Orthopedic postoperative rehabilitation node prediction method based on improved ant colony algorithm

Abstract

The invention discloses an orthopedic postoperative rehabilitation node prediction method based on an improved ant colony algorithm. The method comprises the following steps: S1, extracting postoperative pain level features, limb function recovery progress features and exercise ability evaluation data features of a postoperative patient; s2, generating a standardized postoperative patient feature data set; s3, generating a weight matrix containing each feature weight; s4, initializing parameters of the improved ant colony algorithm based on the standardized feature data set and the weight matrix of the postoperative patient; s5, performing path search on the postoperative rehabilitation nodes by adopting an improved ant colony algorithm; s6, taking the result of the improved ant colony algorithm path search as the prediction result of the postoperative rehabilitation node; and S7, generating a postoperative rehabilitation plan of the patient based on a prediction result. According to the method, obvious improvement is achieved in orthopedic postoperative rehabilitation node prediction, the rehabilitation efficiency of a patient can be effectively improved, and the recurrence risk is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of orthopedics, and in particular to an orthopedic postoperative rehabilitation node prediction method based on an improved ant colony algorithm. Background Art

[0002] With the development of artificial intelligence and optimization algorithms, the medical field has gradually introduced intelligent tools to assist diagnosis and treatment planning. In the field of orthopedic postoperative rehabilitation, the prediction and planning of rehabilitation nodes is an important part of achieving personalized rehabilitation. Accurate prediction of rehabilitation nodes can help doctors formulate scientific rehabilitation plans, improve patients' rehabilitation efficiency and reduce the risk of recurrence.

[0003] At present, the prediction of postoperative rehabilitation nodes in orthopedics mainly relies on the doctors' clinical experience and simple static assessment methods. The static assessment method usually develops a rehabilitation plan by observing the patient's postoperative recovery and referring to the standardized rehabilitation process. Although the traditional experience-based method has certain practical value, its limitations are very significant: on the one hand, since the recovery process of each patient varies significantly due to individual differences, it is difficult to accurately predict the time and functional goals of key rehabilitation nodes based on experience alone; on the other hand, the patient's postoperative recovery is a dynamic, multi-factorial process. Traditional static assessment methods make it difficult to track the patient's recovery progress in real time and dynamically adjust the rehabilitation plan, which may cause the patient to miss the best time for rehabilitation intervention.

[0004] In recent years, prediction methods based on artificial intelligence have begun to be applied in the medical field. Some studies have attempted to use machine learning and deep learning techniques to analyze patients' postoperative data to assist in the prediction of rehabilitation nodes. However, the existing methods still have the following problems: On the one hand, the existing artificial intelligence methods have limited processing capabilities for high-dimensional medical data, which can easily lead to problems of model overfitting or insufficient prediction accuracy; on the other hand, many prediction methods fail to combine patients' real-time monitoring data for dynamic adjustments, making it difficult to cope with the complex changes in patients' recovery process.

[0005] In summary, the existing technology has significant deficiencies in the dynamics, personalization and multi-objective optimization of orthopedic postoperative rehabilitation node prediction, and cannot meet the needs of modern precision medicine for intelligent rehabilitation prediction. Therefore, there is an urgent need for an intelligent method that can combine the patient's multi-dimensional postoperative data, dynamically adjust the rehabilitation plan, and efficiently plan rehabilitation nodes through optimization algorithms to solve the above problems. Summary of the invention

[0006] One purpose of the present invention is to propose a method for predicting orthopedic postoperative rehabilitation nodes based on an improved ant colony algorithm. The present invention has achieved significant improvements in the prediction of orthopedic postoperative rehabilitation nodes, which can effectively improve the patient's rehabilitation efficiency and reduce the risk of recurrence.

[0007] According to an embodiment of the present invention, a method for predicting orthopedic postoperative rehabilitation nodes based on an improved ant colony algorithm comprises the following steps:

[0008] S1. Obtain multi-dimensional data of postoperative patients, extract postoperative pain level characteristics, limb function recovery progress characteristics and motor ability assessment data characteristics of postoperative patients;

[0009] S2. Perform data cleaning, outlier processing, data standardization, and missing value filling for postoperative pain level characteristics, limb function recovery progress characteristics, and motor ability assessment data characteristics to generate a standardized postoperative patient characteristic data set;

[0010] S3. Use the entropy weight method to assign weights to the importance of postoperative pain level features, limb function recovery progress features, and motor ability assessment data features in the standardized postoperative patient feature data set, and generate a weight matrix containing the weights of each feature;

[0011] S4. Initialize the parameters of the improved ant colony algorithm based on the standardized feature dataset and weight matrix of postoperative patients;

[0012] S5. Use improved ant colony algorithm to search the path of postoperative rehabilitation nodes;

[0013] S6. The result of the improved ant colony algorithm path search is used as the prediction result of the postoperative rehabilitation node;

[0014] S7. Generate a postoperative rehabilitation plan for the patient based on the prediction results, including the functional recovery goals of each rehabilitation node, node schedule, and related rehabilitation training recommendations.

[0015] Optionally, the S1 specifically includes:

[0016] S11. Collect multi-dimensional data of the postoperative patient through the postoperative patient monitoring system, the multi-dimensional data including:

[0017] Postoperative physiological index monitoring data, including heart rate, blood oxygen saturation, muscle electrical activity level, and inflammatory marker level;

[0018] Postoperative imaging data, wherein the postoperative imaging data is medical imaging of a specific part of a patient after surgery;

[0019] Postoperative activity assessment data, including range of motion of joints, gait stability score and movement balance ability;

[0020] S12. The postoperative pain level feature P is extracted from the postoperative physiological index monitoring data by using the feature analysis method, the key areas in the image are identified and segmented by using the image analysis algorithm for the postoperative imaging data, and the limb function recovery progress feature F is calculated, and the motion data analysis method is used to calculate the motion ability assessment data feature M for the postoperative activity ability assessment data;

[0021] S13. Based on the extracted postoperative pain level feature P, limb function recovery progress feature F and motor ability assessment data feature M of postoperative patients, a postoperative patient feature dataset D is constructed:

[0022] D={P,F,M}.

[0023] Optionally, the S2 specifically includes:

[0024] S21. Perform data cleaning on the postoperative patient feature data set D to remove obvious outliers in the postoperative pain level feature P, limb function recovery progress feature F, and motor ability assessment data feature M. The determination basis of the outliers is:

[0025] Outliers = {x i ∣x i >Q3+1.5·IQR or x i <Q1-1.5·IQR};

[0026] Among them, x i is the characteristic data point, Q1 and Q3 are the first quartile and the third quartile of the data respectively, and IQR is the interquartile range;

[0027] S22. Perform replacement processing on the outliers detected in the cleaned postoperative patient characteristic data set, use the mean interpolation method to process the outliers of the postoperative pain level characteristics of the postoperative patients, and use the linear interpolation method to process the outliers of the limb function recovery progress characteristics and the motor ability assessment data characteristics;

[0028] S23. The Z-score standardization method was used to standardize the postoperative pain level characteristics, limb function recovery progress characteristics and motor ability assessment data characteristics of the postoperative patients in the postoperative patient characteristic data set;

[0029] S24. Perform missing value detection and completion on the standardized postoperative patient characteristic data set, detect the location of missing values ​​in the postoperative pain level characteristics, limb function recovery progress characteristics and motor ability assessment data characteristics of postoperative patients, and use the nearest neighbor interpolation method to complete the missing values;

[0030] S25. Integrate the postoperative patient characteristic dataset after data cleaning, outlier processing, data standardization and missing value filling into a standardized postoperative patient characteristic dataset:

[0031] D std = {P std ,F std ,M std}.

[0032] Optionally, the S3 specifically includes:

[0033] S31. Standardized postoperative patient characteristics dataset D std The multi-scale discreteness of each feature in is analyzed, and the multi-scale discreteness is calculated by combining the local and global distribution information to calculate the multi-scale discreteness of feature i for:

[0034]

[0035] in, is the local discreteness of the feature, is the global discreteness of the feature, α is the weight coefficient of the local and global discreteness;

[0036] S32. Calculate dynamic weighted information gain based on the change of feature importance during the dynamic rehabilitation process of postoperative patients

[0037]

[0038] Among them, G i is the static information gain value of the feature, ω i (t) is the time-related feature weight change factor, which is calculated based on the gradient change of the historical data of the rehabilitation node:

[0039]

[0040] in, is the recovery loss function at a specific time point t, x i is the eigenvalue, ω i (t) represents the influence of the feature on the rehabilitation outcome at time t;

[0041] S33. Combining multi-scale discreteness and dynamic weighted information gain Calculating spatiotemporal joint weights

[0042]

[0043] Where m is the total number of features, Indicates the multi-scale discreteness of feature i, Represents the dynamic weighted information gain of feature i;

[0044] S34. Based on spatiotemporal joint weight and Construct a multi-dimensional feature weight matrix W multi :

[0045]

[0046] Optionally, the S4 specifically includes:

[0047] S41. Based on the standardized postoperative patient feature data set and the multi-dimensional feature weight matrix, the search space of the rehabilitation node prediction task is defined. The search space consists of the rehabilitation key stage node set N and the time attribute set T:

[0048] N={n1,n2,…,n k},1≤i≤k;

[0049] T={t1,t2,…,t k},1≤i≤k;

[0050] Among them, n i represents the i-th key stage node in the rehabilitation process, t i Represents the node n i The corresponding time attribute, k is the total number of nodes, and the construction and time attributes of the nodes are determined by the feature values ​​in the standardized postoperative patient feature data set and the weight distribution of the multi-dimensional feature weight matrix;

[0051] S42. Initialize the initial pheromone value of each path according to the node set N and the time attribute set T

[0052]

[0053] Among them, C ij For node n i To node n j The comprehensive cost of the path is:

[0054]

[0055] in, and For node n i and node n i Normalized value on feature k of the standardized postoperative patient feature dataset;

[0056] S43. Initialize the global parameters of the improved ant colony algorithm, including:

[0057] The pheromone volatility coefficient ρ, which ranges from 0<ρ<1, is used to control the rate of pheromone forgetting;

[0058] Path heuristic factor

[0059] Maximum number of iterations G max , used to limit the number of searches of the algorithm;

[0060] The number of ants A is used to control the number of ants participating in each round of path search;

[0061] S44. Combined with the characteristics of the rehabilitation node prediction task, a dynamic rule improvement mechanism is introduced to update the path cost based on real-time feedback

[0062]

[0063] Among them, λ is the adjustment coefficient, ΔC ij Indicates the change in path cost updated based on real-time data;

[0064] Dynamically update the path heuristic factor η based on the time attribute set T between nodes ij .

[0065] Optionally, the S5 specifically includes:

[0066] S51. Based on the rehabilitation key stage node set N, time attribute set T, and multi-dimensional feature weight matrix W multi And the comprehensive path cost C ij Constructing an initial recovery pathway:

[0067] P init ={n1→n2→…→n k};

[0068] S52. According to the complexity of the rehabilitation node prediction task, combined with the path comprehensive cost C ij And the dynamics of feature changes between nodes, initialize the number of ants A(t):

[0069]

[0070] Among them, A0 is the basic number of ants, κ is the adjustment coefficient, ΔC ij (t) represents node n i and n j The change in the comprehensive cost of the path at time t is:

[0071]

[0072] in, and Represents node n i and node n j Normalized value over time on feature k, m is the total number of features;

[0073] S53. According to the initialized dynamic ant number A(t) and path selection probability p ij Determine the path selection probability of each ant as:

[0074]

[0075] Among them, τ ij (t) is the path pheromone concentration at the current iteration time, η ij (t) is the heuristic factor of the path, α1 and β are the weight parameters of pheromone and heuristic factor respectively, N i For node n i The set of reachable adjacent nodes of ;

[0076] The dynamic path selection rule preferentially explores rehabilitation nodes with lower path comprehensive costs;

[0077] S54. In each round of iteration, the number of ants A(t+1) is adjusted according to the changes in the comprehensive hanging cost of the paths between nodes and the real-time feedback:

[0078]

[0079] At the same time, the ant colony distribution strategy is dynamically adjusted to make the ants concentrate in the path areas with large changes in comprehensive costs;

[0080] S55. Based on the path sequence generated by the dynamic path selection rule, the pheromone concentration τ on the path is updated after each round of iteration ij (t+1):

[0081] τ ij (t+1)=(1-ρ)·τ ij (t)+Δτ ij (t);

[0082] Among them, ρ is the pheromone volatility coefficient, Δτ ij (t) is the pheromone increment on the path;

[0083] S56. At the maximum number of iterations G max According to the comprehensive cost of the path and the concentration of pheromone in the path, the optimal rehabilitation path P with the lowest comprehensive cost and stability is selected. best :

[0084]

[0085] Optionally, the S6 specifically includes:

[0086] S61. Based on the optimal path P best The time attribute set generates the time distribution matrix T of the rehabilitation node dist :

[0087]

[0088] Each row of the matrix represents a rehabilitation node n i and the corresponding time t i The matching relationship;

[0089] S63. Extracting key stages of rehabilitation by combining standardized postoperative patient feature dataset and multi-dimensional feature weight matrix The function recovery target set G func :

[0090] G func ={g1,g2,…,g c};

[0091] Among them, g i Key stage of rehabilitation critical,i The functional recovery goal is to use the feature weight matrix W multi The weight values ​​in determine the functional recovery features that are prioritized at different stages;

[0092] S64. According to the optimal path P best The comprehensive cost C ij (t) and the characteristic change of real-time feedback Generate a dynamic adjustment suggestion matrix A for rehabilitation nodes dyn :

[0093]

[0094] Among them, a i Represents node n i Dynamic adjustment recommendations, including delaying, advancing or strengthening rehabilitation activities, are calculated as:

[0095]

[0096] Among them, δ min and δ max is a specific threshold used to determine whether the amount of feature change requires adjustment of the timing or content of the rehabilitation node;

[0097] S65. The time distribution matrix T of the rehabilitation node dist , Functional recovery target set G in key rehabilitation stages func and dynamically adjust the suggestion matrix A dyn A postoperative rehabilitation plan is generated comprehensively as the predicted results.

[0098] The beneficial effects of the present invention are:

[0099] (1) The present invention dynamically adjusts the number of ants and optimizes the ant colony distribution strategy in real time according to the dynamic changes in the comprehensive cost of the path between nodes. At key nodes in the rehabilitation process, the number of ants will be adaptively increased according to the complexity of the path and the real-time feature changes, so as to concentrate resources on exploring nodes of high importance. In the stage of relatively stable recovery, the number of ants will be appropriately reduced to improve computing efficiency. Compared with the traditional fixed-scale ant colony algorithm, the present invention can utilize computing resources more efficiently and significantly improve its adaptability to complex rehabilitation scenarios.

[0100] (2) The present invention adopts a spatiotemporal joint weight calculation mechanism based on multi-scale discrete degree and dynamic weighted information gain to construct a multi-dimensional feature weight matrix. The weight matrix comprehensively considers the time series changes and global importance of the patient's postoperative pain level characteristics, limb function recovery progress characteristics and motor ability assessment data characteristics, and accurately captures the key feature changes in the rehabilitation process. Compared with the traditional static weight allocation method, the spatiotemporal joint optimization mechanism of the present invention makes the rehabilitation node prediction more prominent in terms of temporal dynamics and personalized feature extraction.

[0101] (3) The present invention combines real-time feedback data in the process of path optimization, and realizes dynamic adjustment of the predicted results of rehabilitation nodes by adjusting the comprehensive cost of the path and the pheromone volatility coefficient. For the fluctuation of indicators (changes in pain level or delayed functional recovery) that occur during the patient's postoperative recovery process, the time arrangement and goal setting of the rehabilitation nodes can be optimized through a real-time updated adjustment suggestion matrix. The dynamic feedback mechanism can significantly reduce the risk of deviation from the rehabilitation rhythm and ensure that the predicted results are always consistent with the patient's actual recovery status. BRIEF DESCRIPTION OF THE DRAWINGS

[0102] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0103] Figure 1 This is a flow chart of a method for predicting orthopedic postoperative rehabilitation nodes based on an improved ant colony algorithm proposed by the present invention. DETAILED DESCRIPTION

[0104] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0105] refer to Figure 1 , a method for predicting orthopedic postoperative rehabilitation nodes based on improved ant colony algorithm, comprising the following steps:

[0106] S1. Obtain multi-dimensional data of postoperative patients, extract postoperative pain level characteristics, limb function recovery progress characteristics and motor ability assessment data characteristics of postoperative patients;

[0107] S2. Perform data cleaning, outlier processing, data standardization, and missing value filling for postoperative pain level characteristics, limb function recovery progress characteristics, and motor ability assessment data characteristics to generate a standardized postoperative patient characteristic data set;

[0108] S3. Use the entropy weight method to assign weights to the importance of postoperative pain level features, limb function recovery progress features, and motor ability assessment data features in the standardized postoperative patient feature data set, and generate a weight matrix containing the weights of each feature;

[0109] S4. Initialize the parameters of the improved ant colony algorithm based on the standardized feature dataset and weight matrix of postoperative patients;

[0110] S5. Use improved ant colony algorithm to search the path of postoperative rehabilitation nodes;

[0111] S6. The result of the improved ant colony algorithm path search is used as the prediction result of the postoperative rehabilitation node;

[0112] S7. Generate a postoperative rehabilitation plan for the patient based on the prediction results, including the functional recovery goals of each rehabilitation node, node schedule, and related rehabilitation training recommendations.

[0113] In this implementation, S1 specifically includes:

[0114] S11. Collect multi-dimensional data of postoperative patients through the postoperative monitoring system, and the multi-dimensional data includes:

[0115] Postoperative physiological index monitoring data, including heart rate, blood oxygen saturation, muscle electrical activity level and inflammatory marker levels;

[0116] Postoperative imaging data, which are medical images of specific parts of the patient after surgery;

[0117] Postoperative activity assessment data, including joint range of motion, gait stability score and sports balance ability;

[0118] S12. The postoperative pain level feature P is extracted from the postoperative physiological index monitoring data by using the feature analysis method, the key areas in the image are identified and segmented by using the image analysis algorithm for the postoperative imaging data, and the limb function recovery progress feature F is calculated, and the motion data analysis method is used to calculate the motion ability assessment data feature M for the postoperative activity ability assessment data;

[0119] S13. Based on the extracted postoperative pain level feature P, limb function recovery progress feature F and motor ability assessment data feature M of postoperative patients, a postoperative patient feature dataset D is constructed:

[0120] D={P,F,M}.

[0121] In this implementation, S2 specifically includes:

[0122] S21. Perform data cleaning on the postoperative patient feature dataset D to remove obvious outliers in the postoperative pain level feature P, limb function recovery progress feature F, and motor ability assessment data feature M. The basis for determining outliers is:

[0123] Outliers = {x i ∣x i >Q3+1.5·IQR or x i <Q1-1.5·IQR};

[0124] Among them, x i is the characteristic data point, Q1 and Q3 are the first quartile and the third quartile of the data respectively, and IQR is the interquartile range;

[0125] S22. Perform replacement processing on the outliers detected in the cleaned postoperative patient characteristic data set, use the mean interpolation method to process the outliers of the postoperative pain level characteristics of the postoperative patients, and use the linear interpolation method to process the outliers of the limb function recovery progress characteristics and the motor ability assessment data characteristics;

[0126] S23. The Z-score standardization method was used to standardize the postoperative pain level characteristics, limb function recovery progress characteristics and motor ability assessment data characteristics of the postoperative patients in the postoperative patient characteristic data set;

[0127] S24. Perform missing value detection and completion on the standardized postoperative patient characteristic data set, detect the location of missing values ​​in the postoperative pain level characteristics, limb function recovery progress characteristics and motor ability assessment data characteristics of postoperative patients, and use the nearest neighbor interpolation method to complete the missing values;

[0128] S25. Integrate the postoperative patient characteristic dataset after data cleaning, outlier processing, data standardization and missing value filling into a standardized postoperative patient characteristic dataset:

[0129] D std = {P std ,F std ,M std}.

[0130] In this implementation, S3 specifically includes:

[0131] S31. Standardized postoperative patient characteristics dataset D std The multi-scale discreteness of each feature in is analyzed, and the multi-scale discreteness is calculated by combining the local and global distribution information to calculate the multi-scale discreteness of feature i for:

[0132]

[0133] in, is the local discreteness of the feature, is the global discreteness of the feature, α is the weight coefficient of the local and global discreteness;

[0134] S32. Calculate dynamic weighted information gain based on the change of feature importance during the dynamic rehabilitation process of postoperative patients

[0135]

[0136] Among them, G i is the static information gain value of the feature, ω i (t) is the time-related feature weight change factor, which is calculated based on the gradient change of the historical data of the rehabilitation node:

[0137]

[0138] in, is the recovery loss function at a specific time point t, x i is the eigenvalue, ω i (t) represents the influence of the feature on the rehabilitation outcome at time t;

[0139] S33. Combining multi-scale discreteness and dynamic weighted information gain Calculating spatiotemporal joint weights

[0140]

[0141] Where m is the total number of features, Indicates the multi-scale discreteness of feature i, Represents the dynamic weighted information gain of feature i;

[0142] S34. Based on spatiotemporal joint weight and Construct a multi-dimensional feature weight matrix W multi :

[0143]

[0144] In this implementation, S4 specifically includes:

[0145] S41. