Community elderly weakness screening and intervention training method based on artificial intelligence

Abstract

The invention discloses a community elderly weakness screening and intervention training method based on artificial intelligence, and belongs to the technical field of artificial intelligence, and the method comprises the steps: obtaining dynamic physiological parameters and medical data, and generating a weakness evaluation parameter set; determining a target weakness level and generating a personalized training scheme in combination with the user tolerance parameter set; generating a user action quality evaluation result based on real-time action monitoring, and adjusting a training intensity parameter according to a baseline error threshold; and returning an evaluation result, updating the weakness evaluation parameter set, and generating a community-level weakness early warning signal according to a group risk threshold. The time sequence physiological data fusion and decision tree-transfer learning hybrid model of the wearable device is adopted, real-time optimization of a personalized training scheme and dynamic early warning of group weakness risks can be achieved, and screening accuracy and intervention effectiveness are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based method for screening and intervening in frailty among elderly people in a community. Background Art

[0002] With the aging of society, the number of frail patients in the intermediate stage between healthy state and requiring care is increasing. However, these patients find it difficult to detect their own frailty and often wait until they reach the stage where they need care.

[0003] Existing frailty screening and intervention for the elderly mainly rely on static assessment tools, such as the FRAIL scale and the Fried frailty phenotype; single physical fitness tests, such as grip strength and gait speed; and questionnaires to determine the level of frailty, and adopt fixed-cycle universal training programs, such as balance mat exercises and resistance band training.

[0004] This frailty screening and intervention training method relies on periodic manual retesting, resulting in delayed frailty assessments. The training program also lacks a personalized adjustment mechanism, preventing tailored training programs for varying degrees of frailty, which weakens its effectiveness. It also lacks a population-level frailty risk assessment, making it impossible to provide early warning of frailty at the community level. Summary of the Invention

[0005] To solve the above problems, the present invention provides an artificial intelligence-based method for screening and intervention training of frailty among elderly people in the community. It adopts the fusion of dynamic physiological parameters of wearable devices and historical medical data, and triggers real-time motion monitoring of personalized training programs. It can realize dynamic graded assessment of the frail status of the elderly, generate personalized training programs and optimize intensity in real time, and simultaneously trigger community-level risk warning signals, significantly improving the frailty screening capabilities and intervention training effects of the elderly in the community.

[0006] The above objectives can be achieved through the following solutions:

[0007] A method for screening and intervention training for frailty among elderly people in the community based on artificial intelligence, comprising obtaining a set of dynamic physiological parameters collected by a user's wearable device and a preset set of historical medical data to generate a dynamic data set; performing feature extraction on the dynamic data set to generate a set of frailty assessment parameters; determining a target frailty level according to the set of frailty assessment parameters and a preset frailty level classification rule; generating a personalized training plan through an intervention strategy generation model based on the target frailty level and a set of tolerance parameters preset by the user; triggering real-time motion monitoring according to the personalized training plan to generate a user motion quality evaluation result; dynamically adjusting the training intensity parameters of the personalized training plan based on the user motion quality evaluation result and a preset baseline error threshold; transmitting the user motion quality evaluation result back to an update module of the frailty assessment parameter set to generate an updated frailty assessment parameter set; generating a community-level frailty warning signal according to the updated set of frailty assessment parameters and a preset group risk threshold.

[0008] Optionally, generating a frailty assessment parameter set includes: performing a first feature screening on the dynamic data set to obtain an initial feature subset; extracting a key frailty indicator set from the initial feature subset through a preset decision tree feature selection model; and calculating a frailty score value in the frailty assessment parameter set based on the key frailty indicator set and a preset weight distribution rule.

[0009] Optionally, generating a personalized training plan includes: determining a candidate intervention pattern set based on matching the target frailty level with a preset intervention pattern library; screening a target intervention pattern from the candidate intervention pattern set through a mixed integer programming algorithm based on the cardiopulmonary function indicators in the user-preset tolerance parameter set; and generating training cycle parameters and action type sequences of the personalized training plan by combining the target intervention pattern with the gait stability indicator in the frailty assessment parameter set.

[0010] Optionally, the training cycle parameters and action type sequence for generating the personalized training plan include: if the gait stability index is lower than a preset balance imbalance threshold, inserting a balance compensation training unit into the action type sequence; when the number of executions of the balance compensation training unit reaches a preset number of corrections, updating the weight value of the gait stability index and regenerating the action type sequence.

[0011] Optionally, generating the user action quality evaluation result includes: through a preset fuzzy logic controller, real-time collection of the joint angle parameter set when the user performs the training action, and generating a action deviation coefficient; comparing the action deviation coefficient with the standard action template preset in the personalized training plan, and generating a completion score value in the user action quality evaluation result; if the completion score value is lower than the preset action qualification line, triggering the intensity adjustment instruction of the personalized training plan.

