Intelligent rehabilitation monitoring method and device for heart failure patient

Abstract

The invention discloses an intelligent rehabilitation monitoring method and device for a heart failure patient, and relates to the technical field of rehabilitation data processing.The method comprises the steps that on the basis of multiple collection time scales, monitoring item time sequence data sets are obtained, and the monitoring item time sequence data sets comprise the time sequence data set, corresponding to the multiple collection time scales, of each monitoring item; health feature value extraction is carried out on the time sequence data set of the monitoring items, and feature value sequences corresponding to different collection time scales are generated; constructing a multi-time-scale multivariate sequential sequence, and quantifying a physiological system elasticity coefficient value of the heart failure patient by using a preset event elasticity extraction mechanism; and inputting the multi-time-scale multivariate sequential sequence and the elastic coefficient value as monitoring features into a heart failure rehabilitation trajectory prediction model, outputting a health index sequence in a future intervention window, and performing elastic loss critical point identification as an early intervention point. Therefore, the accuracy and foresight of monitoring the rehabilitation health trend of the heart failure patient are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of rehabilitation data processing, and particularly relates to an intelligent rehabilitation monitoring method and device for heart failure patients. BACKGROUND

[0002] Heart failure is a complex clinical syndrome with high morbidity, high readmission rate and high mortality. With the acceleration of population aging, the number of heart failure patients continues to grow, which brings a heavy burden to the medical system. The rehabilitation management of heart failure patients is a long and complex process, which requires continuous monitoring of physiological index changes and timely adjustment of treatment plan.

[0003] The prior art mainly relies on a single sensor or a simple multi-sensor fusion method in the health monitoring of patient rehabilitation training, which usually leads to insufficient completeness and reliability of the monitoring data, and mostly focuses on current data acquisition and state evaluation, lacking the prediction ability of the evolution trend of heart failure pathological state. The acute decompensation of heart failure patients often has a gradual development process, and the transition from the stable period to the pre-decompensation period is usually accompanied by specific physiological mode changes. SUMMARY

[0004] The present application provides an intelligent rehabilitation monitoring method and device for heart failure patients, which improves the accuracy and forward-looking of the rehabilitation health trend monitoring of heart failure patients, and reduces the risk of acute attack.

[0005] The present application provides an intelligent rehabilitation monitoring method for heart failure patients, comprising: S101, based on a preset multi-acquisition time scale, obtaining the time series data of each monitoring item corresponding to each acquisition time scale within a preset time period for heart failure patients, to form a monitoring item time series data set, including a time series data set of each monitoring item corresponding to the multi-acquisition time scale; S102, extracting health feature values from the time series data set of each monitoring item to generate feature value sequences corresponding to different acquisition time scales respectively, the feature value sequence being composed of health feature values of each acquisition point on the corresponding acquisition time scale; S103, based on the feature value sequences of all monitoring items on each acquisition time scale, constructing a multi-time scale multivariate time series, including a multivariate time series matrix corresponding to each time scale; S104, based on the multi-time scale multivariate time series, quantifying the physiological system elasticity coefficient value of the heart failure patient by using a preset event elasticity extraction mechanism; S105, inputting the multi-time scale multivariate time series and the elasticity coefficient value as monitoring features into a pre-trained heart failure rehabilitation trajectory prediction model to output a health index sequence within a future intervention window; S106, identify the elastic loss critical point of the health index sequence as an early intervention point, and synchronize to the monitoring personnel.

[0006] Preferably, the monitoring item time series data set is represented as , P is the total number of monitoring items, is the time series data set of the Pth monitoring item, including the time series data corresponding to the preset multiple collection time scales, , K represents the number of preset collection time scales, is the time series data corresponding to the Kth collection time scale of the monitoring item.

[0007] Preferably, the preset multiple collection time scales are set to three collection time scales, which are high frequency scale , medium frequency scale , and low frequency scale ; The time series data set of each monitoring item in the monitoring item time series data set is divided into three collection time scales: high frequency physiological parameters collected at a first preset time interval, medium frequency trend parameters collected at a second preset time interval, and low frequency cumulative parameters collected at a third preset time interval, wherein the first preset time interval is less than the second preset time interval, and the second preset time interval is less than the third preset time interval.

[0008] Preferably, the health feature value extraction of the time series data set of each monitoring item includes: preprocessing all time series data, including data cleaning and normalization processing; For the time series data of each monitoring item at different collection time scales, the health feature value of each collection point data is extracted respectively to generate a feature value sequence at each collection time scale.

[0009] Preferably, the multi-time scale multivariate time series sequence includes a high frequency scale corresponding multivariate time series sequence matrix , a medium frequency scale corresponding multivariate time series sequence matrix , and a low frequency scale corresponding multivariate time series sequence matrix .

[0010] Preferably, S103 specifically includes: For each time scale , the construction method of the multivariate time series sequence matrix is: for time scale , construct dimension sequence matrix as the multivariate time series sequence matrix corresponding to the time scale:

[0011] wherein, is the time scale the length of the sequence of eigenvalues, P is the number of monitoring items, monitoring items at time health eigenvalues, ∈[ , ]。

[0012] Preferably, the preset event resilience extraction mechanism specifically includes: S201, identifying the disturbance event in the sequence of eigenvalues corresponding to the medium frequency scale, and extracting the data segment corresponding to each disturbance event in the sequence of eigenvalues of the high frequency scale; S202, calculating the health eigenvalue recovery speed based on the extracted data segment corresponding to each disturbance event in the sequence of eigenvalues of the high frequency scale, and quantifying the anti-disturbance ability of the heart failure patient; S203, calculating the physiological system resilience coefficient value based on the health eigenvalue recovery speed corresponding to the disturbance event.

[0013] Preferably, the judgment condition of the disturbance event is set as: For the sequence of eigenvalues of each monitoring item, identify all continuous intervals in which the health eigenvalue exceeds the preset health eigenvalue threshold interval as the disturbance event.

[0014] Preferably, the health eigenvalue recovery speed is calculated according to the following formula:

[0015] wherein, is the recovery rate of the i th disturbance event, is the peak value of the health eigenvalue of the disturbance event in the corresponding data segment, is the stable eigenvalue after the disturbance event in the corresponding data segment, and are and corresponding time points, respectively. The physiological system resilience coefficient value is calculated according to the following formula:

[0016] wherein, E is the resilience coefficient value, M is the total number of disturbance events, is the recovery rate of the i th disturbance event, is the maximum recovery rate corresponding to the heart failure patient.

[0017] The application also provides an intelligent rehabilitation monitoring device for heart failure patients, comprising: a multi-source sensor, a feature extraction module, and a prediction module. The multi-source sensor is configured to obtain time series data of each monitoring item corresponding to each collection time scale within a preset time period based on a preset multi-collection time scale, to form a time series data set of the monitoring item, and to include a time series data set of each monitoring item corresponding to the multi-collection time scale. The feature extraction module is configured to extract a health feature value of the time series data set of each monitoring item to generate a feature value sequence corresponding to different collection time scales, and the feature value sequence is composed of a health feature value of each collection point on the corresponding collection time scale; and to construct a multi-time scale multivariate time series based on the feature value sequence of all monitoring items on each collection time scale, and to include a multivariate time series matrix corresponding to each time scale. The prediction module is configured to quantize a physiological system elasticity coefficient value of the heart failure patient by using a preset event elasticity extraction mechanism based on the multi-time scale multivariate time series, to input the multi-time scale multivariate time series and the elasticity coefficient value as monitoring features into a pre-trained heart failure rehabilitation trajectory prediction model, to output a health index sequence within a future intervention window, and to identify an elasticity loss critical point of the health index sequence as an early intervention point and synchronize the elasticity loss critical point to a monitoring personnel.

[0018] One or more technical solutions provided in the present application have at least the following technical effects or advantages: The multi-source sensor is configured to collect physiological data of a patient at different time scales, to construct a multivariate time series after preprocessing and feature extraction, to quantize a physiological system elasticity coefficient value, to output a future health index sequence by using a prediction model, and to finally identify an elasticity loss critical point and generate an intervention suggestion, which functions to comprehensively and accurately monitor the health status of a heart failure patient, to predict a health change trend in advance, to timely find health problems of the patient and provide professional intervention guidance, and to improve the rehabilitation effect of the heart failure patient. The prediction module is configured to quantize a physiological system elasticity coefficient value of the heart failure patient by using a preset event elasticity extraction mechanism based on the multi-time scale multivariate time series, to identify a disturbance event in a medium frequency scale, to extract a data segment corresponding to a high frequency scale, to calculate a health feature value recovery speed, and to quantize the physiological system elasticity coefficient value, which can accurately evaluate the disturbance resistance ability of the heart failure patient and provide an important basis for judging the health status of the patient; wherein, the disturbance event is identified based on a feature value sequence in the medium frequency scale, the data segment corresponding to the high frequency scale is extracted, the disturbance resistance ability of the patient is quantized by calculating the health feature value recovery speed, and the physiological system elasticity coefficient value is obtained, which is a multi-scale combined elasticity quantization method that breaks through the limitations of traditional single time scale analysis and can more comprehensively and accurately evaluate the physiological system elasticity of the heart failure patient, thereby providing a unique and effective basis for judging the health status of the patient and formulating an intervention strategy. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1A flowchart of an intelligent rehabilitation monitoring method for heart failure patients according to an embodiment of the present application; Figure 2 A structural diagram of an intelligent rehabilitation monitoring device for heart failure patients according to an embodiment of the present application. DETAILED DESCRIPTION

[0020] In order to facilitate the understanding of the present application, the present application will be described more fully below with reference to the accompanying drawings, in which preferred embodiments of the present application are shown. The present application may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and fully convey the scope of the application to those skilled in the art.

[0021] It should be noted that the terms "vertical", "horizontal", "upper", "lower", "left", "right", and similar expressions used herein are for illustrative purposes only and are not intended to be limiting.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the use of the terms "and / or" includes a combination of one or more of the associated listed items.

[0023] Embodiment One: Figure 1 A flowchart of an intelligent rehabilitation monitoring method for heart failure patients according to an embodiment of the present application.

[0024] As shown in Figure 1 , an intelligent rehabilitation monitoring method for heart failure patients includes the following steps: S101, based on a plurality of preset collection time scales, obtaining the time series data of the monitoring items corresponding to each collection time scale of the heart failure patient within a preset time period, and grouping all the time series data into a monitoring item time series data set, including the time series data set of each monitoring item corresponding to the plurality of collection time scales.

[0025] Specifically, by deploying a plurality of multi-source sensors (any one of an electrocardiogram sensor, a pulse oximeter, a respiration sensor, an accelerometer, and a bioimpedance sensor), the monitoring item time series data set of the heart failure patient within the plurality of collection time scales within the preset time period is obtained , P is the total number of monitoring items, the time series data set of the Pth monitoring item, the monitoring item including any one or more of heart rate, blood oxygen saturation, respiratory rate, activity tolerance, and edema index, the time series data set of each monitoring item including time series data corresponding to a preset plurality of collection time scales, i.e., time series data corresponding to each collection time scale constitutes the time series data set of the monitoring item, K represents a number of preset collection time scales, the time series data corresponding to the Kth collection time scale of the monitoring item, i.e., time series data collected according to the collection time scale in a preset time period based on a corresponding time interval.

[0026] It should be noted that the preset time period can be set to one day, or can be dynamically adjusted according to actual conditions.

[0027] In the embodiment of the present application, the preset plurality of collection time scales is set to three collection time scales, i.e., a high-frequency scale , a medium-frequency scale , and a low-frequency scale ; the time series data set of each monitoring item in the monitoring item time series data set is divided into a high-frequency physiological parameter collected at a first preset time interval, a medium-frequency trend parameter collected at a second preset time interval, and a low-frequency cumulative parameter collected at a third preset time interval, wherein the first preset time interval is less than the second preset time interval, the second preset time interval is less than the third preset time interval, and the third preset time interval is much less than the preset time period. That is, in the preset time period, time series data of different monitoring items is periodically collected based on different collection time scales at corresponding preset time intervals, and time series data of different monitoring items corresponding to all collection time scales constitutes the monitoring item time series data set.

[0028] Exemplarily, in the three collection time scales, the first preset time interval can be set to 1 minute, the second preset time interval can be set to 1 hour, and the third preset time interval can be set to 6 hours. The specific values can be dynamically adjusted according to actual monitoring conditions and requirements to ensure that complete physiological information from transient fluctuations to long-term trends is captured. It should be noted that the difference between the parameters collected at different time scales only exists in the difference in collection frequency or collection period, so the parameters are defined as high-frequency physiological parameters, medium-frequency trend parameters, and low-frequency cumulative parameters to distinguish different collection frequencies.

[0029] S102, health feature value extraction is performed on the time series data set of each monitoring item to generate a feature value sequence corresponding to each collection time scale, the feature value sequence being composed of health feature values of each collection point on the corresponding collection time scale.

[0030] Specifically, the health feature value extraction on the time series data set of each monitoring item includes: Preprocess all time series data, including data cleaning and normalization; For each monitoring item with different collection time scales, extract the health characteristic value of each collection point data, and generate a characteristic value sequence for each collection time scale.

[0031] For any monitoring item time series data set At any collection time t, through the preset feature extraction function of the monitoring item Calculate a scalar characteristic value from its original data as the health characteristic value of the monitoring item at that time , represents the original data value of the monitoring item at time t.

[0032] It should be noted that for the monitoring item, the health characteristic value of the data value at a certain collection time (which can be understood as the data value currently monitored by the monitoring item reflecting the health condition of the heart failure patient) can be evaluated using data within a historical time window (which can be set according to the monitoring item, for example, set to 5 minutes), such as rate of change, fluctuation, change period, etc. The feature extraction method can refer to the related prior art, and the present application does not repeat it.

[0033] Exemplarily: For the heart rate monitoring item, extract the heart rate variability standard deviation as the health characteristic value, reflecting the autonomic nervous function state, detect the R wave peak value from the continuous electrocardiogram signal within the historical time window, calculate the time interval between adjacent R waves (RR interval), and calculate the standard deviation based on the RR interval sequence:

[0034] wherein, is the RR interval of the i-th sinus beat, that is, a series of consecutive heartbeat intervals extracted within the historical time window (less than a preset time period, and experts can set the historical time window size according to the characteristics of the monitoring item) of the collection time t, represents the time interval between two consecutive R waves, is the arithmetic mean of all RR intervals in the historical time window, and N is the total number of RR intervals in the historical time window, (i.e. ) is the health characteristic value of the heart rate monitoring item, reflecting the overall degree of heart rate variability.

[0035] For the blood oxygen monitoring item, extract the fluctuation range as the health characteristic value to evaluate the oxygenation stability, and calculate the range of blood oxygen saturation value within the historical time window:

[0036] wherein, S (t) is the blood oxygen saturation value at the collection time t, is a preset historical time window (e.g., the past 5 minutes), , respectively represent the maximum and minimum values of the blood oxygen saturation in the preset historical time window , (i.e. ) represents the range value in the time window, as a health feature value of the blood oxygen monitoring item, reflecting the stability of the blood oxygen level.

[0037] For the respiratory monitoring item, the dominant period is extracted by the autocorrelation function as a health feature value, quantifying the respiratory rhythm, and the autocorrelation function is used to analyze the time series data to find the most significant period component:

[0038]

[0039] wherein, is the autocorrelation function, is the i-th data in the historical time window (the respiratory frequency value at the i-th time point), is the mean value of the respiratory frequency in the historical time window, is the length of the historical time window, is the lag parameter, is a preset respiratory period search range (e.g., corresponding to a respiratory frequency of 8-30 times / minute), (i.e. ) is the lag time corresponding to the maximum absolute value of the autocorrelation function, i.e., the dominant respiratory period, as a health feature value.

[0040] For the activity endurance item (it should be noted that this monitoring item is only introduced when the heart failure patient is in a standing dynamic state), the gait cycle and walking speed are extracted as health feature values: The gait cycle during walking is detected by an acceleration sensor, and the average time interval between consecutive steps is calculated: , is the touch-down time point of the y-th step (determined by acceleration signal peak detection), Y is the total number of steps detected in the historical time window, is the average gait cycle, reflecting the stability of the walking rhythm; Combined with accelerometer and gyroscope data, the average walking speed is calculated by step count and displacement evaluation: , Y is the total number of steps in the historical time window, SL is the average step length, which is estimated by height and gait characteristics, for example, , is the total walking time, is the total walking distance, is the average walking speed, reflecting the activity endurance level, as a health feature value.

[0041] For the edema index item, the bioimpedance change rate is calculated as a health feature value, by bioimpedance spectrum measurement, the relative change rate of the impedance value at the sampling time t is calculated: , is the bioimpedance value closest to the corresponding collection time t (usually the impedance value at 50 kHz frequency), is the bioimpedance value of the previous measurement, is the difference between the two bioimpedance value collection times, is the bioimpedance change rate, a negative value indicates a decrease in impedance, indicating an increase in tissue fluid volume.

[0042] It should be noted that the feature values of all monitoring items are calculated according to the pre-set collection time scale respectively, forming a multi-time scale feature value sequence, providing comprehensive input features for subsequent health index prediction.

[0043] S103, based on the feature value sequence of all monitoring items at each time scale, a multi-time scale multivariate time series sequence is constructed, including a multivariate time series sequence matrix corresponding to a high frequency scale and a multivariate time series sequence matrix corresponding to a low frequency scale .

[0044] In some embodiments, S103 specifically includes: organizing each feature value sequence into a structured matrix, for each time scale , constructing a multivariate time series sequence matrix , where is the length of the feature value sequence of the time scale, P is the number of monitoring items, each row of the matrix corresponds to a collection point (collection time point), and each column corresponds to a feature value sequence of a monitoring item, forming a complete multivariate time series representation.

[0045] Specifically, the multivariate time series sequence matrix is constructed as follows: for a time scale , a dimensional sequence matrix is constructed as the multivariate time series sequence matrix corresponding to the time scale:

[0046] where, is the time scale the length of the sequence of feature values (i.e., the total number of collection time points in the sequence), P is the number of monitoring items, is the monitoring item At time the health feature value, i is between 1 and , .

[0047] S104, based on the multi-time scale multivariate time sequence, using the preset event elasticity extraction mechanism, quantifying the physiological system elasticity coefficient value of the heart failure patient.

[0048] In some embodiments, the preset event elasticity extraction mechanism specifically includes: S201, based on the feature value sequence corresponding to the intermediate frequency scale, identifying the disturbance event therein, and extracting the data segment corresponding to each disturbance event in the feature value sequence of the high frequency scale.

[0049] Specifically, the judgment condition of the disturbance event is set as: for the feature value sequence of each monitoring item on the intermediate frequency scale, identifying all continuous intervals of health feature values exceeding the preset health feature threshold interval (it should be noted that for different monitoring items, the corresponding health feature threshold interval is set in advance to measure the abnormal situation of the corresponding monitoring item, which is determined according to historical data and expert experience) as a disturbance event.

[0050] Specifically, the extraction method of the data segment is: for each disturbance event, a data segment composed of a first window before the disturbance event, a disturbance event period, and a second window after the disturbance event is extracted from the feature value sequence of the high frequency scale. Wherein, the first window and the second window are set according to the actual monitoring situation, the first window is smaller than the second window, and the first window needs to be greater than the first preset time interval corresponding to the high frequency scale, for example, the first window is set to 1 hour, and the second window is set to 2 hours.

[0051] S202, based on the data segment corresponding to each disturbance event in the extracted feature value sequence of the high frequency scale, calculating the health feature value recovery speed, quantifying the anti-disturbance ability of the heart failure patient:

[0052] wherein, is the recovery rate of the i-th disturbance event, is the peak feature value of the disturbance event in the corresponding data segment, is the stable feature value after the disturbance event in the corresponding data segment (i.e., the health feature value corresponding to the case where the health feature values of the disturbance event are not out of the preset health feature threshold interval for two consecutive monitoring times), and​​ respectively and corresponding time.

[0053] S203, based on the health feature value corresponding to the disturbance event recovery speed, the physiological system elasticity coefficient value is calculated, the value is larger, indicating that the elasticity is better.

[0054] Specifically, the physiological system elasticity coefficient value is calculated according to the following formula:

[0055] Wherein, E is the elasticity coefficient value, M is the total number of disturbance events, is the recovery rate of the i-th disturbance event, is the maximum recovery rate corresponding to the heart failure patient (determined based on historical data).

[0056] S105, the multivariate time series sequence of multiple time scales and the elasticity coefficient value are input into the pre-trained heart failure rehabilitation trajectory prediction model as monitoring features, and the health index sequence in the future intervention window is output.

[0057] In some embodiments, the pre-trained heart failure rehabilitation trajectory prediction model adopts a Transformer architecture, the input is a multivariate time series sequence of multiple time scales and the elasticity coefficient value E, the multi-scale time series dependency is learned through a self-attention mechanism, and the output is a health index sequence in a future intervention window (set according to actual monitoring conditions, for example, the intervention window can be set to 1 hour).

[0058] Specifically, the heart failure rehabilitation trajectory prediction model is obtained in the following manner: A1, a large amount of monitoring item time series data set in a pre-set time period of a heart failure patient is obtained, and historical monitoring features and health index sequences in a corresponding actual intervention window are obtained.

[0059] It should be noted that based on each historical monitoring feature, the health index sequence in the corresponding actual intervention window can be based on expert evaluation labeling, that is, the health index of each collection point in the future time window (monitoring in the pre-set intervention window) of the heart failure patient is evaluated (quantified to between 0 and 1, the larger the value, the better the rehabilitation state of the heart failure patient) on the health feature values of different monitoring items in the high-frequency collection time scale. The health feature values of multiple monitoring items can also be dimensionally processed (the evaluation direction of the health index is the same, for example, the larger the health feature value, the higher the health index, indicating the same direction), and a simple weighted fusion is performed to determine the weight based on the heart failure clinical guidelines. This application will not be described here.

[0060] A2, label the historical monitoring features with the health index sequence to obtain a training data set, train a pre-selected neural network structure, and continuously optimize the model to generate a final heart failure rehabilitation trajectory prediction model.

[0061] S106, identify the elastic loss critical point of the health index sequence as an early intervention point, generate a health status report based on the health feature values of each monitoring item of the heart failure patient at the current time, and synchronize it to the monitoring personnel to generate professional intervention suggestions, i.e., early prevention.

[0062] In some embodiments, the way to identify the elastic loss critical point is: Based on the health index sequence in the time sequence, the following conditions are met at the same time: Iterate through each health index in the health index sequence in turn, calculate the health index change rate, when the change rate is lower than the preset health negative change threshold (set according to historical data and expert experience, used to measure health changes, and the health negative change threshold represents the maximum bottom line of the downward trend of the health index) and the physiological system elasticity coefficient is lower than the preset elasticity threshold (set according to expert experience and physical assessment data of heart failure patients, representing the degree of anti-interference ability of the heart failure patient) at the same time, it means that this time point is the elastic loss critical point.

[0063] In summary, through multiple source sensors, based on the preset multiple collection time scales, the time series data of each monitoring item of the heart failure patient in the preset time period is obtained, the monitoring item time series data set is formed, multiple collection time scales (high frequency, medium frequency, low frequency) are set, which can fully capture the complete physiological information of the heart failure patient from transient fluctuations to long-term trends, and provide rich data for subsequent analysis; Extract the health feature values of each monitoring item time series data set to generate feature value sequences corresponding to different collection time scales, preprocess the data to ensure data quality, and extract health feature values for data of different collection time scales, which can accurately reflect the health status of each monitoring item at different time scales, providing a basis for subsequent construction of multi-time series sequences; Based on the feature value sequences of all monitoring items at each time scale, construct a multi-time scale multi-time series sequence, organize the feature value sequences into a structured matrix, form a complete multi-time series representation, and facilitate subsequent use of a preset mechanism to quantify the physiological system elasticity coefficient value and input the heart failure rehabilitation trajectory prediction model; Based on the multi-time scale multi-element time sequence, the preset event elasticity extraction mechanism is used to quantify the physiological system elasticity coefficient value of the heart failure patient, the high frequency scale corresponding data segment is extracted by identifying the medium frequency scale disturbance event, the health characteristic value recovery speed is calculated, and then the physiological system elasticity coefficient value is quantified, which can accurately evaluate the anti-disturbance ability of the heart failure patient, and provide an important basis for judging the health status of the patient; wherein, based on the medium frequency scale characteristic value sequence, the disturbance event is identified, the high frequency scale corresponding data segment is extracted, the health characteristic value recovery speed is calculated to quantify the anti-disturbance ability of the patient, and then the physiological system elasticity coefficient value is obtained. This multi-scale combined elasticity quantization method breaks through the limitation of traditional single time scale analysis, and can more comprehensively and accurately evaluate the physiological system elasticity of the heart failure patient, providing a unique and effective basis for judging the health status of the patient and formulating intervention strategies. The multi-time scale multi-element time sequence and the elasticity coefficient value are input into the pre-trained heart failure rehabilitation trajectory prediction model as monitoring features, and the health index sequence in the future intervention window is output. The model learns the multi-scale time sequence dependence relationship, can accurately predict the health index sequence of the heart failure patient in the future intervention window, and provides a reference for early intervention. The health index sequence is subjected to elasticity loss critical point identification, which is used as an early intervention point. The elasticity loss critical point is identified under specific conditions, which can timely find the time point when the patient's health status may deteriorate, and generate a report and intervention suggestions combined with the current health characteristic value, which is helpful to realize early prevention.

[0064] The technical solutions in the embodiments of the present application have at least the following technical effects or advantages: By constructing a multi-time scale data acquisition system, extracting health characteristic values with clear clinical significance, calculating the physiological system elasticity coefficient, and predicting the health trajectory based on a deep learning model, the key problems existing in traditional heart failure monitoring, such as early warning lag, high false alarm rate, and lack of forward-looking prediction ability, are effectively solved. The transition from passive monitoring to active prediction is realized. By quantifying the physiological system elasticity, the risk of heart failure decompensation can be identified early, and valuable time window is provided for clinical intervention. At the same time, the multi-scale time sequence analysis combined with the prediction model significantly improves the accuracy and forward-looking of health status evaluation, and finally achieves the goal of reducing the risk of acute attack and optimizing the effect of rehabilitation management.

[0065] By using multi-source sensors to collect physiological data of patients at different time scales, constructing multi-element time sequence after preprocessing and feature extraction, quantifying the physiological system elasticity coefficient value, and then outputting the future health index sequence by using the prediction model, and finally identifying the elasticity loss critical point and generating intervention suggestions, the effect is to comprehensively and accurately monitor the health status of heart failure patients, predict the health trend in advance, and provide professional intervention guidance in a timely manner to improve the rehabilitation effect of heart failure patients.

[0066] Further, the embodiment of the present application further provides an intelligent rehabilitation monitoring device for heart failure patients.

[0067] Figure 2 is a structural schematic diagram of the intelligent rehabilitation monitoring device for heart failure patients of the embodiment of the present application.

[0068] As shown in Figure 2 An intelligent rehabilitation monitoring device for heart failure patients comprises a multi-source sensor, a feature extraction module and a prediction module. The multi-source sensor is used to obtain time series data of monitoring items of heart failure patients in each collection time scale in a preset time period based on a preset multi-collection time scale, to form a monitoring item time series data set, and to include a time series data set of each monitoring item in the multi-collection time scale. The feature extraction module is used to extract health feature values of the time series data set of each monitoring item, to generate a feature value sequence corresponding to different collection time scales respectively, and to form the feature value sequence by health feature values of each collection point in the corresponding collection time scale; based on the feature value sequence of all monitoring items in each collection time scale, a multi-time scale multivariate time series sequence is constructed, including a multivariate time series sequence matrix corresponding to each time scale. The prediction module quantifies a physiological system elasticity coefficient value of the heart failure patient by using a preset event elasticity extraction mechanism based on the multi-time scale multivariate time series sequence; the multi-time scale multivariate time series sequence and the elasticity coefficient value are taken as monitoring features to be input into a pre-trained heart failure rehabilitation trajectory prediction model, to output a health index sequence in a future intervention window; the health index sequence is subjected to elasticity loss critical point identification, to serve as an early intervention point, and is synchronized to monitoring personnel.

[0069] It should be noted that other specific implementation contents of the embodiment of the present application can refer to the above-mentioned intelligent rehabilitation monitoring method for heart failure patients.

[0070] The above only describes the preferred embodiments of the present application and is not used to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for intelligent rehabilitation monitoring for heart failure patients, characterized in that, The method comprises: S101, based on a preset multi-acquisition time scale, obtaining time series data of each acquisition time scale corresponding to the monitoring items within a preset time period for a heart failure patient, to form a monitoring item time series data set, including a time series data set of each monitoring item corresponding to the multi-acquisition time scale; S102, extracting health feature values from the time series data set of each monitoring item to generate a feature value sequence corresponding to different acquisition time scales, the feature value sequence being composed of health feature values of each acquisition point on the corresponding acquisition time scale; S103, based on the feature value sequences of all monitoring items on each acquisition time scale, constructing a multi-time scale multivariate time series, including a multivariate time series matrix corresponding to each time scale; S104, based on the multi-time scale multivariate time series, quantifying the physiological system elasticity coefficient value of the heart failure patient by using a preset event elasticity extraction mechanism; S105, inputting the multi-time scale multivariate time series and the elasticity coefficient value as monitoring features into a pre-trained heart failure rehabilitation trajectory prediction model to output a health index sequence within a future intervention window; S106, identifying an elasticity loss critical point in the health index sequence as an early intervention point, and synchronizing to a monitoring personnel.

2. The intelligent rehabilitation monitoring method for heart failure patients as claimed in claim 1 wherein, The time-series data set of the monitoring item is represented as follows: P represents the total number of monitored items. This is the time-series dataset for the P-th monitoring item, which includes time-series data corresponding to multiple preset acquisition time scales. K represents the number of preset data acquisition time scales. For the first monitoring item Each time scale corresponds to a time series data collection.

3. The intelligent rehabilitation monitoring method for heart failure patients as claimed in claim 2, wherein, The preset multiple acquisition time scales are set as three acquisition time scales, respectively, high frequency scale , medium frequency scale , and low frequency scale ; the time series data set of each monitoring item in the monitoring item time series data set is divided into three acquisition time scales: high frequency physiological parameters acquired at a first preset time interval, medium frequency trend parameters acquired at a second preset time interval, and low frequency cumulative parameters acquired at a third preset time interval, wherein the first preset time interval is less than the second preset time interval, and the second preset time interval is less than the third preset time interval.

4. The intelligent rehabilitation monitoring method for heart failure patients as claimed in claim 3 wherein, The health feature value extraction from the time series data set of each monitoring item comprises: preprocessing all time series data, including data cleaning and normalization processing; for the time series data of each monitoring item at different acquisition time scales, health feature values of each acquisition point data are extracted respectively to generate a feature value sequence on each acquisition time scale.

5. The intelligent rehabilitation monitoring method for heart failure patients as claimed in claim 4, wherein, The multi-time scale multivariate time series include a multivariate time series matrix corresponding to a high frequency scale a multivariate time series matrix corresponding to an intermediate frequency scale and a multivariate time series matrix corresponding to a low frequency scale .

6. The smart rehabilitation monitoring method for heart failure patients as claimed in claim 5 wherein, S103 specifically comprises: For each time scale , the multi-variate time series matrix is constructed as follows: for time scale , a -dimensional sequence matrix is constructed as the multi-variate time series matrix corresponding to this time scale: ； wherein, is the eigenvalue sequence length in the time scale P is the number of monitoring items, is the monitoring item is the health eigenvalue at time , ∈[ , ].

7. The intelligent rehabilitation monitoring method for heart failure patients as claimed in claim 6 wherein, The preset event elasticity extraction mechanism specifically comprises: S201, based on the feature value sequence corresponding to the intermediate frequency scale, identifying a disturbance event therein and extracting a data segment corresponding to each disturbance event in the feature value sequence of the high frequency scale; the judgment condition of the disturbance event is set as: for the feature value sequence of each monitoring item, all continuous intervals in which the health feature values exceed a preset health feature threshold interval are identified as disturbance events; S202, based on the data segment corresponding to each disturbance event in the extracted feature value sequence of the high frequency scale, calculating a health feature value recovery speed thereof to quantify the anti-disturbance ability of the heart failure patient; S203, based on the health feature value recovery speed corresponding to the disturbance event, calculating a physiological system elasticity coefficient value.

8. The intelligent rehabilitation monitoring method for heart failure patients as claimed in claim 7, wherein, The health feature value recovery speed is calculated according to the following formula: ； wherein, is the recovery rate for the i-th disturbance event, is the peak value of the health characteristic value for the disturbance event in the corresponding data segment, is the stable characteristic value after the disturbance event in the corresponding data segment, and are, respectively, and the corresponding time instants; The physiological system elasticity coefficient value is calculated according to the following formula: ； wherein E is a value of an elasticity coefficient, M is a total number of perturbation events, is a recovery rate for the i-th perturbation event, is a maximum recovery rate corresponding to the heart failure patient.

9. An intelligent rehabilitation monitoring device for heart failure patients, applied to the intelligent rehabilitation monitoring method of any one of claims 1 to 8, characterized in that, The device comprises a multi-source sensor, a feature extraction module, and a prediction module; The multi-source sensor is used to obtain time series data of each acquisition time scale corresponding to the monitoring items within a preset time period for a heart failure patient based on a preset multi-acquisition time scale, to form a monitoring item time series data set, including a time series data set of each monitoring item corresponding to the multi-acquisition time scale; The feature extraction module is configured to extract health feature values from the time series data set of each monitoring item, to generate a feature value sequence corresponding to different collection time scales respectively, and the feature value sequence is composed of health feature values of each collection point on the corresponding collection time scale; based on the feature value sequences of all monitoring items on each collection time scale, a multi-time scale multivariate time series is constructed, including a multivariate time series matrix corresponding to each time scale; The prediction module uses a preset event elasticity extraction mechanism to quantify the physiological system elasticity coefficient value of the heart failure patient based on the multi-time scale multivariate time series; inputs the multi-time scale multivariate time series and the elasticity coefficient value as monitoring features into a pre-trained heart failure rehabilitation trajectory prediction model, and outputs a health index sequence in a future intervention window; and performs elasticity loss critical point identification on the health index sequence as an early intervention point, and synchronizes to the monitoring personnel.

Images