Intelligent evaluation system for lung rehabilitation detection

Abstract

The invention belongs to the technical field of medical internet of things, and discloses a lung rehabilitation detection intelligent evaluation system, which comprises a database construction module used for constructing an evaluation reference database according to historical evaluation reference data, and the evaluation reference data comprises historical reference patient physiological parameter data and historical lung rehabilitation detection evaluation data; the data collection module is used for collecting physiological parameter data, chest image data and air quality data of the pre-evaluated patient; the data processing module is used for acquiring the physiological parameter data of the patient to be evaluated through the pre-evaluated physiological parameter data of the patient and the evaluation reference database; processing the acquired physiological parameter data, chest image data and air quality data of the to-be-evaluated patient to obtain a comprehensive feature data set of the to-be-evaluated patient; through systematic evaluation and dynamic adjustment, rehabilitation of the patient is more effectively promoted, an evaluation result provides a scientific basis for decision making of a doctor, and a treatment scheme can be more reasonably formulated.

Description

Technical Field

[0001] The present invention relates to the technical field of medical Internet of Things, and more specifically, to an intelligent evaluation system for lung rehabilitation detection. Background Art

[0002] The patent with the patent publication number CN106372390A discloses a self-help health cloud service system for preventing lung cancer based on a deep convolutional neural network, including a convolutional neural network for deep learning and training recognition, a segmentation module for segmenting lung areas from CT images based on a fully convolutional neural network, a deep convolutional neural network for lung cancer diagnosis and classification, and a self-help health cloud service platform for early prevention and treatment based on the identified suspected lung cancer type. The present invention can effectively improve the automation and intelligence level of lung cancer screening based on the mobile Internet, enable more citizens to understand and participate in self-help health testing, evaluation, and guidance, improve the sensitivity, specificity, and accuracy of clinical diagnosis of early lung cancer screening, realize the "early detection, early diagnosis, and early treatment" of lung cancer, and increase self-health management capabilities.

[0003] The existing pulmonary rehabilitation testing and evaluation system has the following main problems:

[0004] Unclear data structure will affect subsequent data retrieval, analysis and maintenance, and may lead to inefficient data processing, increase data processing time and reduce decision-making efficiency; failure to consider the use of collection and structured data storage methods will make it difficult to effectively manage and organize historical evaluation reference data, resulting in data redundancy and confusion, making it difficult to find and use data; lack of sufficient analysis and utilization of the relationship between physiological parameter data and evaluation data will weaken the reliability of evaluation results and affect the scientific basis for patient rehabilitation; failure to introduce constraints in the evaluation reference data graph will make it difficult to accurately establish the correlation between the reference patient physiological parameter data and the pulmonary rehabilitation test evaluation data; the evaluation reference data graph without constraints may cause the connection between data nodes to be too complex or too simple, and cannot accurately reflect the correlation between data;

[0005] The step size cannot be effectively determined, which slows down the convergence of the model during training and increases the training time, especially for large-scale data sets. The model optimization process will not be able to quickly find the global optimal solution. The lack of reasonable adjustment of hyperparameters makes the model more prone to overfitting or underfitting, resulting in poor prediction results and difficulty adapting to different data distributions. Failure to control the update amplitude of hyperparameters may lead to oscillation during model training, that is, the update step size of each iteration is too large or too small, affecting the stability of training.

[0006] Lacking a dynamic adjustment mechanism, the rehabilitation plan may be too heavy, causing the patient's body to be overloaded and increasing the risk of complications or recurrence of the disease; or the rehabilitation plan may be too light, causing the patient's recovery to progress slowly and miss the optimal recovery time window; failure to track the patient's recovery status in a timely manner through quantitative evaluation may cause fluctuations in the rehabilitation progress and fail to maintain a smooth rehabilitation curve, thus affecting the treatment effect; especially when the rehabilitation progress slows down or worsens, failure to detect and adjust in time may prolong the recovery time.

[0007] In view of this, the present invention proposes a pulmonary rehabilitation detection intelligent evaluation system to solve the above problems. Summary of the invention

[0008] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solutions: a lung rehabilitation detection intelligent evaluation system, comprising:

[0009] A database construction module, used to construct an evaluation reference database based on historical evaluation reference data, wherein the evaluation reference data includes reference patient physiological parameter data and pulmonary rehabilitation test evaluation data;

[0010] A data collection module, used to collect pre-assessment patient physiological parameter data, chest imaging data, and air quality data;

[0011] A data processing module is used to obtain the physiological parameter data of the patient to be evaluated through the physiological parameter data of the pre-evaluated patient and the evaluation reference database; the acquired physiological parameter data, chest imaging data and air quality data of the patient to be evaluated are processed to obtain a comprehensive characteristic data set of the patient to be evaluated;

[0012] The pulmonary rehabilitation detection module is used to obtain a pulmonary status index prediction model through training of a comprehensive feature data set of the patient to be evaluated, and to predict the patient's pulmonary rehabilitation status index through the pulmonary status index prediction model;

[0013] A pulmonary rehabilitation assessment module is used to compare the predicted patient pulmonary rehabilitation status index with the preset patient pulmonary rehabilitation status index threshold to assess whether the patient's pulmonary rehabilitation index is abnormal;

[0014] The pulmonary rehabilitation trend assessment module generates early warning information through the pulmonary rehabilitation detection intelligent terminal if the patient's pulmonary rehabilitation indicators are abnormal, and promptly reminds the doctor; if the patient's pulmonary rehabilitation indicators are normal, the patient's future pulmonary rehabilitation trend is predicted based on the assessment reference database and the patient's pulmonary rehabilitation status indicators; the pulmonary rehabilitation detection intelligent terminal evaluates the patient's pulmonary rehabilitation time based on the patient's future pulmonary rehabilitation trend prediction results, and dynamically adjusts the rehabilitation plan;

[0015] The modules are connected to each other via wired and / or wireless means.

[0016] Furthermore, the method for constructing an evaluation reference database based on historical evaluation reference data includes:

[0017] The reference patient physiological parameter data include the patient's respiratory rate, blood oxygen saturation, heart rate, blood pressure, hemoglobin concentration and blood lactate level; the pulmonary rehabilitation test assessment data include the patient's vital capacity, maximum ventilation volume, forced expiratory volume in one second, respiratory airflow rate, carbon dioxide partial pressure and oxygen partial pressure;

[0018] S21. Define a set x to represent the reference patient physiological parameter data; the set x is represented by x = {x1, x2, ..., x p ,...,x M}; where x p is the pth type of patient physiological parameter data; p is the index of the patient physiological parameter data type, p=1,...,M; M is the number of patient physiological parameter data types; each data x in the set x p A node in the corresponding evaluation reference data graph;

[0019] Define set y to represent pulmonary rehabilitation test evaluation data; set y is represented by y = {y1, y2, ..., y q ,...,y N}; where y q is the qth type of pulmonary rehabilitation test evaluation data; q is the index of the type of pulmonary rehabilitation test evaluation data, q=1,...,N; N is the number of types of pulmonary rehabilitation test evaluation data; each data y in the set y q corresponds to another node in the evaluation reference data graph;

[0020] S22, constructing a vertex set V=x∪y in the evaluation reference data graph through all reference patient physiological parameter data nodes and pulmonary rehabilitation test evaluation data nodes, connecting each reference patient physiological parameter data node and pulmonary rehabilitation test evaluation data node through the edge weight ω obtained by calculating the Pearson correlation coefficient, constructing an edge set E of the evaluation reference data graph, and finally obtaining an evaluation reference data graph model G=(V,E);