[0012] Optionally, the dynamic adjustment of the training intensity parameters of the personalized training plan includes: calculating the user's real-time metabolic load value based on the heart rate recovery rate parameter in the user's movement quality evaluation result; if the real-time metabolic load value exceeds the metabolic safety threshold preset in the personalized training plan, reducing the resistance coefficient in the training intensity parameter; after the resistance coefficient is adjusted, synchronously updating the cardiopulmonary function upper limit value in the tolerance parameter set.

[0013] Optionally, generating an updated weakness assessment parameter set includes: extracting the action completion time parameters and joint pressure distribution parameters from the user action quality evaluation results to generate a dynamic feedback data set; performing correlation analysis on the dynamic feedback data set and the weakness assessment parameter set through a preset transfer learning model to update the weight distribution ratio of the weakness score value; and re-dividing the classification boundaries of the target weakness level according to the updated weakness score value.

[0014] Optionally, the weight distribution ratio for updating the frailty score value includes: generating a muscle compensation risk signal when it is detected that the joint pressure distribution parameter continues to deviate from the preset baseline distribution range; triggering the model calibration process of the intervention strategy generation model based on the muscle compensation risk signal; and constraining the calibrated personalized training plan based on the joint pressure distribution parameter and the metabolic load value.

[0015] Optionally, generating a community-level frailty warning signal includes: obtaining an updated set of frailty assessment parameters for all users in the community to generate a group frailty index distribution map; performing spatiotemporal clustering analysis on the group frailty index distribution map to identify frailty index mutation areas; if the proportion of users in the frailty index mutation area exceeds the group risk threshold, triggering a targeted health education instruction and adjusting the warning level of the community-level frailty warning signal.

[0016] Optionally, triggering the targeted health education instruction and adjusting the warning level of the community-level frailty warning signal includes: generating a nutritional intervention recommendation parameter set based on the nutritional intake data of users in the frailty index mutation area; synchronously pushing the nutritional intervention recommendation parameter set and the updated personalized training plan to generate a joint intervention instruction; when it is detected that the execution cycle of the joint intervention instruction is completed, re-collecting the user's dynamic physiological parameter set and updating the group frailty index distribution map.

[0017] Compared with the prior art, the present invention has the following advantages:

[0018] 1. The present invention achieves dynamic closed-loop assessment optimization. By collecting dynamic physiological parameters in real time through wearable devices, dynamically extracting time series features, and updating frailty score weights in real time based on a preset historical medical data set, it solves the problems of high missed detection rates in static assessments and insufficient capture of dynamic physiological features, thus breaking through the limitations of static screening.

[0019] 2. It uses personalized intervention and safety regulation. Based on the target frailty level and the user's preset tolerance parameter set, a preset intervention strategy generation model is used to optimize the balance training target and generate a personalized training plan. According to the plan, real-time action monitoring is triggered and the user's action quality evaluation results are generated. Combining the evaluation results with the preset baseline error threshold, the intensity parameters of the training plan are dynamically adjusted in real time to achieve the coordinated optimization of safety and effectiveness. This solves the problems of homogeneity and lagging safety regulation in traditional intervention plans, and reduces the risk of sports injuries.

[0020] 3. Realize group risk warning; traditional methods lack community-level risk identification capabilities. The present invention generates an updated frailty assessment parameter set by transmitting the user action quality evaluation results back to the update module of the frailty assessment parameter set; then generates a community-level frailty warning signal based on the updated frailty assessment parameter set and the preset group risk threshold, which solves the problem of low group management efficiency of traditional solutions and optimizes resource allocation efficiency.

[0021] Other features and advantages of the present invention will be described in the following description, and in part will become apparent from the description, or will be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0023] Figure 1 It is a flowchart of an artificial intelligence-based community elderly frailty screening and intervention training method according to an embodiment of the present invention.

[0024] Figure 2 3 is a schematic diagram comparing the real-time monitoring of actions according to an embodiment of the present invention and standard actions.

[0025] Figure 3 This is a schematic diagram of the community frailty risk space release according to an embodiment of the present invention.

[0026] Figure 4 2 is a schematic diagram of spatiotemporal clustering risk analysis according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0028] Reference Figure 1 One embodiment of the present invention proposes an artificial intelligence-based method for screening and intervention training for frailty among the elderly in the community. It adopts the fusion of dynamic physiological parameters of wearable devices and historical medical data, and triggers real-time motion monitoring of personalized training programs. It can realize dynamic graded assessment of the frail status of the elderly, generate personalized training programs and optimize intensity in real time, and simultaneously trigger community-level risk warning signals, significantly improving the frailty screening capabilities and intervention training effects of the elderly in the community.

