through hardware collaboration.

[0044] In one embodiment, Figure 1 As shown, a personalized motion generation system suitable for dialysis patients is provided. This embodiment takes the system deployed on an edge computing node that can be integrated into a medical gateway or a local server as an example. At the same time, the system supports a distributed architecture: data acquisition equipment and user terminals serve as edge perception and interaction nodes, and the generation and dynamic adjustment functions of motion plans can be independently operated by edge computing nodes, or realized by collaborative computing of cloud servers. It should be noted that the system can also be adapted to a terminal-server hybrid architecture, for example, the user terminal performs data acquisition and instruction feedback, the server completes model calculation and solution optimization, and the two exchange data through network protocols. In this embodiment, the system includes:

[0045] The data acquisition and standardization module 101 is used to obtain multi-source data sets from medical equipment, physiological monitoring devices and user terminals, and process them using a data standardization method to generate a standardized medical data set.

[0046] Among them, medical equipment can provide data directly related to dialysis treatment, such as dialysis time, dialysate flow, solute clearance and other treatment parameters recorded by the dialysis machine, reflecting the specific process and effect of dialysis treatment. Physiological monitoring devices can monitor patients' physiological indicators in real time, such as heart rate, blood pressure, blood oxygen saturation, blood sugar level, etc., and can intuitively present the patient's physical function status during dialysis and in daily life. User terminals can collect patients' behavioral preference information, such as their daily exercise habits such as exercise type and frequency, dietary preferences and other information. Since the data generated by medical equipment, physiological monitoring devices and user terminals may differ in format, unit, time scale and other aspects, they can be standardized by using a unified data format, aligned time scale and processing of missing values ​​and outliers to obtain a standardized medical data set to provide data support for the calculation of subsequent modules.

[0047] The feature extraction and state modeling module 102 is used to separate treatment cycle features, physiological features and user preference features based on the standardized medical data set using a multimodal feature extraction method to construct a patient state vector.

[0048] Multimodal feature extraction methods can be used to separate treatment cycle features, physiological features, and user preference features. Regarding treatment cycle features, dialysis treatment is typically cyclical, and patients' physical conditions vary at different stages of treatment. By analyzing information such as dialysis time, frequency, and dose from standardized medical datasets, features reflecting the treatment process can be extracted. For example, it can be used to determine whether a patient is in the initial, stable, or adjustment phase of dialysis. Exercise tolerance and needs vary across different phases. Physiological features encompass various patient indicators, such as heart rate, blood pressure, blood oxygen saturation, and renal function indicators such as creatinine and urea nitrogen. Changes in these indicators can directly reflect a patient's physical function and health status, directly influencing the development of an exercise plan. User preference features focus on a patient's personal preferences and habits, including preferences for exercise types such as walking, yoga, and swimming, and exercise times such as morning, afternoon, and evening. Understanding these preferences can improve patient compliance with exercise plans and make them more relevant to patients' daily lives. After separating treatment cycle features, physiological features, and user preference features, these different types of features are integrated to construct a patient state vector. The patient state vector is a multidimensional vector that combines multiple key features in the form of mathematical vectors to comprehensively and comprehensively describe the patient's individual condition. Each dimension represents a characteristic, simplifying complex patient information into a vector form that can be used for calculation and analysis. Subsequent modules can use the patient state vector to generate personalized exercise plans that suit the patient's current physical condition, treatment stage, and personal preferences. This enables precise and personalized exercise intervention, improving the effectiveness and safety of exercise therapy.

[0049] The intelligent exercise plan generation module 103 is used to generate an exercise plan based on the patient's state vector by using a pre-trained neural network model to fuse treatment cycle characteristics, physiological characteristics and preference characteristics.