[0021] S23, defining an optimization objective function, and introducing constraints in the evaluation reference data graph; processing the constraints in the evaluation reference data graph by introducing the Lagrange multiplier method formula, and ensuring that the weight distribution of the evaluation reference data graph meets the optimization goal, the preset optimization goal is the matching degree between the reference patient physiological parameter data and the pulmonary rehabilitation test evaluation data; the Lagrange multiplier method formula is:

[0022] Among them, x i′is the physiological parameter data of the i′th reference patient; i′ is the ith pulmonary rehabilitation test evaluation data; Sim(x i′ ,y i′ ) is used to measure x i′ and i′ The optimization objective function of the matching degree between i′ To measure each pair of x i′ and i′ The weight of; i′ is the index of different data types; n′ is the total number of reference patient physiological parameter data and pulmonary rehabilitation test evaluation data; σ k The kth Lagrange multiplier for the trade-off between the optimization objective function and the constraints in the evaluation reference data graph; g k (x, y) is the constraint condition in the evaluation reference data graph; j is the number of constraints in the evaluation reference data graph; σ is the set of Lagrange multipliers, including all σ k ; k is the index of the constraint in the evaluation reference data graph;

[0023] S24, limiting the number of constraints in the evaluation reference data graph by a constraint number limiting formula, wherein the constraint number limiting formula is: Wherein, j′ is the number of constraints in the restricted evaluation reference data graph; b is the total number of types of reference patient physiological parameter data and pulmonary rehabilitation test evaluation data;

[0024] S25, defining a penalty function corresponding to the constraint conditions in the evaluation reference data graph, and using the penalty function to punish solutions that do not satisfy the constraint conditions in the evaluation reference data graph, wherein the penalty function is Among them, P(x,y) is the penalty function; ρ k To control the penalty coefficient for the degree of violation of the constraints in the evaluation reference data graph;

[0025] S26. Through the API interface provided in the database, the generated evaluation reference data graph is imported into the database, thereby constructing a structured evaluation reference database.

[0026] Furthermore, the pre-assessed patient physiological parameter data includes the patient's respiratory rate, blood oxygen saturation, heart rate, blood pressure, hemoglobin concentration and blood lactate level; the chest imaging data includes the patient's chest X-ray data, chest CT scan image data and magnetic resonance imaging data; the air quality data includes the air pollution index of the patient's living environment.

[0027] Furthermore, the method for obtaining the physiological parameter data of the patient to be evaluated by pre-evaluating the physiological parameter data of the patient and evaluating the reference database includes:

[0028] The pre-evaluation patient physiological parameter data is matched with the historical evaluation reference data in the evaluation reference database by using cosine similarity calculation, the reference patient physiological parameter data that is most similar to the pre-evaluation patient physiological parameter data is found, and the physiological parameter data of the patient to be evaluated is obtained by integration.

[0029] Furthermore, the method of processing the acquired physiological parameter data, chest image data and air quality data of the patient to be evaluated to obtain a comprehensive characteristic data set of the patient to be evaluated includes:

[0030] Use the K-means clustering algorithm to identify outliers in the patient's physiological parameter data, chest imaging data, and air quality data, and remove the identified outliers to obtain processed patient physiological characteristic data sets, chest imaging characteristic data sets, and air quality characteristic data sets;

[0031] The acquired patient physiological characteristic data set, chest imaging characteristic data set and air quality characteristic data set are normalized by standard deviation and converted into a standard normal distribution to obtain a normalized patient physiological characteristic data set, chest imaging characteristic data set and air quality characteristic data set;

[0032] The normalized patient physiological characteristic dataset, chest imaging characteristic dataset and air quality characteristic dataset are fused through a weighted formula to form a comprehensive characteristic dataset of the patient to be evaluated. The patient physiological characteristic dataset is recorded as E1, the chest imaging characteristic dataset is recorded as E2, and the air quality characteristic dataset is recorded as E3;

[0033] The weighted formula is: Hc=E1·δ1+E2·δ2+E3·δ3; wherein δ1 is the weight coefficient of the patient's physiological characteristic data set; δ2 is the weight coefficient of the chest image characteristic data set; and δ3 is the weight coefficient of the air quality characteristic data set.

[0034] Furthermore, the training method of the lung status index prediction model includes:

[0035] The data set is divided into a training set, a validation set and a test set, and a lung status index prediction model is constructed; the sample set is a subset of the data set, and each sample set includes a historical comprehensive feature data set of patients to be evaluated and a corresponding patient lung rehabilitation status index; the input data of the model is a historical comprehensive feature data set of patients to be evaluated; the output label is a patient lung rehabilitation status index; the lung status index prediction model is a gradient boosting decision tree model;

[0036] Initialize the lung status indicator prediction model and set the number of trees, learning rate, and tree depth hyperparameters. Use k-fold cross-validation to train the model using different hyperparameter combinations. Use the cross-validation results to tune the initially set hyperparameters and select the best performing hyperparameter combination.

[0037] Use mean square error as the loss function to measure the difference between the model's predicted value and the actual value; use the training set to train the lung status indicator prediction model, and train a new decision tree based on the residual of the current model in each iteration; adjust the model's hyperparameters to tune the model based on the model performance feedback, and retrain the model with the adjusted hyperparameters. After the entire training process is completed, cross-validation is used to select the most appropriate hyperparameters;

[0038] When the training reaches the preset model complexity, it stops and the final trained lung status index prediction model is obtained; the trained lung status index prediction model is used to predict the comprehensive feature data set of the current patient to be evaluated to obtain the patient's lung rehabilitation status index.

[0039] Furthermore, the method of adjusting the hyperparameters of the model to optimize the model includes:

[0040] S71, calculating the first-order derivative and the second-order derivative of the mean square error loss function with respect to the hyperparameter, determining the change gradient of the mean square error loss function through the first-order derivative, determining the step length of moving along the gradient direction when updating the parameters through the second-order derivative, and accelerating the convergence speed of the mean square error loss function;

[0041] The first-order derivative calculation formula is: Among them, ▽L(θ) is the first-order derivative of the mean square error loss function with respect to the hyperparameter θ, that is, the gradient; is the first-order partial derivative of the mean square error loss function with respect to the hyperparameter θ;

[0042] The second-order derivative calculation formula is: Among them, H(θ) is the second-order derivative of the mean square error loss function with respect to the hyperparameter θ, and the second-order derivative matrix is ​​the Hessian matrix; is the second-order partial derivative of the mean square error loss function with respect to the hyperparameter θ;

[0043] S72. In each iteration, the hyperparameter update formula of the Newton method is used to update the hyperparameters of the model according to the current gradient and the inverse of the Hessian matrix to reduce the L(θ) error;

[0044] The Newton method hyperparameter update formula is: θ t+1 =θ t -H(θ t ) -1 ▽L(θ t ), where θ t+1 is the hyperparameter value after the t+1th iteration update; θ t is the hyperparameter value at the current tth iteration; H(θ t )-1 is the inverse matrix of the Hessian matrix at the current t-th iteration; ▽L(θ t ) is the current hyperparameter θ t The gradient under; t is the number of iterations;

[0045] S73. Continuously iterate the Newton method hyperparameter update formula to adjust the hyperparameters, limit the iterative update of the hyperparameters through the iterative update limit formula, and stop iterating when the hyperparameters of the model tend to be stable;

[0046] The iterative update restriction formula is: in, is the amplitude of hyperparameter update; n is the number of samples in the data set; T is the number of new decision trees; α is a constant for adjusting the convergence speed; β is a factor for adjusting the influence of the number of samples on the convergence speed; γ is a factor for adjusting the influence of the number of decision trees on the convergence speed;

[0047] S74, adjusting the influencing factors β and γ used to adjust the number of samples and the number of decision trees on the convergence speed through the influencing factor adjustment formula, so that the values ​​of β and γ reach the optimal value; the influencing factor adjustment formula is: ln(n β )+ln(T γ )=μ+λ·ln(I)+υ·ln(l); where λ is the coefficient of the influencing factor β; υ is the coefficient of the influencing factor γ; μ is the basic convergence speed of the model; I is the number of iterations of the model; l is the step size moving along the gradient direction when updating the parameters;

[0048] S75. The sum of the coefficient λ of the influencing factor β and the coefficient υ of the influencing factor γ is 1. By keeping the β value unchanged, increasing or decreasing the γ value, observing the convergence speed of the model, and then keeping the γ value unchanged, increasing or decreasing the β value, when the β and γ values ​​are adjusted to the optimal level, the convergence speed of the model is optimal, and the amplitude of the hyperparameter update is optimal. The hyperparameter combination at this time is the optimal hyperparameter combination.