[0029] The method of this embodiment specifically includes:

[0030] Obtain a dynamic physiological parameter set collected by the user's wearable device and a preset historical medical data set to generate a dynamic data set;

[0031] Specifically, the dynamic physiological parameter set is collected in real time through smart wearable devices, including heart rate variability, three-dimensional gait data, and activity energy expenditure, reflecting the user's real-time physical condition. The dynamic data set is a multi-dimensional dynamic dataset generated by integrating a time-series data matrix of real-time physiological status, pathological history, and behavioral patterns, providing a holistic analytical foundation for frailty assessment and personalized intervention.

[0032] performing feature extraction on the dynamic data set to generate a frailty assessment parameter set;

[0033] Specifically, the frailty assessment parameter set is an integration of multi-dimensional indicators that integrate physiological function parameters, metabolic and cardiovascular parameters, and clinical correlation parameters obtained by feature extraction, and is used to quantify the user's frailty status.

[0034] Determining a target frailty level according to the frailty assessment parameter set and a preset frailty level classification rule;

[0035] Specifically, the frailty level is a frailty score generated by a multi-classification SVM model based on the user's frailty assessment parameter set, combined with the frailty degree classification obtained by the preset frailty level classification rules, providing a clear basis for personalized intervention.

[0036] Based on the target frailty level and a set of tolerance parameters preset by the user, a personalized training plan is generated through an intervention strategy generation model;

[0037] Specifically, the intervention strategy generation model uses mixed integer programming combined with a multi-objective genetic algorithm to generate personalized training plans based on the target frailty level and tolerance parameters, tailored to the Fried frailty phenotype, ensuring maximum training effects while minimizing risks.

[0038] triggering real-time action monitoring according to the personalized training plan and generating a user action quality evaluation result;

[0039] Specifically, real-time motion monitoring uses wearable IMU sensors to capture limb vectors, target center of gravity, and movement rhythm, and calculates the motion completion score. Figure 2 As shown, the user action quality evaluation result is obtained by comparing the preset user action quality evaluation standard.

[0040] Dynamically adjusting the training intensity parameters of the personalized training program based on the user action quality evaluation result and a preset baseline error threshold;

[0041] Specifically, the movement quality score is divided into grade thresholds: if the movement quality score is greater than or equal to 85, the user is in a low-risk zone for frailty, and the training intensity is allowed to increase gradually, with the training intensity increasing by 5% or the duration increasing by 10% each week; if the movement quality score is between 75 and 85, the user is in a frailty warning zone and needs to maintain the current training intensity and current parameters, and the training intensity upgrade will be triggered after the cumulative number of times the standard is met is ≥3 times; if the movement quality score is less than 75, the user is in a high-risk zone and needs to be forced to downgrade the training.

[0042] Transmitting the user action quality evaluation result back to the weakness assessment parameter set update module to generate an updated weakness assessment parameter set;

[0043] Specifically, a standardized dynamic feedback dataset was constructed by extracting the movement completion time deviation rate and joint pressure distribution parameters in the user movement quality evaluation; based on the transfer learning model, the data distribution differences were analyzed and the loss function was optimized, and the weight distribution ratio of the frailty score features was dynamically adjusted; finally, the fuzzy clustering algorithm was used to update the class center position, and the hyperplane translation strategy of the support vector machine was combined to reconstruct the frailty risk classification boundary, realizing dynamic adaptation of the evaluation parameters and accurate level division.

[0044] Generating a community-level frailty warning signal based on the updated frailty assessment parameter set and a preset group risk threshold;

[0045] Specifically, the proportion of high-risk users is obtained through the updated frailty assessment parameter set, combined with the average value of the community frailty assessment parameter, and compared with the preset group risk threshold to obtain the community frailty risk space release, such as Figure 3 , identify community-level early warning signs of frailty, such as Figure 4 As shown, the location in the target circle is the location where the community-level frailty warning signal is generated.

[0046] Optionally, the step of determining the target frailty level according to the frailty assessment parameter set and a preset frailty level classification rule includes:

[0047] Performing a first feature screening on the dynamic data set to obtain an initial feature subset;

[0048] Specifically, feature screening was performed on a dynamic data set containing dynamic physiological parameters from wearable devices and medical history data. This feature screening used the Pearson correlation coefficient method to calculate the correlation between each feature parameter and the pre-set frailty assessment criteria. Parameters with a correlation coefficient greater than 0.3 were retained to form the initial feature subset.