[0050] The neural network model can employ a modality fusion network architecture, comprising a feature encoding layer, a cross-modal attention mechanism, and a feature fusion layer. It is trained using a historical dialysis patient dataset and continuously optimized through a reinforcement learning framework. The trained neural network model integrates treatment cycle characteristics, physiological characteristics, and preference features to capture and understand the nonlinear relationships between these features. For example, for patients with poor renal function and a high heart rate during the dialysis stable phase, given their preference for evening exercise, a low-intensity walk and minimal exercise duration may be more appropriate. By analyzing and processing the patient's state vector, the intelligently generated exercise plan is highly personalized, encompassing key elements such as intensity, duration, and type of exercise. Exercise intensity can be measured using heart rate zones and metabolic equivalents of exercise, ensuring that patients receive the desired effect without experiencing excessive physical strain. Exercise duration can be adjusted based on the patient's physical endurance and treatment stage to avoid excessive fatigue. The type of exercise is selected based on the patient's preferences and physical condition. For example, for patients with poor joint function, high-impact exercises like long-distance running and rope skipping are avoided, while low-impact exercises like swimming or cycling are recommended. The generated exercise plan takes into account multiple factors to better promote the recovery and health management of dialysis patients.

[0051] In one embodiment, Figure 2 As shown, the system further includes an exercise plan dynamic adjustment module 104, which is used to:

[0052] The physiological monitoring device collects the patient's real-time physiological sign data stream during the exercise execution;

[0053] Use dynamic fluctuation detection algorithm to analyze real-time physiological sign data stream in real time to identify abnormal physiological fluctuation signals;

[0054] When abnormal physiological fluctuation signals are detected, the safety constraint rule matching engine is triggered to perform risk assessment;

[0055] Based on the preset physiological feedback adjustment strategy, updated exercise instructions are generated according to risk assessment, and the updated exercise instructions are used to dynamically adjust the exercise intensity parameters or exercise duration parameters in the exercise plan.

[0056] Specifically, physiological monitoring devices can be wearable devices such as smart bracelets, smart watches, or other professional medical monitoring devices. They can continuously monitor key physiological indicators such as heart rate, blood pressure, blood oxygen saturation, and respiratory rate, providing real-time insights into the patient's physical responses during exercise. A dynamic fluctuation detection algorithm can be used to analyze the real-time physiological sign data stream in real time to identify abnormal physiological fluctuation signals. When an abnormal physiological fluctuation signal is detected, a safety constraint rule matching engine is triggered to perform a risk assessment. The safety constraint rule matching engine, which pre-sets safety rules and standards, assesses the potential risks associated with the current exercise based on the detected abnormal physiological fluctuation signals and the patient's basic health data, such as age and underlying medical conditions. If the patient's heart rate exceeds the upper limit of the safe heart rate for their age within a short period of time, the safety constraint rule matching engine, taking into account factors such as the patient's cardiovascular history, determines the risk level of the abnormality. Updated exercise instructions are used to dynamically adjust the exercise intensity or duration parameters within the exercise plan. If the risk assessment results indicate that the current exercise poses a certain risk, adjustments can be made based on pre-set strategies. If the risk is determined to be mild, the exercise intensity may be appropriately reduced, such as by lowering the target heart rate during exercise. If the risk is high, the exercise duration may be shortened or even suspended. This dynamic adjustment ensures the patient's safety during exercise and optimizes the exercise plan based on the patient's real-time physical condition, thereby improving the effectiveness of exercise therapy.

[0057] In one embodiment, the motion scheme dynamic adjustment module 104 is further configured to construct a dynamic fluctuation detection algorithm using the following formula:

[0058]

[0059] Among them, A(t) represents the abnormal fluctuation index, ω(τ) represents the time decay weight, represents the wavelet domain gradient operator, SWT(x(τ)) represents the stationary wavelet transform coefficient, KL(p||q) τ represents KL divergence, CUSUM k (t) represents the cumulative sum statistic, ζ represents the mutation component amplification coefficient, and Δt represents the dynamic detection window.

[0060] For example, the time decay weight ω(τ) reflects the change in the degree of influence of data at different moments on the current abnormal fluctuation index over time. As time goes by, the importance of data at earlier moments to the current abnormal judgment gradually decreases, and the value of ω(τ) will decrease, reflecting the timeliness of the data. It can be obtained based on experience, data change characteristics, or through analysis and fitting of a large amount of physiological sign data. During the exercise of dialysis patients, recent physiological data can better reflect the current physical state. By setting a suitable ω(τ), the influence of recent data on abnormal fluctuation judgment can be highlighted, avoiding misjudgment due to interference from long-term data. Wavelet domain gradient operator It is used to calculate the gradient of stationary wavelet transform coefficients in the wavelet domain. When processing physiological sign data streams, it can capture the changing trend and rate of change of data at different frequencies and time scales. The calculation of can find the local change characteristics of the data, such as when analyzing heart rate data, It can accurately locate the heart rate mutation point and its change amplitude, providing a basis for subsequent abnormality judgment. By decomposing the physiological sign data stream into different frequency sub-bands, the obtained stationary wavelet transform coefficient SWT(x(τ)) contains the characteristic information of the signal at different frequency components. It can help analyze the details and trends of the signal. For example, when analyzing blood pressure data, the coefficients at different frequencies can reflect the different period and amplitude characteristics of blood pressure fluctuations. KL divergence KL(p||q) τ It is used to measure the difference between two probability distributions p and q. In this algorithm, they represent the distribution of physiological sign data under normal physiological state and the current monitoring state respectively. By calculating the KL divergence, the degree of difference between the current data distribution and the normal distribution can be quantified. The greater the difference, the more the current physiological state deviates from the normal state and the more likely there is abnormal fluctuation. The cumulative sum statistic CUSUM is obtained by cumulatively summing the changes in the data. k (t), can effectively detect small change trends in data. In physiological sign data monitoring, CUSUM k (t) It can gradually accumulate data change information and discover small, gradually developing abnormal changes. For example, when monitoring blood sugar data, if the blood sugar value has an abnormal trend of slowly rising or falling, CUSUM kBy continuously accumulating changes, the algorithm can detect trends promptly and avoid missing early abnormal signals. The mutation component amplification factor ζ is used to enhance sensitivity to sudden changes in the data. In physiological sign data, certain sudden abnormal changes may be crucial for assessing exercise risk, but these changes may account for a small portion of the overall data and be easily overlooked. ζ amplifies these sudden changes in the abnormal fluctuation index calculation, making the algorithm more sensitive to sudden anomalies. The dynamic detection window Δt defines the time range the algorithm focuses on when analyzing data. A shorter Δt can capture instantaneous abnormal changes more promptly but may be affected by noise; a longer Δt can smooth noise but delay the detection of anomalies. This can be determined by combining the frequency of change and noise characteristics of physiological sign data. For example, for relatively fast-changing data such as heart rate, a shorter Δt can be used; for relatively slow-changing data such as blood pressure, a longer Δt can be used to balance detection timeliness and accuracy.

[0061] In one embodiment, the system further includes an effect evaluation module 105 for:

[0062] Align the time stamps of the updated motion instructions and real-time physiological sign data streams and extract motion execution record data;

[0063] Statistical modeling methods are used to process exercise execution record data to generate exercise execution indexes and physical sign change curves;

[0064] The exercise execution index, physical sign change curve and exercise plan are correlated and mapped to generate an effect evaluation dataset containing multi-dimensional features.

[0065] Specifically, timestamp alignment ensures data accuracy and consistency by chronologically matching updated exercise instructions with real-time physiological sign data streams, ensuring that each exercise instruction corresponds to the corresponding physiological sign data. Exercise execution record data is extracted from the aligned data. This data contains key information about the patient's actual exercise process, such as exercise start and end times, changes in exercise intensity, and the execution status of each exercise instruction. Using statistical modeling methods to process exercise execution record data can extract valuable information from complex data. The exercise execution index measures the patient's degree of adherence to the exercise plan. This quantitative indicator is calculated by analyzing factors such as exercise intensity, duration, and frequency in the exercise execution record data and combining them with pre-set exercise plan standards. If the exercise plan stipulates that the patient's exercise intensity should be maintained within a certain heart rate range, and the patient's heart rate remains within this range for the majority of the time during exercise, the exercise execution index will be high. Vital sign change curves are generated through statistical analysis of real-time physiological sign data such as heart rate, blood pressure, and blood oxygen saturation. They visually demonstrate the temporal trends of physiological indicators during exercise, such as how heart rate gradually increases after the start of exercise, stabilizes after reaching a certain intensity, and then gradually decreases after the end of exercise. By observing these vital sign change curves, we can understand the impact of exercise on the patient's body and determine whether the exercise plan is appropriate. By mapping exercise execution indicators, vital sign change curves, and exercise plans, we can comprehensively evaluate the effectiveness of exercise plans. By comparing and analyzing these with actual exercise execution indicators and vital sign change curves, we can identify any implementation issues. A low exercise execution indicator may indicate that the exercise plan is too difficult or that patient compliance is poor. If the vital sign change curves indicate abnormal physiological reactions during exercise, such as elevated blood pressure or elevated heart rate, it may indicate that exercise intensity needs to be adjusted. By integrating this information to generate a multi-dimensional evaluation dataset, we can provide data support for subsequent optimization of exercise plans.

[0066] In one embodiment, the effect evaluation module 105 is further configured to:

[0067] Based on the parameter association matrix, a dynamic association analysis algorithm is used to calculate the association strength between each treatment cycle feature and the physical sign change curve, and the features with association strength exceeding the preset threshold are screened to form an optimized decision parameter set;

[0068] If the key indicators in the optimized decision parameter set are lower than the preset threshold, the reinforcement learning model is triggered to update the weight parameters of the neural network model.

[0069] For example, a dynamic correlation analysis algorithm is a method that dynamically measures the degree of association between variables. By calculating the strength of association between each treatment cycle characteristic and the vital sign change curve, this algorithm can provide a deeper understanding of the inherent connection between the different stages of dialysis treatment and the patient's physical response. Treatment cycle characteristics may include the number of dialysis sessions, the duration of dialysis sessions, and changes in dialysate composition. The vital sign change curve encompasses the dynamic changes in indicators such as heart rate, blood pressure, and blood oxygen saturation. If the correlation strength between a treatment cycle characteristic, such as an increase in the number of dialysis sessions, and the heart rate change curve exceeds a preset threshold, indicating that this characteristic has a significant impact on heart rate changes, it is selected and included in the optimized decision parameter set. This optimized decision parameter set is a key factor influencing the effectiveness of the exercise program and the patient's physiological response, providing an important basis for subsequent optimization of the exercise program. The key indicators in the optimized decision parameter set reflect important parameters of the exercise program's effectiveness and the patient's physiological state. The preset thresholds can be set based on clinical experience, medical research, and extensive practical data. If these key indicators fall below the preset threshold, it indicates that the current exercise program may not achieve the expected results or is not significantly improving the patient's physical condition, and the exercise program generation model needs to be adjusted. At this time, the reinforcement learning model is triggered to update the weight parameters of the neural network model. The reinforcement learning model learns the optimal strategy by interacting with the environment and based on the reward mechanism, taking the key indicators in the optimized decision parameter set as feedback. By continuously adjusting the weights of the neural network model and changing the way the model handles input features such as treatment cycle characteristics, physiological characteristics, and user preference characteristics, it can generate an exercise plan that better meets the patient's actual needs. If it is found that a patient's physical recovery is not good after exercising according to the current exercise plan at a specific stage of dialysis treatment, that is, the key indicators are lower than the preset threshold, the weights of the neural network model will be adjusted. When generating the exercise plan later, more attention will be paid to the characteristics related to physical recovery in this treatment cycle stage, thereby optimizing the exercise plan and improving the effect of exercise intervention.

[0070] In one embodiment, the system further includes a visualization module 106 for:

[0071] The multidimensional feature parameters in the effect evaluation dataset are integrated with the historical medical record data, and a time series correlation matrix is ​​constructed using a sliding time window algorithm.

[0072] The time series correlation matrix is ​​processed using the t-SNE dimensionality reduction algorithm and mapped to a three-dimensional space coordinate system;

[0073] Based on the three-dimensional spatial coordinates mapped by the time series correlation matrix, a visualization rendering engine is used to generate visualization images, which include dynamic heat maps, three-dimensional scatter matrix and time axis synchronization comparison chart.

[0074] Specifically, the multidimensional feature parameters in the effect evaluation dataset include information such as exercise execution indicators and vital sign change curves that reflect the effectiveness of the exercise program. Historical medical record data covers medical information such as the patient's past diagnosis results and treatment process. Combining the two can more comprehensively present the patient's health status and the changing trends in the effectiveness of exercise interventions. The sliding time window algorithm slides a fixed-length window over time series data for analysis. By continuously sliding the time window, the correlation between the data within each window is calculated, capturing the association between data at different time points. Taking heart rate data as an example, the sliding time window algorithm can analyze the correlation between heart rate changes in adjacent time periods, discover patterns in heart rate fluctuations, and potential connections with other physiological indicators or treatment stages. The time series correlation matrix contains a large number of data dimensions, making direct visualization difficult and incomprehensible. The t-SNE dimensionality reduction algorithm (t-distributed stochastic neighbor embedding algorithm) can be used to map high-dimensional data into a low-dimensional space, in this example, a three-dimensional spatial coordinate system, while preserving the relative relationships between the data, making complex high-dimensional data more intuitive. The three-dimensional coordinate data after dimensionality reduction can highlight the clustering structure and distribution characteristics of the data, so that the differences and similarities between the effects of different exercise programs, the physiological status of patients, and the treatment stages can be clearly presented. A variety of visual images are generated using a visual rendering engine, which can include: dynamic heat maps, which use color depth to show the distribution density and changes of data in different areas, reflecting the concentration trend and abnormality of patient physiological indicators under different exercise programs and different treatment stages; three-dimensional scatter matrix, which displays data points in the form of scattered points in three-dimensional space, each point represents a data sample, and by observing the distribution of scatter points, the relationship between different features and the clustering of data can be seen, which helps to discover potential patterns in the data; time axis synchronous comparison chart, combined with the time dimension, synchronously compares and displays data at different time points, allowing users to intuitively observe the changing trends of various indicators over time and the differences in the effects of different exercise programs at different times, providing a powerful tool for medical staff and patients to evaluate the long-term effects of exercise programs.

[0075] In one embodiment, the system further includes a remote monitoring module 107 for:

[0076] Synchronously collect the patient's physiological characteristic data and motion trajectory spatial coordinates during exercise to generate a monitoring data stream;

[0077] De-noise the monitoring data stream and identify abnormal physiological fluctuations and movement trajectory deviations;

[0078] When physiological fluctuations or movement trajectory deviations exceed the preset threshold, a graded alarm is triggered according to the preset alarm mechanism.

[0079] For example, physiological characteristic data can include key indicators such as heart rate, blood pressure, and blood oxygen saturation, reflecting the patient's physical function during exercise. The spatial coordinates of the motion trajectory record the patient's position changes during exercise. By synchronously collecting these two types of data, the patient's overall condition during exercise can be grasped. Combining them into a monitoring data stream provides basic data for comprehensive assessment of exercise status in multiple dimensions for subsequent analysis and processing. Denoising the monitoring data stream removes noise interference introduced during data collection, improving data accuracy and reliability, and reducing false positives. Abnormal physiological fluctuations can be identified using the same algorithm as the dynamic fluctuation detection described above, based on preset normal physiological indicator ranges and fluctuation thresholds. Motion trajectory deviations can be identified by comparing the patient's actual motion trajectory with a preset motion trajectory. The preset motion trajectory can be based on the planned exercise plan or the patient's past normal exercise patterns. When the physiological fluctuation or motion trajectory deviation exceeds a preset threshold, it indicates a potential exercise risk, triggering a graded alarm according to the preset alarm mechanism. Graded alarms are categorized by the severity of the abnormality, including mild, moderate, and severe. A mild alarm may indicate a slight abnormality in a patient's physiological indicators or movement trajectory, but it will not pose a major threat to their health for the time being. It can prompt medical staff or patients to pay attention and make appropriate adjustments to their exercise routines. A moderate alarm indicates a more obvious abnormality and requires timely action, such as pausing exercise for inspection. A severe alarm indicates an emergency situation that may endanger the patient's life and requires immediate notification to relevant medical staff for emergency treatment. The graded alarm mechanism enables relevant personnel to take appropriate response measures based on different risk levels, ensuring the safety of patients during exercise in a timely manner.

[0080] The personalized exercise generation system for dialysis patients, described above, addresses many challenges in traditional dialysis patient exercise management through the collaborative operation of multiple modules. The data acquisition and standardization module integrates and standardizes multi-source heterogeneous data to construct a structured medical dataset, overcoming the difficulty of collaborative analysis of traditional multi-source data and laying the foundation for accurate feature extraction. The feature extraction and state modeling module utilizes a multimodal feature extraction method to separate and integrate treatment cycle, physiological, and user preference features to construct a multidimensional patient state vector. This overcomes the limitations of traditional reliance on a single data dimension and enables refined modeling of individual pathological states and behavioral preferences. The intelligent exercise plan generation module leverages a pretrained neural network model to integrate multidimensional features to generate personalized exercise plans, fully accounting for individual differences and the dynamic relationship between treatment cycles, thereby enhancing the scientific validity of exercise interventions. The dynamic exercise plan adjustment module collects physiological data in real time during patient exercise, identifies anomalies using a dynamic fluctuation detection algorithm, and adjusts exercise intensity or duration based on pre-set strategies after risk assessment to ensure exercise safety. The effectiveness evaluation module processes exercise execution record data to generate exercise execution indicators and vital sign change curves, which are mapped to the exercise plan to form a multidimensional effectiveness evaluation dataset. Based on this, it selects and optimizes decision parameters, triggers the reinforcement learning model to update the neural network model weights, and continuously optimizes the exercise plan. The visualization module integrates multidimensional feature parameters with historical medical records to generate a variety of visualization images, providing medical staff and patients with intuitive basis for evaluating the effectiveness of exercise plans. The remote monitoring module simultaneously collects physiological characteristic data and motion trajectory coordinates, identifies anomalies after noise reduction processing, and issues graded alarms to further ensure patient exercise safety.

[0081] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0082] Based on the same inventive concept, the embodiments of the present application also provide a method for implementing the aforementioned personalized motion generation system for dialysis patients. The solution to the problem provided by this method is similar to the solution described in the aforementioned personalized motion generation system for dialysis patients. Therefore, the specific limitations of one or more embodiments of the personalized motion generation method for dialysis patients provided below can be found in the above-mentioned limitations of the personalized motion generation system for dialysis patients, and will not be repeated here.

[0083] In an exemplary embodiment, Figure 3 As shown, a personalized motion generation method suitable for dialysis patients is provided, comprising:

[0084] S11, acquiring a multi-source data set from medical equipment, physiological monitoring devices, and user terminals, and processing the data using a data standardization method to generate a standardized medical data set;

[0085] S12, based on the standardized medical dataset, uses a multimodal feature extraction method to separate treatment cycle features, physiological features, and user preference features to construct a patient state vector;

[0086] S13, based on the patient state vector, a pre-trained neural network model is used to fuse treatment cycle characteristics, physiological characteristics, and preference characteristics to generate an exercise plan.