[0049] Furthermore, the method of comparing the predicted patient's lung rehabilitation status index with a preset patient's lung rehabilitation status index threshold to evaluate whether the patient's lung rehabilitation index is abnormal includes:

[0050] If the predicted patient's lung rehabilitation status index is less than or equal to a preset patient's lung rehabilitation status index threshold, the patient's lung rehabilitation index is determined to be normal;

[0051] If the predicted patient's lung rehabilitation status index is greater than the preset patient's lung rehabilitation status index threshold, the patient's lung rehabilitation index is determined to be abnormal.

[0052] Furthermore, the method for predicting the patient's future pulmonary rehabilitation trend based on the evaluation reference database and the patient's pulmonary rehabilitation status index includes:

[0053] The data set is divided into a training set, a validation set and a test set, and a pulmonary rehabilitation trend prediction model is constructed; the sample set is a subset of the data set, and each sample set includes a historical reference database and a patient's pulmonary rehabilitation status indicator and the corresponding patient's future pulmonary rehabilitation trend; the pulmonary rehabilitation trend prediction model is constructed using the deep learning library TensorFlow; the pulmonary rehabilitation trend prediction model includes an input layer, an LSTM layer and an output layer; the pulmonary rehabilitation trend prediction model is an LSTM model;

[0054] Define L2 regularized mean square error as the loss function of the model to measure the difference between the patient's future pulmonary rehabilitation trend output by the LSTM layer and the actual patient's future pulmonary rehabilitation trend;

[0055] The gradient of the loss function with respect to the model parameters is calculated through the chain rule, and the Adam optimizer is used to update the parameters and tune the model. The performance of the model in the prediction task is evaluated through the test set. When the training reaches the preset model complexity, it is stopped to obtain a trained pulmonary rehabilitation trend prediction model. The trained pulmonary rehabilitation trend prediction model is used to predict the current reference database and the patient's pulmonary rehabilitation status indicators to obtain the patient's future pulmonary rehabilitation trend.

[0056] Furthermore, the method of evaluating the patient's lung rehabilitation time and dynamically adjusting the rehabilitation plan based on the patient's future lung rehabilitation trend prediction results includes:

[0057] The patient's future pulmonary rehabilitation trend prediction result includes the pulmonary rehabilitation trends at z future time points; a pulmonary rehabilitation time evaluation model is constructed to evaluate the patient's pulmonary rehabilitation time according to the patient's future pulmonary rehabilitation trend prediction result; the pulmonary rehabilitation time evaluation model is: Among them, T′ is the assessed lung rehabilitation time of the patient; S is the lung rehabilitation speed of the patient; AP is the air pollution index of the patient's living environment; VC is the vital capacity of the patient; ξ1 is the weight coefficient for adjusting the influence of the patient's lung rehabilitation speed on the patient's lung rehabilitation time; ξ2 is the weight parameter for controlling the influence of air pollution index and vital capacity on the patient's lung rehabilitation time; S(t′) is the function expression of the patient's lung rehabilitation speed over time; VC(t′) is the function expression of the patient's vital capacity over time; t′ is the time required for the rehabilitation process; C is a constant term of a fixed influencing factor not included in the model;

[0058] Adjust the intensity of the rehabilitation plan based on the assessed patient's pulmonary rehabilitation time, combined with the patient's pulmonary rehabilitation speed and future pulmonary rehabilitation trends; if the rehabilitation time is long and the recovery speed is slow, increase the intensity of the rehabilitation plan; if the rehabilitation time is short and the recovery speed is fast, maintain or reduce the intensity of the rehabilitation plan.

[0059] The technical effects and advantages of the intelligent evaluation system for lung rehabilitation detection of the present invention are as follows:

[0060] The present invention can effectively manage and organize historical evaluation reference data by defining sets and constructing evaluation reference data graphs, making the data structure clearer; this structured data storage method facilitates subsequent data retrieval, analysis and maintenance; the Lagrange multiplier method and the optimization objective function can ensure that the matching degree between the reference patient physiological parameter data and the pulmonary rehabilitation test evaluation data is maximized, thereby improving the prediction accuracy of the model, effectively discovering and utilizing the intrinsic relationship between the data, and enhancing the reliability of the evaluation results; the weights of the edges in the data graph are adjusted by constraint conditions and penalty functions, which can effectively reflect the relative importance of different data points and prevent the complexity of the graph from being too high;

[0061] By calculating the first-order and second-order derivatives of the mean square error loss function, the update step size can be determined more effectively, thereby accelerating the convergence of the model. Using Newton's method combined with the inverse of the Hessian matrix can help find the optimal solution faster and reduce training time. Reasonable adjustment of hyperparameters can effectively prevent overfitting and underfitting problems of the model, making the model more generalizable. By dynamically adjusting the complexity and flexibility of the model, it can adapt to different data distributions. Continuously iterating and optimizing hyperparameters makes the model stable in multiple trainings, ensuring the reliability and repeatability of the model. By iteratively updating the limit formula, the amplitude of the hyperparameter update can be controlled, effectively preventing shocks during model training, and ensuring that each update is meaningful, which is especially important for large-scale data sets. Using the influencing factor adjustment formula, we can clearly understand the impact of each hyperparameter on the convergence speed of the model, which helps researchers and engineers make more rational adjustments and decisions.

[0062] By dynamically adjusting the intensity of the rehabilitation plan, the patient's rehabilitation can be promoted more effectively, unnecessary delays or excessive loads in the rehabilitation process can be avoided, and patients can be ensured to be able to recover in the best condition; the evaluation model provides a quantitative basis, allowing doctors to rely on data to make scientific decisions when formulating and adjusting rehabilitation plans, thereby improving the rationality and effectiveness of treatment. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a schematic diagram of the structure of an intelligent evaluation system for lung rehabilitation detection according to the present invention;

[0064] Figure 2 The figure is a flow chart of an intelligent evaluation method for lung rehabilitation testing according to the present invention. DETAILED DESCRIPTION

[0065] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions 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 creative work are within the scope of protection of the present invention.

[0066] Embodiment 1

[0067] See also Figure 1 As shown, this embodiment provides a lung rehabilitation detection intelligent assessment system, including:

[0068] A database construction module, used to construct an evaluation reference database based on historical evaluation reference data, wherein the evaluation reference data includes reference patient physiological parameter data and pulmonary rehabilitation test evaluation data;

[0069] A data collection module, used to collect pre-assessment patient physiological parameter data, chest imaging data, and air quality data;

[0070] A data processing module is used to obtain the physiological parameter data of the patient to be evaluated through the physiological parameter data of the pre-evaluated patient and the evaluation reference database; the acquired physiological parameter data, chest imaging data and air quality data of the patient to be evaluated are processed to obtain a comprehensive characteristic data set of the patient to be evaluated;

[0071] The pulmonary rehabilitation detection module is used to obtain a pulmonary status index prediction model through training of a comprehensive feature data set of the patient to be evaluated, and to predict the patient's pulmonary rehabilitation status index through the pulmonary status index prediction model;

[0072] A pulmonary rehabilitation assessment module is used to compare the predicted patient pulmonary rehabilitation status index with the preset patient pulmonary rehabilitation status index threshold to assess whether the patient's pulmonary rehabilitation index is abnormal;

[0073] The pulmonary rehabilitation trend assessment module generates early warning information through the pulmonary rehabilitation detection intelligent terminal if the patient's pulmonary rehabilitation indicators are abnormal, and promptly reminds the doctor; if the patient's pulmonary rehabilitation indicators are normal, the patient's future pulmonary rehabilitation trend is predicted based on the assessment reference database and the patient's pulmonary rehabilitation status indicators; the pulmonary rehabilitation detection intelligent terminal evaluates the patient's pulmonary rehabilitation time based on the patient's future pulmonary rehabilitation trend prediction results, and dynamically adjusts the rehabilitation plan;

[0074] The modules are connected to each other via wired and / or wireless means.

[0075] Methods for constructing an evaluation reference database based on historical evaluation reference data include:

[0076] Reference patient physiological parameter data include the patient's respiratory rate, blood oxygen saturation, heart rate, blood pressure, hemoglobin concentration and blood lactate level; pulmonary rehabilitation test evaluation data include the patient's vital capacity, maximum ventilation, forced expiratory volume in one second, respiratory airflow rate, carbon dioxide partial pressure and oxygen partial pressure;

[0077] S21. Define a set x to represent the reference patient physiological parameter data; the set x is represented by x = {x1, x2, ..., x p ,...,x M}; where x p is the pth type of patient physiological parameter data; p is the index of the patient physiological parameter data type, p=1,...,M; M is the number of patient physiological parameter data types; each data x in the set x p A node in the corresponding evaluation reference data graph;

[0078] Define set y to represent pulmonary rehabilitation test evaluation data; set y is represented by y = {y1, y2, ..., y q ,...,y N}; where y q is the qth type of pulmonary rehabilitation test evaluation data; q is the index of the type of pulmonary rehabilitation test evaluation data, q=1,...,N; N is the number of types of pulmonary rehabilitation test evaluation data; each data y in the set y q corresponds to another node in the evaluation reference data graph;

[0079] S22, constructing a vertex set V=x∪y in the evaluation reference data graph through all reference patient physiological parameter data nodes and pulmonary rehabilitation test evaluation data nodes, connecting each reference patient physiological parameter data node and pulmonary rehabilitation test evaluation data node through the edge weight ω obtained by calculating the Pearson correlation coefficient, constructing an edge set E of the evaluation reference data graph, and finally obtaining an evaluation reference data graph model G=(V,E);

[0080] S23. Define an optimization objective function and introduce constraints in the evaluation reference data graph; process the constraints in the evaluation reference data graph by introducing the Lagrange multiplier method formula, and ensure that the weight distribution of the evaluation reference data graph meets the optimization goal. The preset optimization goal is the matching degree between the reference patient physiological parameter data and the pulmonary rehabilitation test evaluation data; the Lagrange multiplier method formula is:

[0081] Among them, x i′ is the physiological parameter data of the i′th reference patient; i′is the ith pulmonary rehabilitation test evaluation data; Sim(x i′ ,y i′ ) is used to measure x i′ and i′ The optimization objective function of the matching degree between i′ To measure each pair of x i′ and i′ The weight of; i′ is the index of different data types; n′ is the total number of reference patient physiological parameter data and pulmonary rehabilitation test evaluation data; σ k The kth Lagrange multiplier for the trade-off between the optimization objective function and the constraints in the evaluation reference data graph; g k (x, y) is the constraint condition in the evaluation reference data graph; j is the number of constraints in the evaluation reference data graph; σ is the set of Lagrange multipliers, including all σ k ; k is the index of the constraint in the evaluation reference data graph;

[0082] S24. The number of constraints in the evaluation reference data graph is limited by a constraint number limiting formula. The constraint number limiting formula is: Wherein, j′ is the number of constraints in the restricted evaluation reference data graph; b is the total number of types of reference patient physiological parameter data and pulmonary rehabilitation test evaluation data;

[0083] By quantifying and controlling the number of constraints, this formula can support decision making. When tuning the model, researchers can evaluate whether the current settings are reasonable based on the calculated j′ value and make necessary adjustments;

[0084] For example, suppose that the data of a pulmonary rehabilitation assessment system is being processed. There are 3 types of reference patient physiological parameter data (such as respiratory rate, blood oxygen saturation, heart rate) and 3 types of pulmonary rehabilitation test assessment data (such as vital capacity, maximum ventilation, and forced expiratory volume in one second) in the system, a total of 6 parameter pairs. To make this formula more understandable, assume that the weight and number of samples for each parameter pair are as follows:

[0085] The total number of reference patient physiological parameter data and pulmonary rehabilitation test evaluation data is 10, the patient's respiratory rate is 18, blood oxygen saturation is 96, heart rate is 75, vital capacity is 3.5, maximum ventilation volume is 100, forced expiratory volume in one second is 4, weight ω1 is 0.5, weight ω2 is 0.3, weight ω3 is 0.2, then the number of constraints in the restricted evaluation reference data diagram is indivual.

[0086] S25. Define a penalty function corresponding to the constraint conditions in the evaluation reference data graph, and use the penalty function to punish the solution that does not meet the constraint conditions in the evaluation reference data graph. The penalty function is Among them, P(x,y) is the penalty function; ρ k To control the penalty coefficient for the degree of violation of the constraints in the evaluation reference data graph;

[0087] S26. Through the API interface provided in the database, the generated evaluation reference data graph is imported into the database, thereby constructing a structured evaluation reference database.

[0088] The pre-assessment patient physiological parameter data include the patient's respiratory rate, blood oxygen saturation, heart rate, blood pressure, hemoglobin concentration and blood lactate level; the chest imaging data include the patient's chest X-ray data, chest CT scan image data and magnetic resonance imaging data; the air quality data include the air pollution index of the patient's living environment.

[0089] The method for obtaining the physiological parameter data of the patient to be evaluated by pre-evaluating the physiological parameter data of the patient and evaluating the reference database includes:

[0090] The pre-evaluation patient physiological parameter data is matched with the historical evaluation reference data in the evaluation reference database by using cosine similarity calculation, the reference patient physiological parameter data that is most similar to the pre-evaluation patient physiological parameter data is found, and the physiological parameter data of the patient to be evaluated is obtained by integration.

[0091] The method of processing the acquired physiological parameter data, chest image data and air quality data of the patient to be evaluated to obtain a comprehensive characteristic data set of the patient to be evaluated includes:

[0092] Use the K-means clustering algorithm to identify outliers in the patient's physiological parameter data, chest imaging data, and air quality data, and remove the identified outliers to obtain processed patient physiological characteristic data sets, chest imaging characteristic data sets, and air quality characteristic data sets;

[0093] The acquired patient physiological characteristic data set, chest imaging characteristic data set and air quality characteristic data set are normalized by standard deviation and converted into a standard normal distribution to obtain a normalized patient physiological characteristic data set, chest imaging characteristic data set and air quality characteristic data set;

[0094] The normalized patient physiological characteristic dataset, chest imaging characteristic dataset and air quality characteristic dataset are fused through a weighted formula to form a comprehensive characteristic dataset of the patient to be evaluated. The patient physiological characteristic dataset is recorded as E1, the chest imaging characteristic dataset is recorded as E2, and the air quality characteristic dataset is recorded as E3;

[0095] The weighted formula is: Hc=E1·δ1+E2·δ2+E3·δ3; where δ1 is the weight coefficient of the patient's physiological characteristic data set; δ2 is the weight coefficient of the chest image characteristic data set; and δ3 is the weight coefficient of the air quality characteristic data set.

[0096] The training method of the lung status index prediction model includes:

[0097] The data set is divided into a training set, a validation set, and a test set, and a lung status indicator prediction model is constructed; the sample set is a subset of the data set, and each sample set includes a historical comprehensive feature data set of patients to be evaluated and the corresponding patient lung rehabilitation status indicators; the input data of the model is a historical comprehensive feature data set of patients to be evaluated; the output label is the patient lung rehabilitation status indicator; the lung status indicator prediction model is a gradient boosting decision tree model;

[0098] Initialize the lung status indicator prediction model and set the number of trees, learning rate, and tree depth hyperparameters. Use k-fold cross-validation to train the model using different hyperparameter combinations. Use the cross-validation results to tune the initially set hyperparameters and select the best performing hyperparameter combination.

[0099] The mean square error is used as the loss function to measure the difference between the model's predicted value and the actual value; the mean square error loss function is: Where L(θ) is the error of the current model under the hyperparameter θ; n is the number of samples in the data set; Y i is the true pulmonary rehabilitation status indicator of the i-th sample in the dataset; is the lung rehabilitation status index predicted by the model's hyperparameter θ for the i-th sample in the data set; i is the index of the sample; θ is the model's hyperparameter combination;

[0100] Use the training set to train the lung status indicator prediction model, and train a new decision tree based on the residual of the current model in each iteration; adjust the model's hyperparameters to tune the model according to the model performance feedback, and retrain the model with the adjusted hyperparameters. After the entire training process is completed, cross-validation is used to select the most appropriate hyperparameters;

[0101] When the training reaches the preset model complexity, it stops and the final trained lung status index prediction model is obtained; the trained lung status index prediction model is used to predict the comprehensive feature data set of the current patient to be evaluated to obtain the patient's lung rehabilitation status index.

[0102] Methods for adjusting the model's hyperparameters to optimize the model include:

[0103] S71, calculating the first-order derivative and the second-order derivative of the mean square error loss function with respect to the hyperparameter, determining the change gradient of the mean square error loss function through the first-order derivative, determining the step length of moving along the gradient direction when updating the parameters through the second-order derivative, and accelerating the convergence speed of the mean square error loss function;

[0104] The formula for calculating the first-order derivative is: Among them, ▽L(θ) is the first-order derivative of the mean square error loss function with respect to the hyperparameter θ, that is, the gradient; is the first-order partial derivative of the mean square error loss function with respect to the hyperparameter θ;

[0105] The second-order derivative calculation formula is: Among them, H(θ) is the second-order derivative of the mean square error loss function with respect to the hyperparameter θ, and the second-order derivative matrix is ​​the Hessian matrix; is the second-order partial derivative of the mean square error loss function with respect to the hyperparameter θ;

[0106] S72. In each iteration, the hyperparameter update formula of the Newton method is used to update the hyperparameters of the model according to the current gradient and the inverse of the Hessian matrix to reduce the L(θ) error;

[0107] The Newton method hyperparameter update formula is: t+1 =θ t -H(θ t ) -1 ▽L(θ t ), where θ t+1 is the hyperparameter value after the t+1th iteration update; θ t is the hyperparameter value at the current tth iteration; H(θ t ) -1 is the inverse matrix of the Hessian matrix at the current t-th iteration; ▽L(θ t ) is the current hyperparameter θ t The gradient under; t is the number of iterations;

[0108] S73. Continuously iterate the Newton method hyperparameter update formula to adjust the hyperparameters, limit the iterative update of the hyperparameters through the iterative update limit formula, and stop iterating when the hyperparameters of the model tend to be stable;

[0109] The iterative update limit formula is: in, is the amplitude of hyperparameter update; n is the number of samples in the data set; T is the number of new decision trees; α is a constant for adjusting the convergence speed; β is a factor for adjusting the influence of the number of samples on the convergence speed; γ is a factor for adjusting the influence of the number of decision trees on the convergence speed;

[0110] S74. The influencing factors β and γ used to adjust the number of samples and the number of decision trees on the convergence speed are adjusted by the influencing factor adjustment formula to optimize the values ​​of β and γ. The influencing factor adjustment formula is: ln(n β )+ln(T γ)=μ+λ·ln(I)+υ·ln(l); where λ is the coefficient of the influencing factor β; υ is the coefficient of the influencing factor γ; μ is the basic convergence speed of the model; I is the number of iterations of the model; l is the step size moving along the gradient direction when updating the parameters;

[0111] S75. The sum of the coefficient λ of the influencing factor β and the coefficient υ of the influencing factor γ is 1. By keeping the β value unchanged, increasing or decreasing the γ value, observing the convergence speed of the model, and then keeping the γ value unchanged, increasing or decreasing the β value, when the β and γ values ​​are adjusted to the optimal level, the convergence speed of the model reaches the best, and the amplitude of the hyperparameter update reaches the optimal level. The hyperparameter combination at this time is the optimal hyperparameter combination.

[0112] Explain this equation: Left side: Indicates the relative change of the hyperparameters in the current iteration. The smaller it is, the smaller the amplitude of parameter update is, and the adjustment of the model has stabilized. Right: It is a dynamic convergence condition, which determines whether the iteration should be stopped according to the number of samples and the number of decision trees. As n and T increase, the value on the right becomes smaller, that is, the hyperparameter update needs to become more stringent to meet the stopping condition.

[0113] This equation shows that when the relative change of the hyperparameter When the right side is equal to the right side, stop training. This dynamic stopping condition can adapt to the model complexity and data size to avoid overfitting or underfitting.

[0114] The method of comparing the predicted patient's pulmonary rehabilitation status index with the preset patient's pulmonary rehabilitation status index threshold to evaluate whether the patient's pulmonary rehabilitation index is abnormal includes:

[0115] If the predicted patient's lung rehabilitation status index is less than or equal to a preset patient's lung rehabilitation status index threshold, the patient's lung rehabilitation index is determined to be normal;

[0116] If the predicted patient's lung rehabilitation status index is greater than the preset patient's lung rehabilitation status index threshold, the patient's lung rehabilitation index is determined to be abnormal.

[0117] Based on the evaluation reference database and the patient's pulmonary rehabilitation status indicators, the methods for predicting the patient's future pulmonary rehabilitation trend include:

[0118] The data set is divided into a training set, a validation set, and a test set to build a pulmonary rehabilitation trend prediction model; the sample set is a subset of the data set, and each sample set includes a historical reference database and a patient's pulmonary rehabilitation status indicator as well as the corresponding patient's future pulmonary rehabilitation trend; the deep learning library TensorFlow is used to build a pulmonary rehabilitation trend prediction model; the pulmonary rehabilitation trend prediction model includes an input layer, an LSTM layer, and an output layer; the pulmonary rehabilitation trend prediction model is an LSTM model;

[0119] The L2 regularized mean square error is defined as the loss function of the model to measure the difference between the future pulmonary rehabilitation trend of the patient output by the LSTM layer and the actual future pulmonary rehabilitation trend of the patient; the L2 regularized mean square error loss function is:

[0120] Where N′ is the number of samples; is the predicted value of the ath sample; Y a is the actual value of the ath sample; η is the regularization coefficient, which controls the regularization strength; w u is the trainable parameter of the model; u is the index of the training parameter;

[0121] The gradient of the loss function with respect to the model parameters is calculated through the chain rule, and the Adam optimizer is used to update the parameters and tune the model. The performance of the model in the prediction task is evaluated through the test set. When the training reaches the preset model complexity, it is stopped to obtain a trained pulmonary rehabilitation trend prediction model. The trained pulmonary rehabilitation trend prediction model is used to predict the current reference database and the patient's pulmonary rehabilitation status indicators to obtain the patient's future pulmonary rehabilitation trend.