[0049] A set of key frailty indicators is extracted from the initial feature subset through a preset decision tree feature selection model;

[0050] Specifically, the initial feature subset is input into a preset screening model - the model consists of a two-layer neural network, the first layer is a 64-node fully connected layer, and the second layer is a 6-node feature compression layer - the model learns the nonlinear relationship between features through the back propagation algorithm, and outputs a set of three key frailty indicators including gait stability, resting heart rate oscillation amplitude, and electromyographic signal entropy value.

[0051] According to the key frailty indicator set and the preset weight distribution rules, the frailty score value in the frailty assessment parameter set is calculated.

[0052] Specifically, according to the weight distribution rules determined by the doctor consensus meeting, the frailty score is calculated by weighting the gait index (40%), the heart rate index (35%), and the electromyography index (25%). , where GSI is the ratio of the standard deviation of the gait cycle to the mean, HRS is the range of the difference between adjacent RR intervals in the resting state, and MSE is the sample entropy of the electromyographic signal.

[0053] For example, the dynamic data of a 65-year-old user includes gait sequences recorded by an accelerometer and electrocardiogram (ECG) monitoring data. The system first calculates three parameters: stride dispersion of 0.15 (correlation coefficient with clinical frailty criteria of 0.4), heart rate variability of 28 (correlation coefficient of 0.35), and electromyographic signal periodicity deviation of 0.22 (correlation coefficient of 0.25) to form the initial feature subset. After training, the screening model identifies gait periodicity variability (GSI) of 0.18, resting heart rate oscillation amplitude (HRS) of 32, and electromyographic signal entropy (MSE) of 0.25. Using the weighted formula, the frailty score (FS) of 0.169 is calculated. At this point, a threshold of 0.15 corresponds to mild frailty. When the user's GSI drops to 0.12 after training intervention, the system automatically updates the FS to 0.129, achieving dynamic downgrade of the frailty level.

[0054] Optionally, the steps of generating a personalized training program through an intervention strategy based on the target frailty level and a set of tolerance parameters preset by the user include:

[0055] Based on the target frailty level, a preset intervention mode library is matched to determine a candidate intervention mode set; based on the cardiopulmonary function indicators in the user's tolerance parameter set, a mixed integer programming algorithm is used to screen the target intervention mode from the candidate intervention mode set;

[0056] Specifically, pattern matching is performed in a preset intervention pattern library based on the target frailty level. The intervention pattern library is divided into a three-level storage structure according to the frailty level, where mild frailty corresponds to pattern codes L1-L3, moderate frailty corresponds to M1-M5, and severe frailty corresponds to S1-S4. Each pattern code contains a basic energy consumption value EB and an action complexity coefficient CA (a normalized value of 0-1). After the user's target frailty level is input into the pattern matching module, all pattern codes matching the level are extracted to form a candidate intervention pattern set. By loading the cardiopulmonary function indicators in the tolerance parameter set preset by the user, including peak oxygen uptake and resting heart rate HRrest. Set the filter conditions to And HRrest is less than or equal to the baseline heart rate threshold HRb. The expected oxygen uptake is calculated based on age, gender and weight, and HRb takes the 75th percentile value of the user's age group. Each mode code in the candidate intervention mode set is conditionally judged, and only those that meet the requirements at the same time are retained. and items with CA ≤ 0.5 × (1-HRrest / HRb) were used as target intervention patterns.

[0057] The target intervention mode is combined with the gait stability index in the frailty assessment parameter set to generate the training cycle parameters and action type sequence of the personalized training program.

[0058] Specifically, the training parameters are generated according to the following rules by combining the muscle group activation sequence parameters recorded in the target intervention mode with the gait stability index GS in the weakness assessment parameter set: training cycle length ,in The maximum complexity coefficient in the selected mode; the action type sequence is based on a three-day cycle unit, alternating the strength training and balance training action types in the target mode. When GS exceeds 0.25, additional posture control training is inserted into the sequence.

[0059] For example, a moderately frail user's VO2peak is measured at 23 mL / kg / min, HRrest is 78 bpm, and the frailty assessment results in a gait stability index GS=0.28. Pattern matching obtains five candidate patterns M1-M5, whose EB values ​​are 2.1, 2.5, 3.0, 3.2, and 3.8 MET, respectively. First, screen and retain Therefore, all candidate sets meet the conditions; retain Pattern and . Combined When the threshold exceeds 0.25, the training cycle parameter TC = 2.32 weeks is generated, which is rounded up to 2 weeks; posture control training is inserted every two days in the action sequence, forming a three-day cycle unit including strength training, balance training, and posture control.

[0060] Optionally, generating a training cycle parameter and action type sequence for a personalized training plan further includes:

[0061] If the gait stability index is lower than the preset balance imbalance threshold, a balance compensation training unit is inserted into the action type sequence.