[0087] In one embodiment, the method further comprises:

[0088] S21, collecting real-time physiological sign data streams of the patient during exercise execution through a physiological monitoring device;

[0089] S22, using a dynamic fluctuation detection algorithm to perform real-time analysis on the real-time physiological sign data stream to identify abnormal physiological fluctuation signals;

[0090] S23, when an abnormal physiological fluctuation signal is detected, triggering the safety constraint rule matching engine to perform risk assessment;

[0091] S24, based on the preset physiological feedback adjustment strategy, generating updated exercise instructions according to the risk assessment, and the updated exercise instructions are used to dynamically adjust the exercise intensity parameters or exercise duration parameters in the exercise plan.

[0092] In one embodiment, the method further includes: S31, constructing a dynamic fluctuation detection algorithm using the following formula:

[0093]

[0094] Among them, A(t) represents the abnormal fluctuation index, ω(τ) represents the time decay weight, represents the wavelet domain gradient operator, SWT(x(τ)) represents the stationary wavelet transform coefficient, KL(p||q) τ represents KL divergence, CUSUM k (t) represents the cumulative sum statistic, ζ represents the mutation component amplification coefficient, and Δt represents the dynamic detection window.

[0095] In one embodiment, the method further comprises:

[0096] S41, aligning the timestamps of the updated motion instructions and the real-time physiological sign data stream and extracting the motion execution record data;

[0097] S42, processing the exercise execution record data using a statistical modeling method to generate an exercise execution index and a physical sign change curve;

[0098] S43, mapping the exercise execution index, the physical sign change curve and the exercise plan to generate an effect evaluation data set containing multi-dimensional features.

[0099] In one embodiment, the method further comprises:

[0100] S51, based on the parameter association matrix, using a dynamic association analysis algorithm to calculate the association strength between each treatment cycle feature and the physical sign change curve, and screening features with association strength exceeding a preset threshold to form an optimized decision parameter set;

[0101] S52: If the key indicator in the optimized decision parameter set is lower than the preset threshold, the reinforcement learning model is triggered to update the weight parameters of the neural network model.

[0102] In one embodiment, the method further comprises:

[0103] S61, the multidimensional feature parameters in the effect evaluation dataset are integrated with the historical medical record data, and a time series correlation matrix is ​​constructed using a sliding time window algorithm;

[0104] S62, using the t-SNE dimensionality reduction algorithm to process the time series correlation matrix and map it to a three-dimensional space coordinate system;

[0105] S63, based on the three-dimensional spatial coordinates mapped by the time series correlation matrix, generates visualization images using a visualization rendering engine. The visualization images include dynamic heat maps, three-dimensional scatter matrixes, and time axis synchronization comparison maps.

[0106] In one embodiment, the method further comprises:

[0107] S71, synchronously collecting physiological characteristic data and motion trajectory spatial coordinates of the patient during exercise to generate a monitoring data stream;

[0108] S72, performing noise reduction processing on the monitoring data stream and identifying abnormal physiological fluctuations and movement trajectory deviations;

[0109] S73, when the physiological fluctuation or movement trajectory deviation exceeds the preset threshold, a graded alarm is triggered according to the preset alarm mechanism.

[0110] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the aforementioned personalized motion generation system for dialysis patients.

[0111] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0112] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial description of the method embodiments. The device embodiments described above are merely illustrative, wherein the components described as separate parts may or may not be physically separated, and the parts displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the disclosed solution. A person of ordinary skill in the art can understand and implement it without expending creative work.

[0113] The above-described embodiments merely represent several implementation methods of the embodiments of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that a person skilled in the art may make various modifications and improvements without departing from the concept of the embodiments of the present application, and these modifications and improvements fall within the scope of protection of the embodiments of the present application.

Claims

1. A personalized motion generation system for dialysis patients, characterized in that: The system comprises: The data acquisition and standardization module is used to obtain multi-source data sets from medical equipment, physiological monitoring devices and user terminals, and process them using data standardization methods to generate standardized medical data sets; a feature extraction and state modeling module, configured to separate treatment cycle features, physiological features, and user preference features using a multimodal feature extraction method based on the standardized medical dataset, and construct a patient state vector; An intelligent exercise plan generation module is used to generate an exercise plan based on the patient state vector and using a pre-trained neural network model to fuse the treatment cycle characteristics, physiological characteristics and preference characteristics.

2. The system according to claim 1, wherein: The system further includes a motion plan dynamic adjustment module, which is used to: The physiological monitoring device collects real-time physiological sign data streams of the patient during the exercise execution process; Using a dynamic fluctuation detection algorithm to perform real-time analysis on the real-time physiological sign data stream to identify abnormal physiological fluctuation signals; When abnormal physiological fluctuation signals are detected, the safety constraint rule matching engine is triggered to perform risk assessment; Based on a preset physiological feedback adjustment strategy, an updated exercise instruction is generated according to the risk assessment, and the updated exercise instruction is used to dynamically adjust the exercise intensity parameter or the exercise duration parameter in the exercise plan.

3. The system according to claim 2, characterized in that The motion scheme dynamic adjustment module is further configured to construct a dynamic fluctuation detection algorithm using the following formula: Among them, A(t) represents the abnormal fluctuation index, ω(τ) represents the time decay weight, represents the wavelet domain gradient operator, SWT(x(τ)) represents the stationary wavelet transform coefficient, KL(p||q) τ represents KL divergence, CUSUM k (t) represents the cumulative sum statistic, ζ represents the mutation component amplification coefficient, and Δt represents the dynamic detection window.

4. The system according to claim 2, wherein: The system further includes an effect evaluation module, which is used to: Performing time stamp alignment on the updated motion instructions and the real-time physiological sign data stream and extracting motion execution record data; Processing the exercise execution record data using a statistical modeling method to generate an exercise execution index and a physical sign change curve; The exercise execution index, the physical sign change curve and the exercise plan are correlated and mapped to generate an effect evaluation data set containing multi-dimensional features.

5. The system according to claim 4, characterized in that The effect evaluation module is also used to: Based on the parameter association matrix, a dynamic association analysis algorithm is used to calculate the association strength between each treatment cycle feature and the physical sign change curve, and features with an association strength exceeding a preset threshold are screened to form an optimized decision parameter set; If the key indicator in the optimization decision parameter set is lower than the preset threshold, the reinforcement learning model is triggered to update the weight parameters of the neural network model.

6. The system according to claim 4, characterized in that The system further includes a visualization module for: The multidimensional feature parameters in the effect evaluation data set are integrated with the historical medical record data, and a time series correlation matrix is ​​constructed using a sliding time window algorithm; The time series correlation matrix is ​​processed using the t-SNE dimensionality reduction algorithm and mapped to a three-dimensional space coordinate system; Based on the three-dimensional spatial coordinates mapped by the time series correlation matrix, a visualization rendering engine is used to generate a visualization image, which includes a dynamic heat map, a three-dimensional scatter matrix and a time axis synchronization comparison map.

7. The system according to claim 1, wherein: The system also includes a remote monitoring module for: Synchronously collect the patient's physiological characteristic data and motion trajectory spatial coordinates during exercise to generate a monitoring data stream; De-noising the monitoring data stream and identifying abnormal physiological fluctuations and movement trajectory deviations; When the physiological fluctuation or movement trajectory deviation exceeds a preset threshold, a graded alarm is triggered according to a preset alarm mechanism.

8. A personalized motion generation method for dialysis patients, characterized in that: The method comprises: Acquire multi-source data sets from medical equipment, physiological monitoring devices, and user terminals, and process them using data standardization methods to generate standardized medical data sets; Based on the standardized medical dataset, a multimodal feature extraction method is used to separate treatment cycle features, physiological features, and user preference features, and construct a patient state vector; According to the patient state vector, a pre-trained neural network model is used to fuse the treatment cycle characteristics, physiological characteristics and preference characteristics to generate an exercise plan.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images