A respiratory patient rehabilitation assessment system

Abstract

The present invention relates to the technical field of intelligent assessment of respiratory rehabilitation, and specifically discloses a respiratory patient rehabilitation assessment system. By real-time monitoring of the differential oscillation frequency and resistance load data of the patient during respiratory training, the system calculates the airway response characteristic value reflecting airway adaptability and the respiratory muscle endurance characteristic value reflecting the functional state of the respiratory muscles, and constructs a comprehensive respiratory rehabilitation characteristic vector input trained random forest model to achieve intelligent assessment of the patient's rehabilitation level. Based on the assessment results, the training parameters are dynamically adjusted to form a closed-loop feedback mechanism, thereby improving the personalization and effectiveness of rehabilitation training. The present invention integrates multidimensional physiological information with machine learning technology, solves the problems of strong subjectivity and poor real-time performance of traditional assessment methods, and provides scientific and intelligent technical support for respiratory rehabilitation.

Description

Technical Field

[0001] The present invention relates to the technical field of respiratory rehabilitation intelligent assessment, and in particular to a respiratory patient rehabilitation assessment system. Background Art

[0002] Real-time monitoring and comprehensive assessment of a patient's airway responsiveness and respiratory muscle endurance during respiratory rehabilitation training is a key technical challenge facing current clinical rehabilitation programs. Traditional rehabilitation assessment systems rely primarily on subjective physician judgment, lung function testing, or analysis of single physiological parameters. These systems struggle to fully reflect a patient's true adaptability during dynamic training. This is particularly true given the large individual variability and frequent changes in training load. The lack of effective means for simultaneous quantitative assessment of airway sensitivity and muscle fatigue makes it difficult to personalize and dynamically optimize training programs.

[0003] In the existing technology, although some studies have attempted to introduce signal processing and machine learning systems for respiratory function assessment, there are generally problems such as a single feature extraction dimension, insufficient model generalization ability, and lack of closed-loop feedback mechanism. For example, some systems only judge the fatigue state of respiratory muscles based on time domain statistics or spectral peaks, ignoring the dynamic changes in the distribution of signal energy in different frequency bands; other methods use artificial intelligence models for rehabilitation prediction, but the input features lack physiological interpretability and cannot support real-time reverse regulation of training parameters. Therefore, how to build a respiratory rehabilitation assessment system that can integrate multimodal physiological information, has highly robust feature extraction capabilities and intelligent decision-making mechanisms has become a technical bottleneck that urgently needs to be solved.

[0004] In response to the above-mentioned problems, the present invention proposes a comprehensive evaluation system that integrates differential oscillation frequency stability analysis, wavelet packet energy entropy ratio calculation and random forest model prediction, realizing the simultaneous quantitative identification of airway response and respiratory muscle endurance. Furthermore, through a closed-loop control mechanism driven by rehabilitation scoring, it effectively solves the shortcomings of existing technologies in terms of weak feature expression ability, strong subjectivity in evaluation, and fixed training strategies, providing a new technical path for the intelligent upgrade of personalized respiratory rehabilitation systems. Summary of the Invention

[0005] The object of the present invention is to provide a respiratory patient rehabilitation assessment system to solve the above-mentioned problems.

[0006] The purpose of the present invention can be achieved through the following technical solutions:

[0007] A respiratory patient rehabilitation assessment system, comprising:

[0008] A real-time data monitoring module, which monitors the patient's differential oscillation frequency and resistance load data in real time during the patient's breathing training;

[0009] An airway response characteristic evaluation module, which analyzes the patient's differential oscillation frequency data and calculates an airway response characteristic value based on the stability of frequency changes to evaluate the patient's respiratory system's adaptation to training;

[0010] a respiratory muscle endurance characteristic evaluation module, which analyzes the patient's resistance load data and calculates a respiratory muscle endurance characteristic value based on the load change trend to determine whether the patient's respiratory muscles have been strengthened by training;

[0011] A rehabilitation feature fusion and intelligent assessment module, which constructs a comprehensive respiratory rehabilitation feature vector from the patient's airway response characteristic values ​​and respiratory muscle endurance characteristic values, inputs this vector into a pre-trained artificial intelligence model for fusion analysis, and assesses the patient's current rehabilitation level;

[0012] A personalized parameter dynamic control module adjusts the patient's training parameters, including oscillation frequency and resistance load, according to the rehabilitation level, to achieve dynamic optimization of the personalized rehabilitation program and improve the rehabilitation effect.

[0013] As a further embodiment of the present invention, the evaluation of the patient's respiratory system's adaptability to training specifically includes:

[0014] During the patient's breathing training, the patient's differential oscillation frequency data is monitored in real time, and the patient's differential oscillation frequency data is analyzed. Based on the stability of the frequency change, the airway response characteristic value is calculated to determine whether the patient's airway response characteristic value is greater than or equal to the preset threshold. If so, the patient's respiratory system is not adapted to the training; if not, the patient's respiratory system is adapted to the training.

[0015] As a further solution of the present invention: the process of obtaining the airway response characteristic value is:

[0016] Collect the patient's real-time differential oscillation frequency data during breathing training to form a time series signal;

[0017] Divide the time series into multiple blocks of length , and perform mean removal on each subsequence to obtain a decentralized subsequence. The mean removal is the value of the sampling point in each subsequence minus the average value of all sampling points in the corresponding subsequence.

[0018] Construct the covariance matrix based on all subsequences;

[0019] Perform eigenvalue decomposition on the covariance matrix to obtain all its eigenvalues;

[0020] Based on random matrix theory, a Marchenko-Pastur distribution model is established under a noise background, the probability density function is calculated, and the maximum theoretical eigenvalue and minimum theoretical eigenvalue of the Marchenko-Pastur distribution model are obtained;

[0021] The difference between the maximum theoretical eigenvalue and the minimum theoretical eigenvalue is calculated to obtain the eigenvalue extreme difference, the average value of all eigenvalues ​​is calculated, and the ratio of the eigenvalue extreme difference to the average value of all eigenvalues ​​is calculated to obtain the airway response eigenvalue.

[0022] As a further embodiment of the present invention, the step of determining whether the patient's respiratory muscles have been strengthened by training specifically includes:

[0023] During the patient's respiratory training, the patient's resistance load data is monitored in real time, the patient's resistance load data is analyzed, and the respiratory muscle endurance characteristic value is calculated based on the load change trend to determine whether the patient's respiratory muscle endurance characteristic value is greater than or equal to the preset threshold. If so, the patient's respiratory muscle groups are strengthened due to training; if not, the patient's respiratory muscle groups are not strengthened due to training.

[0024] As a further solution of the present invention: the process of obtaining the respiratory muscle endurance characteristic value is:

[0025] During the patient's breathing training, the patient's resistance load time series data is collected in real time;

[0026] Perform wavelet packet decomposition on the patient's resistance load time series data and select The wavelet packet node set of the layer is used to obtain the wavelet coefficients of each frequency band;

[0027] Calculate the energy of each wavelet packet node;

[0028] Normalize the energy of all nodes to obtain the energy probability distribution corresponding to each node;

[0029] Calculating the wavelet packet energy entropy in the current time period according to the energy probability distribution;

[0030] The reference wavelet packet energy entropy of the preset starting stage is obtained as a reference value, and the wavelet packet energy entropy in the current time period is ratio-calculated with the reference wavelet packet energy entropy to obtain the respiratory muscle endurance characteristic value.

[0031] As a further solution of the present invention: the patient's airway response characteristic value and respiratory muscle endurance characteristic value are constructed into a comprehensive respiratory rehabilitation characteristic vector, which is input into a pre-trained artificial intelligence model for fusion analysis, specifically including:

[0032] During the patient's respiratory training process, the patient's airway response characteristic value and respiratory muscle endurance characteristic value are obtained, and the airway response characteristic value and respiratory muscle endurance characteristic value are constructed into a comprehensive respiratory rehabilitation characteristic vector as the input of the artificial intelligence model. Minimizing the error between the predicted rehabilitation score and the actual rehabilitation score is used as the training goal of the artificial intelligence model. Based on the trained artificial intelligence model, the patient's rehabilitation score is output. The artificial intelligence model is a random forest model.

[0033] As a further solution of the present invention: the training process of the artificial intelligence model is:

[0034] Multiple historical sets of comprehensive respiratory rehabilitation feature vectors and corresponding rehabilitation score labels were obtained and constructed into a training dataset. The objective function was to minimize the mean square error between the predicted output rehabilitation score and the actual rehabilitation score. Cross-validation was used for model training. By optimizing the Gini impurity, each decision tree in the random forest model learned the optimal splitting rule on different subsamples and feature subsets, gradually improving the overall prediction performance. The trained random forest model can predict the rehabilitation score of the newly input comprehensive respiratory rehabilitation feature vector.

[0035] As a further embodiment of the present invention, the assessment of the patient's current recovery level specifically includes:

[0036] During the patient's breathing training, it is determined whether the patient's rehabilitation score is greater than or equal to a preset first threshold. If so, the patient's rehabilitation level is excellent. If not, it is determined whether the patient's rehabilitation score is less than or equal to a preset second threshold. If so, the patient's rehabilitation level is poor. If not, the patient's rehabilitation level is average.

[0037] As a further solution of the present invention: the adjusting of the patient's training parameters according to the rehabilitation level specifically includes:

[0038] Based on the rehabilitation score results output by the artificial intelligence model, the current patient's respiratory training parameters are dynamically adjusted. If the patient's rehabilitation level is judged to be excellent, the oscillation frequency and resistance load intensity of the current training are reduced, and the frequency of training feedback is increased; if the patient's rehabilitation level is average, the current training parameters are maintained unchanged and the rehabilitation trend is continuously monitored; if the patient's rehabilitation level is poor, the set values ​​of the oscillation frequency and resistance load are gradually increased to enhance the training stimulation effect, thereby realizing personalized, closed-loop dynamic regulation of the patient's respiratory training parameters.

[0039] Beneficial effects of the present invention:

[0040] (1) The present invention constructs a comprehensive respiratory rehabilitation feature vector with the ability to represent multi-dimensional physiological information by fusing the airway response feature value extracted based on differential oscillation frequency stability analysis and the respiratory muscle endurance feature value calculated based on the wavelet packet energy entropy ratio. It then introduces the feature vector as input into an artificial intelligence model optimized based on the random forest algorithm to perform nonlinear feature fusion and rehabilitation level prediction. This evaluation mechanism not only fully exploits the intrinsic relationship between the dynamic response characteristics of the airway and the changes in respiratory muscle function, but also leverages the powerful nonlinear modeling capabilities of the machine learning model to achieve high-precision mapping of the patient's overall rehabilitation status in a complex feature space. Compared with the traditional method of relying on clinical experience or a single indicator to judge the rehabilitation effect, this system, from a data-driven perspective, constructs an objective, quantifiable, and repeatable intelligent evaluation system, which significantly improves the accuracy, robustness, and physiological interpretability of the rehabilitation score, providing a scientific basis and technical support for personalized respiratory rehabilitation training.

[0041] (2) Based on the rehabilitation scoring results output by the artificial intelligence model, the present invention constructs a set of refined hierarchical control mechanisms, which can dynamically adjust the patient's key training parameters such as oscillation frequency and resistance load according to different rehabilitation stages, and truly realize a personalized, closed-loop respiratory rehabilitation training model. Specifically, the system divides the rehabilitation status into three levels of "excellent", "general" and "poor" according to the rehabilitation score, and sets adaptive adjustment functions for frequency and resistance, respectively executing strategies of reducing intensity, maintaining the status quo or gradually increasing. At the same time, combined with the intelligent adjustment of feedback frequency, a complete closed-loop control process of "physiological data real-time monitoring - multi-dimensional feature fusion evaluation - AI-driven decision feedback - dynamic optimization of training parameters" is formed. This system not only effectively improves the safety and individual adaptability of the training process, avoids fatigue accumulation or training failure caused by improper training intensity, but also enhances the patient's training compliance and functional recovery efficiency by continuously optimizing the training stimulation level. Technically, the whole process from data collection to intelligent decision-making to intervention execution is automated, providing a scalable and replicable personalized respiratory rehabilitation solution for the clinic, with significant engineering application value and clinical transformation potential. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] The present invention will be further described below with reference to the accompanying drawings.

[0043] Figure 1 It is a flow chart of a respiratory patient rehabilitation assessment system of the present invention. DETAILED DESCRIPTION

[0044] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0045] Example 1, please refer to Figure 1 As shown, the present invention is a respiratory patient rehabilitation assessment system, comprising:

[0046] A real-time data monitoring module, which monitors the patient's differential oscillation frequency and resistance load data in real time during the patient's breathing training;

[0047] An airway response characteristic evaluation module, which analyzes the patient's differential oscillation frequency data and calculates an airway response characteristic value based on the stability of frequency changes to evaluate the patient's respiratory system's adaptation to training;

[0048] a respiratory muscle endurance characteristic evaluation module, which analyzes the patient's resistance load data and calculates a respiratory muscle endurance characteristic value based on the load change trend to determine whether the patient's respiratory muscles have been strengthened by training;

[0049] A rehabilitation feature fusion and intelligent assessment module, which constructs a comprehensive respiratory rehabilitation feature vector from the patient's airway response characteristic values ​​and respiratory muscle endurance characteristic values, inputs this vector into a pre-trained artificial intelligence model for fusion analysis, and assesses the patient's current rehabilitation level;

[0050] A personalized parameter dynamic control module adjusts the patient's training parameters, including oscillation frequency and resistance load, according to the rehabilitation level, to achieve dynamic optimization of the personalized rehabilitation program and improve the rehabilitation effect.

[0051] In the real-time data monitoring module, during the patient's breathing training, the patient's differential oscillation frequency and resistance load data are monitored in real time, including:

[0052] During breathing training, the present invention uses an integrated physiological signal acquisition device to monitor the patient's differential oscillation frequency and resistance load data in real time. This differential oscillation frequency is acquired using a high-precision airflow sensor and pressure sensing module, located in the mask interface and the exhalation / inhalation channels, respectively. These sensors detect dynamic pressure fluctuations and flow changes within the patient's airway during different training phases. By performing frequency domain conversion on the continuous time series of pressure-flow signals, the differential oscillation frequency values, which characterize the airway response characteristics, are extracted. The sampling frequency is set to no less than 200Hz to ensure that subtle frequency variations are captured.

[0053] Furthermore, the collection of resistance load data is achieved through a variable impedance adjustment unit and a force feedback sensor built into the respiratory training device. The sensor records in real time the force exerted by the patient's respiratory muscles during inhalation and exhalation, and calculates the actual load intensity in combination with a preset resistance setting value; the system samples the resistance load data multiple times per second and performs sliding window filtering to eliminate instantaneous noise interference and retain a stable load change trend; all collected data is transmitted to the central processing unit via an embedded data acquisition module for use by subsequent feature extraction and rehabilitation assessment modules, thereby building a complete closed-loop monitoring system for physiological parameters.

[0054] In the airway response characteristic assessment module, the patient's differential oscillation frequency data is analyzed. Based on the stability of frequency changes, the airway response characteristic value is calculated to assess the patient's respiratory system's adaptation to training. Specifically, the following are included:

[0055] During the patient's breathing training, the patient's differential oscillation frequency data is monitored in real time, and the patient's differential oscillation frequency data is analyzed. Based on the stability of the frequency change, the airway response characteristic value is calculated to determine whether the patient's airway response characteristic value is greater than or equal to the preset threshold. If so, the patient's respiratory system is not adapted to the training; if not, the patient's respiratory system is adapted to the training.

[0056] The process of obtaining the airway response characteristic value is as follows:

[0057] Collect the patient's real-time differential oscillation frequency data during breathing training to form a time series signal;

[0058] Divide the time series into multiple blocks of length , and perform mean removal on each subsequence to obtain a decentralized subsequence. The mean removal is the value of the sampling point in each subsequence minus the average value of all sampling points in the corresponding subsequence.

[0059] Construct the covariance matrix based on all subsequences, and the calculation expression is: ;

[0060] in, Indicates the The subsequence and The covariance between the subsequences, and Represents two different subsequences, represents the length of the subsequence, represents the number of sampling points, Indicates the The first of the subsequences The value of the sampling point, Indicates the The average value of the subsequences, Indicates the The first of the subsequences The value of the sampling point, Indicates the The average value of subsequences;

[0061] Perform eigenvalue decomposition on the covariance matrix to obtain all its eigenvalues;

[0062] According to the random matrix theory, the Marchenko-Pastur distribution model under the noise background is established, and its probability density function is: ;

[0063] in, represents the variance of the system noise, represents the probability density function of the Marchenko-Pastur distribution model, represents the eigenvalue, represents the maximum theoretical eigenvalue of the Marchenko-Pastur distribution model, represents the minimum theoretical eigenvalue of the Marchenko-Pastur distribution model;

[0064] The difference between the maximum theoretical eigenvalue and the minimum theoretical eigenvalue is calculated to obtain the eigenvalue extreme difference, the average value of all eigenvalues ​​is calculated, and the ratio of the eigenvalue extreme difference to the average value of all eigenvalues ​​is calculated to obtain the airway response eigenvalue.