[0122] Methods for evaluating the patient's pulmonary rehabilitation time and dynamically adjusting the rehabilitation plan based on the patient's future pulmonary rehabilitation trend prediction results include:

[0123] The prediction results of the patient's future pulmonary rehabilitation trend include the pulmonary rehabilitation trends at z future time points; a pulmonary rehabilitation time evaluation model is constructed to evaluate the patient's pulmonary rehabilitation time according to the prediction results of the patient's future pulmonary rehabilitation trend; the pulmonary rehabilitation time evaluation model is:

[0124] Among them, T′ is the assessed lung rehabilitation time of the patient; S is the lung rehabilitation speed of the patient; AP is the air pollution index of the patient's living environment; VC is the patient's vital capacity; ξ1 is the weight coefficient for adjusting the influence of the patient's lung rehabilitation speed on the patient's lung rehabilitation time; ξ2 is the weight parameter for controlling the influence of the air pollution index and vital capacity on the patient's lung rehabilitation time; S(t′) is the functional expression of the patient's lung rehabilitation speed over time; VC(t′) is the functional expression of the patient's vital capacity over time; t′ is the time required for the rehabilitation process; C is a constant term of a fixed influencing factor not included in the model, which can reflect the patient's age, medical history, physical condition, living habits, etc. These factors will have an impact on the rehabilitation time, but it is difficult to directly quantify them with the variables in the formula, so they are expressed by the constant C;

[0125] For example, the function expression of the patient's lung recovery speed changing with time is: S(t′)=0.03×(1-0.01t); the function expression of the patient's vital capacity changing with time is: VC(t′)=3+0.001t; the air pollution index AP of the patient's living environment is 100, the weight coefficient ξ1 for adjusting the patient's lung recovery speed on the patient's lung recovery time is 0.4, the weight parameter ξ2 for controlling the influence of the air pollution index and vital capacity on the patient's lung recovery time is 0.6, the constant term C of the fixed influencing factors not included in the model is 20, and the time t′ required for the recovery process is 30, then the estimated patient's lung recovery time T′ is 0.0084+2.0976+0.25+20≈22 days.

[0126] Adjust the intensity of the rehabilitation plan based on the assessed patient's pulmonary rehabilitation time, combined with the patient's pulmonary rehabilitation speed and future pulmonary rehabilitation trends; if the rehabilitation time is long and the recovery speed is slow, increase the intensity of the rehabilitation plan; if the rehabilitation time is short and the recovery speed is fast, maintain or reduce the intensity of the rehabilitation plan.

[0127] The preset threshold of the patient's lung rehabilitation status indicator is set by the staff. The lung rehabilitation status indicators of different patients are collected through the lung rehabilitation detection intelligent terminal, and the average value of the lung rehabilitation status indicators of multiple patients is taken as the preset threshold of the patient's lung rehabilitation status indicator.

[0128] In this embodiment, by defining a set and constructing an evaluation reference data graph, historical evaluation reference data can be effectively managed and organized, making the data structure clearer; this structured data storage method facilitates subsequent data retrieval, analysis and maintenance; the Lagrange multiplier method and the optimization objective function can ensure that the matching degree between the reference patient physiological parameter data and the pulmonary rehabilitation test evaluation data is maximized, thereby improving the prediction accuracy of the model, effectively discovering and utilizing the intrinsic relationship between the data, and enhancing the reliability of the evaluation results; by adjusting the weights of the edges in the data graph through constraints and penalty functions, the relative importance of different data points can be effectively reflected to prevent the complexity of the graph from being too high;

[0129] By calculating the first-order and second-order derivatives of the mean square error loss function, the update step size can be determined more effectively, thereby accelerating the convergence of the model. Using Newton's method combined with the inverse of the Hessian matrix can help find the optimal solution faster and reduce training time. Reasonable adjustment of hyperparameters can effectively prevent overfitting and underfitting problems of the model, making the model more generalizable. By dynamically adjusting the complexity and flexibility of the model, it can adapt to different data distributions. Continuously iterating and optimizing hyperparameters makes the model stable in multiple trainings, ensuring the reliability and repeatability of the model. By iteratively updating the limit formula, the amplitude of the hyperparameter update can be controlled, effectively preventing shocks during model training, and ensuring that each update is meaningful, which is especially important for large-scale data sets. Using the influencing factor adjustment formula, we can clearly understand the impact of each hyperparameter on the convergence speed of the model, which helps researchers and engineers make more rational adjustments and decisions.

[0130] By dynamically adjusting the intensity of the rehabilitation plan, the patient's rehabilitation can be promoted more effectively, unnecessary delays or excessive loads in the rehabilitation process can be avoided, and patients can be ensured to be able to recover in the best condition; the evaluation model provides a quantitative basis, allowing doctors to rely on data to make scientific decisions when formulating and adjusting rehabilitation plans, thereby improving the rationality and effectiveness of treatment.

[0131] Embodiment 2

[0132] See also Figure 2 As shown, the part not described in detail in this embodiment is described in Example 1, which provides an intelligent evaluation method for lung rehabilitation detection, including:

[0133] S1. Construct an evaluation reference database based on historical evaluation reference data, where the evaluation reference data includes reference patient physiological parameter data and pulmonary rehabilitation test evaluation data;

[0134] S2, collect pre-assessment patient physiological parameter data, chest imaging data and air quality data;

[0135] S3. Acquire the physiological parameter data of the patient to be evaluated through the physiological parameter data of the pre-evaluated patient and the evaluation reference database; process the acquired physiological parameter data, chest imaging data and air quality data of the patient to be evaluated to obtain a comprehensive characteristic data set of the patient to be evaluated;

[0136] S4. A lung state index prediction model is obtained by training the comprehensive feature data set of the patient to be evaluated, and the lung rehabilitation state index of the patient is predicted by the lung state index prediction model;

[0137] S5, comparing the predicted patient's pulmonary rehabilitation status index with a preset patient's pulmonary rehabilitation status index threshold to assess whether the patient's pulmonary rehabilitation index is abnormal;

[0138] S6. If the patient's lung rehabilitation indicators are abnormal, the lung rehabilitation detection intelligent terminal will generate early warning information to remind the doctor in time; if the patient's lung rehabilitation indicators are normal, the patient's future lung rehabilitation trend will be predicted based on the evaluation reference database and the patient's lung rehabilitation status indicators; the lung rehabilitation detection intelligent terminal will evaluate the patient's lung rehabilitation time based on the patient's future lung rehabilitation trend prediction results, and dynamically adjust the rehabilitation plan.

[0139] Since the electronic device introduced in this embodiment is an electronic device used to implement a lung rehabilitation detection intelligent evaluation system in the embodiment of this application, based on the lung rehabilitation detection intelligent evaluation system introduced in the embodiment of this application, the technical personnel of this field can understand the specific implementation of the electronic device of this embodiment and its various variations, so how the electronic device implements the method in the embodiment of this application is not described in detail here. As long as the technical personnel of this field implement the electronic device used in the lung rehabilitation detection intelligent evaluation system in the embodiment of this application, it belongs to the scope of protection of this application.

[0140] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and thresholds in the formula are set by technicians in this field according to actual conditions.

[0141] The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technical users in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A pulmonary rehabilitation detection intelligent assessment system, characterized in that: include: A database construction module, used to construct an evaluation reference database based on historical evaluation reference data, wherein the evaluation reference data includes reference patient physiological parameter data and pulmonary rehabilitation test evaluation data; A data collection module, used to collect pre-assessment patient physiological parameter data, chest imaging data, and air quality data; A data processing module, used for obtaining the physiological parameter data of the patient to be evaluated through the physiological parameter data of the pre-evaluated patient and the evaluation reference database; Processing the acquired physiological parameter data, chest imaging data and air quality data of the patient to be evaluated to obtain a comprehensive characteristic data set of the patient to be evaluated; The pulmonary rehabilitation detection module is used to obtain a pulmonary status index prediction model through training of a comprehensive feature data set of the patient to be evaluated, and to predict the patient's pulmonary rehabilitation status index through the pulmonary status index prediction model; A pulmonary rehabilitation assessment module is used to compare the predicted patient pulmonary rehabilitation status index with the preset patient pulmonary rehabilitation status index threshold to assess whether the patient's pulmonary rehabilitation index is abnormal; The pulmonary rehabilitation trend assessment module generates early warning information through the pulmonary rehabilitation detection intelligent terminal to remind doctors in time if the patient's pulmonary rehabilitation indicators are abnormal; If the patient's pulmonary rehabilitation indicators are normal, the patient's future pulmonary rehabilitation trend is predicted based on the evaluation reference database and the patient's pulmonary rehabilitation status indicators; The pulmonary rehabilitation detection intelligent terminal estimates the patient's pulmonary rehabilitation time based on the patient's future pulmonary rehabilitation trend prediction results, and dynamically adjusts the rehabilitation plan; The modules are connected to each other via wired and / or wireless means.

2. The pulmonary rehabilitation detection intelligent assessment system according to claim 1, characterized in that: The method for constructing an evaluation reference database based on historical evaluation reference data comprises: The reference patient physiological parameter data include the patient's respiratory rate, blood oxygen saturation, heart rate, blood pressure, hemoglobin concentration and blood lactate level; the pulmonary rehabilitation test assessment data include the patient's vital capacity, maximum ventilation volume, forced expiratory volume in one second, respiratory airflow rate, carbon dioxide partial pressure and oxygen partial pressure; S21. Define a set x to represent the reference patient physiological parameter data; the set x is represented by x = {x1, x2, ..., x p ,...,x M }; where x p is the pth type of patient physiological parameter data; p is the index of the patient physiological parameter data type, p=1,...,M; M is the number of patient physiological parameter data types; each data x in the set x p A node in the corresponding evaluation reference data graph; Define set y to represent pulmonary rehabilitation test evaluation data; set y is represented by y = {y1, y2, ..., y q ,...,y N }; where y q is the qth type of pulmonary rehabilitation test evaluation data; q is the index of the type of pulmonary rehabilitation test evaluation data, q=1,...,N; N is the number of types of pulmonary rehabilitation test evaluation data; each data y in the set y q corresponds to another node in the evaluation reference data graph; S22, constructing a vertex set V=x∪y in the evaluation reference data graph through all reference patient physiological parameter data nodes and pulmonary rehabilitation test evaluation data nodes, connecting each reference patient physiological parameter data node and pulmonary rehabilitation test evaluation data node through the edge weight ω obtained by calculating the Pearson correlation coefficient, constructing an edge set E of the evaluation reference data graph, and finally obtaining an evaluation reference data graph model G=(V,E); S23, defining an optimization objective function, and introducing constraints in the evaluation reference data graph; processing the constraints in the evaluation reference data graph by introducing the Lagrange multiplier method formula, and ensuring that the weight distribution of the evaluation reference data graph meets the optimization goal, the preset optimization goal is the matching degree between the reference patient physiological parameter data and the pulmonary rehabilitation test evaluation data; the Lagrange multiplier method formula is: Among them, x i′ is the physiological parameter data of the i′th reference patient; i′ is the ith pulmonary rehabilitation test evaluation data; Sim(x i′ ,y i′ ) is used to measure x i′ and i′ The optimization objective function of the matching degree between i′ To measure each pair of x i′ and i′ The weight of; i′ is the index of different data types; n′ is the total number of reference patient physiological parameter data and pulmonary rehabilitation test evaluation data; σ k The kth Lagrange multiplier for the trade-off between the optimization objective function and the constraints in the evaluation reference data graph; g k (x, y) is the constraint condition in the evaluation reference data graph; j is the number of constraints in the evaluation reference data graph; σ is the set of Lagrange multipliers, including all σ k ; k is the index of the constraint in the evaluation reference data graph; S24, limiting the number of constraints in the evaluation reference data graph by a constraint number limiting formula, wherein the constraint number limiting formula is: Wherein, j′ is the number of constraints in the restricted evaluation reference data graph; b is the total number of types of reference patient physiological parameter data and pulmonary rehabilitation test evaluation data; S25, defining a penalty function corresponding to the constraint conditions in the evaluation reference data graph, and using the penalty function to punish solutions that do not satisfy the constraint conditions in the evaluation reference data graph, wherein the penalty function is Among them, P(x,y) is the penalty function; ρ k To control the penalty coefficient for the degree of violation of the constraints in the evaluation reference data graph; S26. Through the API interface provided in the database, the generated evaluation reference data graph is imported into the database, thereby constructing a structured evaluation reference database.

3. The intelligent evaluation system for pulmonary rehabilitation testing according to claim 2, characterized in that: The pre-assessed patient physiological parameter data includes the patient's respiratory rate, blood oxygen saturation, heart rate, blood pressure, hemoglobin concentration and blood lactate level; the chest imaging data includes the patient's chest X-ray data, chest CT scan image data and magnetic resonance imaging data; the air quality data includes the air pollution index of the patient's living environment.

4. The intelligent evaluation system for pulmonary rehabilitation testing according to claim 3, characterized in that: The method for obtaining the physiological parameter data of the patient to be evaluated by pre-evaluating the physiological parameter data of the patient and evaluating the reference database includes: The pre-evaluation patient physiological parameter data is matched with the historical evaluation reference data in the evaluation reference database by using cosine similarity calculation, the reference patient physiological parameter data that is most similar to the pre-evaluation patient physiological parameter data is found, and the physiological parameter data of the patient to be evaluated is obtained by integration.

5. The intelligent evaluation system for pulmonary rehabilitation testing according to claim 4, characterized in that: The method of processing the acquired physiological parameter data, chest image data and air quality data of the patient to be evaluated to obtain a comprehensive characteristic data set of the patient to be evaluated includes: Use the K-means clustering algorithm to identify outliers in the patient's physiological parameter data, chest imaging data, and air quality data, and remove the identified outliers to obtain processed patient physiological characteristic data sets, chest imaging characteristic data sets, and air quality characteristic data sets; The acquired patient physiological characteristic data set, chest imaging characteristic data set and air quality characteristic data set are normalized by standard deviation and converted into a standard normal distribution to obtain a normalized patient physiological characteristic data set, chest imaging characteristic data set and air quality characteristic data set; The normalized patient physiological characteristic dataset, chest imaging characteristic dataset and air quality characteristic dataset are fused through a weighted formula to form a comprehensive characteristic dataset of the patient to be evaluated. The patient physiological characteristic dataset is recorded as E1, the chest imaging characteristic dataset is recorded as E2, and the air quality characteristic dataset is recorded as E3; The weighted formula is: Hc=E1·δ1+E2·δ2+E3·δ3; wherein δ1 is the weight coefficient of the patient's physiological characteristic data set; δ2 is the weight coefficient of the chest image characteristic data set; and δ3 is the weight coefficient of the air quality characteristic data set.

6. The intelligent evaluation system for pulmonary rehabilitation testing according to claim 5, characterized in that: The training method of the lung status index prediction model includes: The data set is divided into a training set, a validation set and a test set, and a lung status index prediction model is constructed; the sample set is a subset of the data set, and each sample set includes a historical comprehensive feature data set of patients to be evaluated and a corresponding patient lung rehabilitation status index; the input data of the model is a historical comprehensive feature data set of patients to be evaluated; the output label is a patient lung rehabilitation status index; the lung status index prediction model is a gradient boosting decision tree model; Initialize the lung status indicator prediction model and set the number of trees, learning rate, and tree depth hyperparameters. Use k-fold cross-validation to train the model using different hyperparameter combinations. Use the cross-validation results to tune the initially set hyperparameters and select the best performing hyperparameter combination. Use mean square error as the loss function to measure the difference between the model's predicted value and the actual value; use the training set to train the lung status indicator prediction model, and train a new decision tree based on the residual of the current model in each iteration; adjust the model's hyperparameters to tune the model based on the model performance feedback, and retrain the model with the adjusted hyperparameters. After the entire training process is completed, cross-validation is used to select the most appropriate hyperparameters; When the training reaches the preset model complexity, it stops and the final trained lung status index prediction model is obtained; the trained lung status index prediction model is used to predict the comprehensive feature data set of the current patient to be evaluated to obtain the patient's lung rehabilitation status index.

7. The intelligent evaluation system for pulmonary rehabilitation testing according to claim 6, characterized in that: The method for adjusting the hyperparameters of the model to optimize the model includes: S71, calculating the first-order derivative and the second-order derivative of the mean square error loss function with respect to the hyperparameter, determining the change gradient of the mean square error loss function through the first-order derivative, determining the step length of moving along the gradient direction when updating the parameters through the second-order derivative, and accelerating the convergence speed of the mean square error loss function; The first-order derivative calculation formula is: Among them, ▽L(θ) is the first-order derivative of the mean square error loss function with respect to the hyperparameter θ, that is, the gradient; is the first-order partial derivative of the mean square error loss function with respect to the hyperparameter θ; The second-order derivative calculation formula is: Among them, H(θ) is the second-order derivative of the mean square error loss function with respect to the hyperparameter θ, and the second-order derivative matrix is ​​the Hessian matrix; is the second-order partial derivative of the mean square error loss function with respect to the hyperparameter θ; S72. In each iteration, the hyperparameter update formula of the Newton method is used to update the hyperparameters of the model according to the current gradient and the inverse of the Hessian matrix to reduce the L(θ) error; The Newton method hyperparameter update formula is: θ t+1 =θ t -H(θ t ) -1 ▽L(θ t ), where θ t+1 is the hyperparameter value after the t+1th iteration update; θ t is the hyperparameter value at the current tth iteration; H(θ t ) -1 is the inverse matrix of the Hessian matrix at the current t-th iteration; ▽L(θ t ) is the current hyperparameter θ t The gradient under; t is the number of iterations; S73. Continuously iterate the Newton method hyperparameter update formula to adjust the hyperparameters, limit the iterative update of the hyperparameters through the iterative update limit formula, and stop iterating when the hyperparameters of the model tend to be stable; The iterative update restriction formula is: in, is the amplitude of hyperparameter update; n is the number of samples in the data set; T is the number of new decision trees; α is a constant for adjusting the convergence speed; β is a factor for adjusting the influence of the number of samples on the convergence speed; γ is a factor for adjusting the influence of the number of decision trees on the convergence speed; S74, adjusting the influencing factors β and γ used to adjust the number of samples and the number of decision trees on the convergence speed through the influencing factor adjustment formula, so that the values ​​of β and γ reach the optimal value; the influencing factor adjustment formula is: ln(n β )+ln(T γ )=μ+λ·ln(I)+υ·ln(l); where λ is the coefficient of the influencing factor β; υ is the coefficient of the influencing factor γ; μ is the basic convergence speed of the model; I is the number of iterations of the model; l is the step size moving along the gradient direction when updating the parameters; S75. The sum of the coefficient λ of the influencing factor β and the coefficient υ of the influencing factor γ is 1. By keeping the β value unchanged, increasing or decreasing the γ value, observing the convergence speed of the model, and then keeping the γ value unchanged, increasing or decreasing the β value, when the β and γ values ​​are adjusted to the optimal level, the convergence speed of the model is optimal, and the amplitude of the hyperparameter update is optimal. The hyperparameter combination at this time is the optimal hyperparameter combination.

8. The intelligent evaluation system for pulmonary rehabilitation testing according to claim 7, characterized in that: The method of comparing the predicted patient's pulmonary rehabilitation status index with a preset patient's pulmonary rehabilitation status index threshold to evaluate whether the patient's pulmonary rehabilitation status index is abnormal includes: If the predicted patient's lung rehabilitation status index is less than or equal to a preset patient's lung rehabilitation status index threshold, the patient's lung rehabilitation index is determined to be normal; If the predicted patient's lung rehabilitation status index is greater than the preset patient's lung rehabilitation status index threshold, the patient's lung rehabilitation index is determined to be abnormal.

9. The intelligent evaluation system for lung rehabilitation testing according to claim 8, characterized in that: The method for predicting the future pulmonary rehabilitation trend of a patient based on the evaluation reference database and the patient's pulmonary rehabilitation status index includes: The data set is divided into a training set, a validation set and a test set, and a pulmonary rehabilitation trend prediction model is constructed; the sample set is a subset of the data set, and each sample set includes a historical reference database and a patient's pulmonary rehabilitation status indicator and the corresponding patient's future pulmonary rehabilitation trend; the pulmonary rehabilitation trend prediction model is constructed using the deep learning library TensorFlow; the pulmonary rehabilitation trend prediction model includes an input layer, an LSTM layer and an output layer; the pulmonary rehabilitation trend prediction model is an LSTM model; Define L2 regularized mean square error as the loss function of the model to measure the difference between the patient's future pulmonary rehabilitation trend output by the LSTM layer and the actual patient's future pulmonary rehabilitation trend; The gradient of the loss function with respect to the model parameters is calculated through the chain rule, and the Adam optimizer is used to update the parameters and tune the model. The performance of the model in the prediction task is evaluated through the test set. When the training reaches the preset model complexity, it is stopped to obtain a trained pulmonary rehabilitation trend prediction model. The trained pulmonary rehabilitation trend prediction model is used to predict the current reference database and the patient's pulmonary rehabilitation status indicators to obtain the patient's future pulmonary rehabilitation trend.

10. The intelligent evaluation system for pulmonary rehabilitation testing according to claim 9, characterized in that: The method of evaluating the patient's pulmonary rehabilitation time and dynamically adjusting the rehabilitation plan based on the patient's future pulmonary rehabilitation trend prediction results includes: The patient's future pulmonary rehabilitation trend prediction result includes the pulmonary rehabilitation trends at z future time points; a pulmonary rehabilitation time evaluation model is constructed to evaluate the patient's pulmonary rehabilitation time according to the patient's future pulmonary rehabilitation trend prediction result; the pulmonary rehabilitation time evaluation model is: Among them, T′ is the assessed lung rehabilitation time of the patient; S is the lung rehabilitation speed of the patient; AP is the air pollution index of the patient's living environment; VC is the vital capacity of the patient; ξ1 is the weight coefficient for adjusting the influence of the patient's lung rehabilitation speed on the patient's lung rehabilitation time; ξ2 is the weight parameter for controlling the influence of air pollution index and vital capacity on the patient's lung rehabilitation time; S(t′) is the function expression of the patient's lung rehabilitation speed over time; VC(t′) is the function expression of the patient's vital capacity over time; t′ is the time required for the rehabilitation process; C is a constant term of a fixed influencing factor not included in the model; Adjust the intensity of the rehabilitation plan based on the assessed patient's pulmonary rehabilitation time, combined with the patient's pulmonary rehabilitation speed and future pulmonary rehabilitation trends; if the rehabilitation time is long and the recovery speed is slow, increase the intensity of the rehabilitation plan; if the rehabilitation time is short and the recovery speed is fast, maintain or reduce the intensity of the rehabilitation plan.

Images