[0062] Specifically, when the gait stability index GS is calculated from the gait cycle data collected by the inertial sensor, it is specifically the ratio of the standard deviation of the single-foot support period in 10 consecutive gait cycles to its mean. The balance imbalance threshold is set to 0.3. When GS exceeds this threshold, the system automatically starts the balance compensation training unit insertion mechanism. The unit insertion mechanism rule is to determine the arrangement position of the strength training module in the original action type sequence, and insert the balance compensation training unit after each strength training module. The insertion frequency is determined by the correction coefficient F. , where floor is a rounding function. The balance compensation training unit includes two modes: lateral stride training and heel-toe walking training. The duration of each unit is minute.

[0063] When the execution times of the balance compensation training unit reaches the preset correction times, the weight value of the gait stability index is updated and the action type sequence is regenerated.

[0064] Specifically, when the cumulative number of executions of the balance compensation training unit N in the action type sequence reaches the preset number of corrections (The default value is 8 times), the system automatically starts the weight value update process. When updating the weight, first calculate the historical GS fluctuation value ΔG as the difference between the maximum and minimum GS values ​​in the last N records. (Initial value 0.4) Adjust according to the comparison result of ΔG and 0.2, the new weight , where tanh is the hyperbolic tangent function. The updated weight value must not exceed the range constraint of 0.2-0.6, and finally the action type sequence is regenerated based on the new weight value.

[0065] For example, a user's first evaluation GS value is 0.31, triggering F=4 balance compensation training insertions. During the 8 consecutive training sessions, the recorded GS value fluctuates between 0.28-0.33, with ΔG=0.05. Due to the lower bound constraint, the weight is reset to 0.2. Based on the new weight, the system reduces the proportion of gait stability in the overall weakness assessment and generates a new action sequence to reduce the frequency of balance compensation training to F = 3 times.

[0066] Optionally, triggering real-time motion monitoring according to the personalized training plan and generating user motion quality evaluation results may include:

[0067] Through the preset fuzzy logic controller, the joint angle parameter set of the user when performing training movements is collected in real time to generate the movement deviation coefficient;

[0068] Specifically, a wearable motion capture device is used to collect nine-axis inertial measurement data in real time while the user performs training movements. The device is attached to the shoulder, elbow, hip, and knee joints according to standard human anatomical positions. The processor performs Kalman filtering on the original signal to reduce noise and decompose it to obtain a set of real-time angle parameters for each joint in the sagittal, coronal, and horizontal planes. , where i=1~6 corresponds to the left shoulder, right shoulder, left elbow, right elbow, left knee, and right knee three-axis angles. Take the ideal angle at the corresponding time point in the standard action template As a benchmark, the actual angle is collected every 0.1 seconds within a single motion cycle. For each joint, calculate the root mean square value of the angle deviation of the three planes , then the overall action deviation coefficient , weight They are 0.25 for shoulder, 0.2 for elbow, 0.3 for hip and 0.25 for knee.

[0069] The action deviation coefficient is compared with the standard action template preset in the personalized training plan to generate a completion score value in the user action quality evaluation result; if the completion score value is lower than the preset action passing line, the intensity adjustment instruction of the personalized training plan is triggered.

[0070] Specifically, the calculation formula for the completion score Q is: , set the action passing line =75. When the arithmetic mean of the Q value of three consecutive training cycles is lower than 75, the system starts the intensity adjustment algorithm: determine the current training intensity level , the target intensity Set to ,in is the preset minimum intensity level, and max is the maximum function of the two. Meanwhile, in the next cycle, the number of standard action video demonstrations will be increased to twice the original plan.

[0071] Optionally, dynamically adjusting the training intensity parameters of the personalized training program includes:

[0072] The user's real-time metabolic load value is calculated based on the heart rate recovery rate parameter in the user's movement quality evaluation result; if the real-time metabolic load value exceeds the metabolic safety threshold preset in the personalized training plan, the resistance coefficient in the training intensity parameter is reduced; after the resistance coefficient is adjusted, the cardiopulmonary function upper limit value in the tolerance parameter set is synchronously updated.

[0073] Specifically, a heart rate belt is used to continuously monitor the user's heart rate drop curve after exercise and calculate the heart rate recovery rate parameter ,in It is defined as the ratio between the time from the peak heart rate of exercise to the resting heart rate and the corresponding heart rate drop difference. ,have: ,in is the standard recovery rate benchmark value based on the user's age, The age correction factor is obtained by querying the preset age segment table, with 1.0 as the benchmark for those under 40 years old, and decreases by 0.08 for every 5 years of age.

[0074] Setting metabolic safety thresholds When the ML calculated in real time exceeds the threshold, the intensity is adjusted and the new training intensity is ,have: ,in , and the adjusted Must not fall below the preset minimum intensity level =3.

[0075] Optionally, transmitting the user action quality evaluation result back to the frailty assessment parameter set updating module to generate an updated frailty assessment parameter set includes:

[0076] Extract the action completion time parameters and joint pressure distribution parameters from the user action quality evaluation results to generate a dynamic feedback data set;

[0077] Specifically, the pressure distribution matrix of the user's foot during training is collected through pressure sensing insoles and wearable timers. (i=1-8 corresponds to the forefoot to heel area, j is the pressure value) and the time it takes to complete the action The dynamic feedback dataset is constructed as a five-dimensional vector ,in is the standard completion time ratio ( is the health benchmark value), is the maximum pressure area value, is the sum of pressure distribution deviations, is the pressure symmetry index of the left and right feet, is the coefficient of variation of five consecutive training sessions.

[0078] Through a preset transfer learning model, the dynamic feedback data set is associated with the frailty assessment parameter set and analyzed to update the weight distribution ratio of the frailty score value;

[0079] Specifically, the weight distribution ratio is updated using a sliding window mechanism, and the window size is the most recent N=5 training records. Calculate the Pearson correlation coefficient between each dimension in the dynamic feedback dataset D and the baseline frailty score FS (k=1-5), new weight ,in is the original weight, is the Pearson correlation coefficient between the mth dimension in the dynamic feedback dataset D and the baseline frailty score FS, is the original weight of the mth dimension.

[0080] The classification boundary of the target frailty level is re-divided according to the updated frailty score value.

[0081] Specifically, for the frailty score ,have: ,in The min-max normalized value of each dimension of the dynamic feedback dataset D is used. The classification boundary update adopts the binary relocation method to reposition the original three level boundary points B1 and B2 according to the current user group. The 25% and 75% percentile values ​​of the distribution are dynamically adjusted to ensure a balanced proportion of users at all levels.

[0082] For example, the maximum foot pressure recorded by a user during five training sessions is 280kPa (standard 250kPa), and the time is =1.15, pressure deviation =85, symmetry index =0.72, and the temporal coefficient of variation T_var=0.18. The correlations between each dimension and FS were calculated as R=[0.55, 0.68, 0.42, 0.37, 0.61]. The original weights W=[0.3, 0.25, 0.2, 0.15, 0.1]. After calculation, W'=[0.237, 0.213, 0.105, 0.069, 0.076], normalized to [0.33, 0.30, 0.15, 0.10, 0.11]. The new frailty score increases the weights of time and stress, raising the user's FS_new from 62 to 68, moving their level from Class 2 to Class 3, triggering a more intensive intervention.

[0083] Optionally, updating the weight distribution ratio of the frailty score value includes:

[0084] When the joint pressure distribution parameter is detected to continuously deviate from the preset baseline distribution range, a muscle compensation risk signal is generated;

[0085] Specifically, the pressure values ​​of the metatarsal area, arch area and calcaneus area are collected through the plantar pressure sensor array to form a three-dimensional pressure distribution curve. The baseline distribution range is defined as the spatial envelope of pressure distribution under similar movements in healthy elderly people. When the pressure value of a certain area is detected to have exceeded the baseline envelope boundary for five consecutive movement cycles, and the duration accounts for more than 30% of the total training time, the system determines that it is in a continuous deviation state. The probability density function comparison method is used to calculate the overlap DS value between the actual pressure distribution and the baseline distribution. When DS < 0.6 for 3 minutes, a level I muscle compensation risk signal is triggered; when DS < 0.4 for 5 minutes, it is upgraded to a level II muscle compensation risk signal.

[0086] A model calibration process for triggering the intervention strategy generation model based on the muscle compensation risk signal;

[0087] Specifically, when the muscle compensation risk signal is triggered, the intensity increase module of the current exercise mode is immediately stopped and the historical metabolic load peak data is called. , the dynamic relationship between joint pressure P and ML is established through linear regression: , where k is the proportional coefficient and b is the compensation value. The constraint optimization algorithm is used to recalculate the allowable load range of each action unit to ensure and .

[0088] The calibrated personalized training program is constrained based on the joint pressure distribution parameters and the metabolic load value.

[0089] Specifically, establish the joint constraint equation ,in is the maximum plantar pressure value, i.e., the joint pressure distribution parameter, ML is the metabolic load value, and the weight coefficient =0.6, =0.4, is the safety threshold.

[0090] Optionally, generating a community-level frailty warning signal further includes:

[0091] Obtain the updated frailty assessment parameter set for all users in the community and generate a group frailty index distribution graph;

[0092] Specifically, the user's geographic location with evaluation timestamp Association, build space-time cube, divide the community into grid units and time window The population decline index for each unit Defined as: ,in is the number of users in the unit, and the missing units are filled by cubic spline interpolation of adjacent space-time grids to form a continuous exponential distribution covering the entire community.

[0093] Performing spatiotemporal cluster analysis on the frailty index distribution map of the group to identify frailty index mutation areas;

[0094] Specifically, the density peak algorithm is used to identify high-density clusters, and for each cluster Calculate the average frailty index , and compared to the community's historical baseline , calculate the mutation intensity ,like The mutation region is marked is the historical standard deviation of the community, the threshold Determined by the significance level.

[0095] If the proportion of users in the frailty index mutation area exceeds the group risk threshold, a targeted health education instruction is triggered and the warning level of the community-level frailty warning signal is adjusted.

[0096] Specifically, the proportion of risky users is obtained by comparing the total number of users covered in the mutation area set with the total number of users in the community. , compared to the preset group risk threshold :If less than or equal to No warning; if greater than Less than or equal to A low-risk warning is triggered and regular health reminders are given; if it is greater than A high-risk warning will be triggered and medical resources will be coordinated to intervene.

[0097] Optionally, triggering a targeted health education instruction and adjusting the warning level of the community-level frailty warning signal further includes:

[0098] generating a nutritional intervention recommendation parameter set based on the nutritional intake data of users in the frailty index mutation area;

[0099] Specifically, the nutritional intake of the past seven days was obtained through the user's diet record APP, and compared with the recommended values ​​in the "Dietary Guidelines for Chinese Elderly People", and the deficiency was defined as: ,in is the deficiency ratio of the jth nutrient; Recommended daily intake; The actual daily intake of the user. The weight is set according to the correlation between nutrients and frailty , calculate the comprehensive intervention score: , prioritize the nutrients and increase the daily intake by 15% based on the nutrient deficiency and nutrient priority.

[0100] Synchronously push the nutritional intervention recommendation parameter set and the updated personalized training plan to generate a joint intervention instruction;

[0101] Specifically, the optimal nutrient intake period is extracted from the knowledge base and combined with the user's personalized training plan to generate joint timing instructions: ,in is the start time of the kth training; It is a nutritional replenishment window of 30 minutes to 2 hours after training.

[0102] When it is detected that the execution cycle of the joint intervention instruction is completed, the dynamic physiological parameter set of the user is re-collected and the group frailty index distribution map is updated.

[0103] Specifically, after the execution cycle ends, the wearable device collects dynamic physiological parameters such as the user's resting heart rate variability (HRV) and the modified SPPB physical fitness score. The new data set is input into the transfer learning model to update the individual frailty index: ,in is the improvement degree of HRV index, is the individual baseline value. The grid unit group frailty index is recalculated based on the updated frailty index. , lift the warning or maintain intervention in the mutation area.

[0104] For example, in a community in District D, 12 users were found to have combined vitamin D and protein deficiency during the early warning period. The system obtained their average daily vitamin D intake of 3.2μg (lower than the recommended value of 10μg) and protein of 48g (recommended value of 60g). The calculated deficiency ratios were 68% for vitamin D and 20% for protein, and the comprehensive score was The generated recommendation was to add 400 IU (equivalent to 10 μg) of vitamin D supplements daily and a high-protein snack (providing 12g of protein). The combined instruction was set to perform 30 minutes of balance training at 10:00 AM daily, followed by nutritional supplements at 11:00 AM. After 28 days of implementation, the user's average HRV value increased from 52ms to 68ms, and the system lowered the D-zone group frailty index from 0.71 to 0.63, removing the red alert. This method addresses the pain point of limited effectiveness of single rehabilitation methods by precisely matching nutrients with training periods. A dynamic monitoring data-driven scoring update mechanism ensures that group risk assessments reflect intervention effectiveness in real time. The combined instruction push not only improves convenience for seniors but also enhances nutrient absorption through temporal correlation, forming a quantifiable frailty reversal loop.

[0105] That is, any equivalent changes and modifications made according to the teachings of the present invention are still within the scope of the present invention. Those skilled in the art will readily conceive of other embodiments of the present invention after considering the disclosure of the specification and practical truths. This application is intended to cover any variations, uses, or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary technical means in the art not described herein.

Claims

1. An artificial intelligence-based method for screening and intervention training for frailty in the elderly in the community, characterized by: The method comprises: Obtain a dynamic physiological parameter set collected by the user's wearable device and a preset historical medical data set to generate a dynamic data set; performing feature extraction on the dynamic data set to generate a frailty assessment parameter set; Determining a target frailty level according to the frailty assessment parameter set and a preset frailty level classification rule; generating a personalized training program based on the target frailty level and a set of tolerance parameters preset by the user; generating a user movement quality evaluation result based on real-time movement monitoring of the personalized training program; Adjusting the training intensity parameters of the personalized training program based on the user action quality evaluation result and a preset baseline error threshold; Returning the user action quality evaluation result and updating the weakness assessment parameter set to generate an updated weakness assessment parameter set; A community-level frailty warning signal is generated based on the updated frailty assessment parameter set and the preset group risk threshold.

2. The artificial intelligence-based frailty screening and intervention training method for community elderly people according to claim 1, characterized in that: The step of generating a set of frailty assessment parameters comprises: Performing feature screening on the dynamic data set to obtain an initial feature subset; Extracting a set of key frailty indicators from the initial feature subset using a preset screening model; The frailty score value in the frailty assessment parameter set is calculated according to the key frailty indicator set and a preset weight allocation rule.

3. The artificial intelligence-based community elderly frailty screening and intervention training method according to claim 1 is characterized in that: The step of generating a personalized training plan includes: Determining a candidate intervention pattern set based on matching the target frailty level with a preset intervention pattern library; Screening a target intervention mode according to the cardiopulmonary function indicators in the tolerance parameter set preset by the user; The target intervention mode is combined with the gait stability index in the frailty assessment parameter set to generate training cycle parameters and action type sequences of the personalized training program.

4. The artificial intelligence-based community elderly frailty screening and intervention training method according to claim 3 is characterized in that: The training cycle parameters and action type sequence for generating the personalized training program include: If the gait stability index is lower than a preset balance imbalance threshold, inserting a balance compensation training unit into the action type sequence; When the execution times of the balance compensation training unit reaches a preset correction times, the weight value of the gait stability index is updated and the action type sequence is regenerated.

5. The artificial intelligence-based community elderly frailty screening and intervention training method according to claim 1 is characterized in that: The step of generating the user action quality evaluation result includes: The joint angle parameter set of the user when performing training movements is collected in real time to generate movement deviation coefficients; Comparing the movement deviation coefficient with the standard movement template preset in the personalized training program to generate a completion score value in the user movement quality evaluation result; If the completion score is lower than the preset action passing line, the intensity of the personalized training program is adjusted.

6. The artificial intelligence-based frailty screening and intervention training method for community elderly people according to claim 1 is characterized in that: The step of dynamically adjusting the training intensity parameters of the personalized training program includes: generating a real-time metabolic load value of the user according to the heart rate recovery rate parameter in the user movement quality evaluation result; If the real-time metabolic load value exceeds the metabolic safety threshold preset in the personalized training program, reducing the training intensity parameter; After the training intensity is adjusted, the upper limit value of the cardiopulmonary function in the tolerance parameter set is synchronously updated.

7. The artificial intelligence-based community elderly frailty screening and intervention training method according to claim 6 is characterized in that: The step of generating an updated frailty assessment parameter set comprises: Extracting action completion time parameters and joint pressure distribution parameters from the user action quality evaluation results to generate a dynamic feedback data set; Performing correlation analysis on the dynamic feedback data set and the frailty assessment parameter set, and updating the weight distribution ratio of the frailty score value; The classification boundary of the target frailty level is re-divided according to the updated frailty score value.

8. The artificial intelligence-based frailty screening and intervention training method for community elderly people according to claim 7, characterized in that: The method further comprises: When it is detected that the joint pressure distribution parameter continuously deviates from the preset baseline distribution range, a muscle compensation risk signal is generated; Calibrate and generate a personalized training plan based on the muscle compensation risk signal; The calibrated personalized training program is constrained based on the joint pressure distribution parameters and the metabolic load value.

9. The artificial intelligence-based frailty screening and intervention training method for community elderly people according to claim 1, characterized in that: Generating a community-level frailty warning signal includes: Obtain the updated frailty assessment parameter set for all users in the community and generate a group frailty index distribution graph; Analyzing the frailty index distribution map of the group to identify frailty index mutation areas; If the proportion of users in the frailty index mutation area exceeds the group risk threshold, targeted health education is carried out and the warning level of the community-level frailty warning signal is adjusted.

10. The artificial intelligence-based frailty screening and intervention training method for community elderly people according to claim 9, characterized in that: The method further comprises: generating a nutritional intervention recommendation parameter set based on the nutritional intake data of users in the frailty index mutation area; Synchronously push the nutritional intervention recommendation parameter set and the updated personalized training plan to generate a joint intervention instruction; When it is detected that the execution cycle of the joint intervention instruction is completed, the dynamic physiological parameter set of the user is re-collected and the group frailty index distribution map is updated.

Images