[0065] It should be noted that in the process of extracting airway response eigenvalues, the present invention innovatively introduces covariance matrix eigenvalue analysis and Marchenko-Pastur distribution model modeling in random matrix theory. By subsequence division, demeaning, covariance matrix construction and eigenvalue decomposition of the differential oscillation frequency time series, combined with the ratio calculation of the theoretical eigenvalue range under a noise background, a quantitative evaluation indicator with high robustness and physiological interpretability - the airway response eigenvalue is constructed. This system not only effectively eliminates trend terms and redundant fluctuations in the signal, improving the sensitivity to the stability of airway dynamic response, but also enhances the ability to identify real physiological changes by correlating the measured eigenvalues ​​with the theoretical noise distribution, avoiding the risk of misjudgment caused by noise interference in traditional systems. Technically, a breakthrough has been achieved in extracting structured features from nonlinear physiological signals for respiratory system adaptability assessment, providing a scientific basis and real-time control foundation for personalized respiratory rehabilitation training, and significantly improving the intelligence level and clinical practicality of rehabilitation assessment.

[0066] In the respiratory muscle endurance characteristic assessment module, the patient's resistance load data is analyzed, and the respiratory muscle endurance characteristic value is calculated based on the load change trend to determine whether the patient's respiratory muscles have been strengthened by training. Specifically, it includes:

[0067] During the patient's respiratory training, the patient's resistance load data is monitored in real time, the patient's resistance load data is analyzed, and the respiratory muscle endurance characteristic value is calculated based on the load change trend to determine whether the patient's respiratory muscle endurance characteristic value is greater than or equal to the preset threshold. If so, the patient's respiratory muscle groups are strengthened due to training; if not, the patient's respiratory muscle groups are not strengthened due to training.

[0068] The process of obtaining the respiratory muscle endurance characteristic value is as follows:

[0069] During the patient's breathing training, the patient's resistance load time series data is collected in real time;

[0070] Perform wavelet packet decomposition on the patient's resistance load time series data and select The wavelet packet node set of the layer, where , get the wavelet coefficients of each frequency band;

[0071] Calculate the energy of each wavelet packet node. The calculation expression is: ;

[0072] in, Indicates the first Tier The signal sample value at each node, Indicates the time point collection point, represents each signal sample value, represents the number of layers of wavelet packet decomposition, represents the first nodes;

[0073] Normalize the energy of all nodes to obtain the energy probability distribution corresponding to each node. The calculation expression is: ;

[0074] in, represents the first Layer, Energy probability at each node;

[0075] According to the energy probability distribution, the wavelet packet energy entropy in the current time period is calculated, and the calculation expression is: ;

[0076] in, represents the energy entropy of wavelet packet;

[0077] The reference wavelet packet energy entropy of the preset starting stage is obtained as a reference value, and the wavelet packet energy entropy in the current time period is ratio-calculated with the reference wavelet packet energy entropy to obtain the respiratory muscle endurance characteristic value.

[0078] It should be noted that the present invention innovatively integrates wavelet packet transform and energy entropy analysis techniques in the extraction of respiratory muscle endurance eigenvalues. By performing multi-resolution frequency band decomposition on the patient's resistance load time series, it accurately captures the energy distribution changes under muscle fatigue and adaptation during respiratory training. This system not only efficiently processes non-stationary and nonlinear physiological signals but also quantifies the functional stability and adaptability trends of the respiratory muscles at different training stages by constructing energy probability distributions in each frequency band and calculating their information entropy. Furthermore, by incorporating a baseline energy entropy at the initial training stage as a reference, the respiratory muscle endurance eigenvalue is constructed in the form of a ratio, effectively eliminating the influence of individual differences and enhancing the comparability and robustness of the assessment results. This technical breakthrough breaks through the limitations of traditional endurance state assessment based on time-domain mean or peak values. For the first time, the dynamic ratio of energy entropy of complex signals has been applied to respiratory rehabilitation assessment, significantly improving the accuracy of identifying respiratory muscle training effects and the ability to provide real-time feedback, providing strong theoretical support and engineering implementation path for personalized, intelligent respiratory rehabilitation.

[0079] In the rehabilitation feature fusion and intelligent assessment module, the patient's airway response characteristic values ​​and respiratory muscle endurance characteristic values ​​are constructed into a comprehensive respiratory rehabilitation feature vector, which is input into a pre-trained artificial intelligence model for fusion analysis to assess the patient's current rehabilitation level. Specifically, the following are included:

[0080] During the patient's respiratory training process, the patient's airway response characteristic value and respiratory muscle endurance characteristic value are obtained, and the airway response characteristic value and respiratory muscle endurance characteristic value are constructed into a comprehensive respiratory rehabilitation characteristic vector as the input of the artificial intelligence model. Minimizing the error between the predicted rehabilitation score and the actual rehabilitation score is used as the training goal of the artificial intelligence model. Based on the trained artificial intelligence model, the patient's rehabilitation score is output. The artificial intelligence model is a random forest model.

[0081] Multiple historical sets of comprehensive respiratory rehabilitation feature vectors and corresponding rehabilitation score labels were obtained and constructed into a training dataset. The objective function was to minimize the mean square error between the predicted output rehabilitation score and the actual rehabilitation score. Cross-validation was used for model training. By optimizing the Gini impurity, each decision tree in the random forest model learned the optimal splitting rule on different subsamples and feature subsets, gradually improving the overall prediction performance. The trained random forest model can predict the rehabilitation score of the newly input comprehensive respiratory rehabilitation feature vector.

[0082] It should be noted that the present invention innovatively introduces multi-feature fusion and artificial intelligence modeling technology in the rehabilitation assessment link, constructs a structured comprehensive respiratory rehabilitation feature vector based on the extracted airway response eigenvalues ​​and respiratory muscle endurance eigenvalues, and uses a random forest model to perform nonlinear fusion analysis on it, thereby achieving an accurate quantitative assessment of the patient's rehabilitation level. By constructing a training mechanism with the goal of minimizing the mean square error, the system effectively captures the complex interactive relationship between airway adaptability and muscle endurance in the feature space, significantly improving the stability and generalization ability of rehabilitation score prediction; at the same time, combined with the Gini impurity optimization strategy and cross-validation mechanism, the model still has excellent robustness and discrimination ability in the face of individual differences and training data noise. Technically, it has broken through the limitations of the traditional single indicator for evaluating rehabilitation status, and realized the intelligent mapping from multi-dimensional physiological characteristics to the overall rehabilitation level, providing key technical support and clinical application value for personalized and dynamic respiratory rehabilitation systems.

[0083] In the personalized parameter dynamic control module, the patient's training parameters, including oscillation frequency and resistance load, are adjusted according to the patient's rehabilitation level to achieve dynamic optimization of the personalized rehabilitation plan and improve the rehabilitation effect. Specifically, it includes:

[0084] Based on the rehabilitation score results output by the artificial intelligence model, the current patient's respiratory training parameters are dynamically adjusted. If the patient's rehabilitation level is judged to be excellent, the oscillation frequency and resistance load intensity of the current training are reduced, and the frequency of training feedback is increased; if the patient's rehabilitation level is average, the current training parameters are maintained unchanged and the rehabilitation trend is continuously monitored; if the patient's rehabilitation level is poor, the set values ​​of the oscillation frequency and resistance load are gradually increased to enhance the training stimulation effect, thereby realizing personalized, closed-loop dynamic regulation of the patient's respiratory training parameters.

[0085] Example 2: To verify the feasibility and effectiveness of the present invention in respiratory patient rehabilitation assessment, the following implementation parameters and test procedures were selected for exemplary verification:

[0086]

[0087]

[0088] Data Description:

[0089] Differential oscillation frequency: reflects the dynamic airway pressure fluctuation characteristics. The larger the numerical fluctuation amplitude (the larger the standard deviation), the worse the frequency stability and the higher the corresponding airway response characteristic value.

[0090] Resistance load: Adjust with training intensity, increase to enhance stimulation when recovery level is poor, and decrease to reduce load when recovery level is excellent.

[0091] Airway responsiveness characteristic values:

[0092] The threshold is set to 1.2 (≥1.2 indicates unsuitable training) based on the calculation of the covariance matrix eigenvalues ​​and the Marchenko-Pastur distribution model.

[0093] For example, if the characteristic value is 1.32>1.2 at 10-15 minutes, it indicates that the respiratory system is not adapted to the current training.

[0094] Respiratory muscle endurance characteristic values:

[0095] The threshold was set at 1.1 (≥1.1 indicated enhanced muscle function) based on the calculation of the wavelet packet energy entropy ratio.

[0096] At 20-25 minutes, the eigenvalue was 1.15>1.1, indicating that the respiratory muscles were strengthened due to training.

[0097] Rehabilitation scoring and adjustment strategies:

[0098] The scoring range is 0-100. For excellent (≥80), the parameters are reduced and feedback is increased. For poor (≤39), the parameters are increased to enhance stimulation. For average (40-79), the parameters are maintained.

[0099] It should be noted that the data simulates the patient's fluctuation process from "average" to "poor" to "excellent" during 40 minutes of training, reflecting the closed-loop mechanism of system dynamic monitoring, feature analysis and parameter adjustment.

[0100] Working principle of the present invention: The present invention aims to achieve real-time monitoring and intelligent evaluation of the adaptability of the patient's respiratory system and respiratory muscle endurance. The patient's differential oscillation frequency and resistance load data are acquired in real time through an integrated physiological signal acquisition device; the covariance matrix eigenvalue decomposition and Marchenko-Pastur distribution modeling technology are used to extract airway response eigenvalues ​​with high robustness to evaluate the patient's airway adaptation to training; the wavelet packet energy entropy and baseline entropy ratio calculation method is introduced to construct respiratory muscle endurance eigenvalues ​​to accurately identify the enhancement trend of respiratory muscle groups during training; the above two eigenvalues ​​are constructed into a comprehensive respiratory rehabilitation feature vector, and input into an artificial intelligence model trained based on the random forest algorithm for nonlinear fusion analysis, and a rehabilitation score reflecting the patient's overall rehabilitation level is output; personalized closed-loop regulation of training parameters is achieved based on the rehabilitation score results, and the oscillation frequency and resistance load intensity are dynamically adjusted to ensure the safety, effectiveness and individual adaptability of the training program. This invention breaks through the limitations of traditional reliance on subjective judgment or a single indicator to evaluate rehabilitation status, and realizes the full process automation from multi-dimensional physiological data collection, feature extraction to intelligent evaluation and feedback regulation, significantly improving the scientificity and intelligence level of respiratory rehabilitation training, and providing reliable technical support and implementation path for clinical personalized rehabilitation intervention.

[0101] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0102] The above is a detailed description of an embodiment of the present invention. However, the content described is only a preferred embodiment of the present invention and should not be considered to limit the scope of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of the patent coverage of the present invention.

Claims

1. A respiratory patient rehabilitation assessment system, characterized in that: include: A real-time data monitoring module, which monitors the patient's differential oscillation frequency and resistance load data in real time during the patient's breathing training; The airway response characteristic evaluation module analyzes the patient's differential oscillation frequency data and calculates the airway response characteristic value based on the stability of the frequency change to evaluate the patient's respiratory system's adaptation to the training. Specifically, it includes: During the patient's respiratory training, the patient's differential oscillation frequency data is monitored in real time, the differential oscillation frequency data is analyzed, and the airway response characteristic value is calculated based on the stability of the frequency change to determine whether the patient's airway response characteristic value is greater than or equal to a preset threshold. If so, the patient's respiratory system is not adapted to the training; if not, the patient's respiratory system is adapted to the training; The process of obtaining the airway response characteristic value is as follows: Collect the patient's real-time differential oscillation frequency data during breathing training to form a time series signal; Divide the time series into multiple blocks of length , and perform mean removal on each subsequence to obtain a decentralized subsequence. The mean removal is the value of the sampling point in each subsequence minus the average value of all sampling points in the corresponding subsequence. Construct the covariance matrix based on all subsequences; Perform eigenvalue decomposition on the covariance matrix to obtain all its eigenvalues; Based on random matrix theory, a Marchenko-Pastur distribution model is established under a noise background, the probability density function is calculated, and the maximum theoretical eigenvalue and minimum theoretical eigenvalue of the Marchenko-Pastur distribution model are obtained; Calculate the difference between the maximum theoretical eigenvalue and the minimum theoretical eigenvalue to obtain the eigenvalue extreme difference, calculate the average value of all eigenvalues, and calculate the ratio of the eigenvalue extreme difference to the average value of all eigenvalues ​​to obtain the airway response characteristic value; The respiratory muscle endurance characteristic evaluation module analyzes the patient's resistance load data and calculates the respiratory muscle endurance characteristic value based on the load change trend to determine whether the patient's respiratory muscles have been strengthened due to training. Specifically, it includes: Determining whether the patient's respiratory muscles are strengthened due to training specifically includes: During the patient's respiratory training, the patient's resistance load data is monitored in real time, the patient's resistance load data is analyzed, and according to the load change trend, the respiratory muscle endurance characteristic value is calculated to determine whether the patient's respiratory muscle endurance characteristic value is greater than or equal to a preset threshold. If so, the patient's respiratory muscle group has been strengthened due to the training; if not, the patient's respiratory muscle group has not been strengthened due to the training; The process of obtaining the respiratory muscle endurance characteristic value is as follows: During the patient's breathing training, the patient's resistance load time series data is collected in real time; Perform wavelet packet decomposition on the patient's resistance load time series data and select The wavelet packet node set of the layer is used to obtain the wavelet coefficients of each frequency band; Calculate the energy of each wavelet packet node; Normalize the energy of all nodes to obtain the energy probability distribution corresponding to each node; Calculating the wavelet packet energy entropy in the current time period according to the energy probability distribution; Obtaining a reference wavelet packet energy entropy at a preset start stage as a reference value, calculating the ratio of the wavelet packet energy entropy in the current time period to the reference wavelet packet energy entropy to obtain a respiratory muscle endurance characteristic value; A rehabilitation feature fusion and intelligent assessment module, which constructs a comprehensive respiratory rehabilitation feature vector from the patient's airway response characteristic values ​​and respiratory muscle endurance characteristic values, inputs this vector into a pre-trained artificial intelligence model for fusion analysis, and assesses the patient's current rehabilitation level; A personalized parameter dynamic control module adjusts the patient's training parameters, including oscillation frequency and resistance load, according to the rehabilitation level, to achieve dynamic optimization of the personalized rehabilitation program and improve the rehabilitation effect.

2. A respiratory patient rehabilitation assessment system according to claim 1, characterized in that: The patient's airway response characteristic value and respiratory muscle endurance characteristic value are constructed into a comprehensive respiratory rehabilitation feature vector, which is input into a pre-trained artificial intelligence model for fusion analysis, specifically including: During the patient's respiratory training process, the patient's airway response characteristic value and respiratory muscle endurance characteristic value are obtained, and the airway response characteristic value and respiratory muscle endurance characteristic value are constructed into a comprehensive respiratory rehabilitation characteristic vector as the input of the artificial intelligence model. Minimizing the error between the predicted rehabilitation score and the actual rehabilitation score is used as the training goal of the artificial intelligence model. Based on the trained artificial intelligence model, the patient's rehabilitation score is output. The artificial intelligence model is a random forest model.

3. A respiratory patient rehabilitation assessment system according to claim 2, characterized in that: The training process of the artificial intelligence model is as follows: Multiple historical sets of comprehensive respiratory rehabilitation feature vectors and corresponding rehabilitation score labels were obtained and constructed into a training dataset. The objective function was to minimize the mean square error between the predicted output rehabilitation score and the actual rehabilitation score. Cross-validation was used for model training. By optimizing the Gini impurity, each decision tree in the random forest model learned the optimal splitting rule on different subsamples and feature subsets, gradually improving the overall prediction performance. The trained random forest model can predict the rehabilitation score of the newly input comprehensive respiratory rehabilitation feature vector.

4. A respiratory patient rehabilitation assessment system according to claim 1, characterized in that: The assessment of the patient's current recovery level specifically includes: During the patient's breathing training, it is determined whether the patient's rehabilitation score is greater than or equal to a preset first threshold. If so, the patient's rehabilitation level is excellent. If not, it is determined whether the patient's rehabilitation score is less than or equal to a preset second threshold. If so, the patient's rehabilitation level is poor. If not, the patient's rehabilitation level is average.

5. A respiratory patient rehabilitation assessment system according to claim 1, characterized in that: The adjusting of the patient's training parameters according to the rehabilitation level specifically includes: Based on the rehabilitation score results output by the artificial intelligence model, the current patient's respiratory training parameters are dynamically adjusted. If the patient's rehabilitation level is judged to be excellent, the oscillation frequency and resistance load intensity of the current training are reduced, and the frequency of training feedback is increased; if the patient's rehabilitation level is average, the current training parameters are maintained unchanged and the rehabilitation trend is continuously monitored; if the patient's rehabilitation level is poor, the set values ​​of the oscillation frequency and resistance load are gradually increased to enhance the training stimulation effect, thereby realizing personalized, closed-loop dynamic regulation of the patient's respiratory training parameters.

Images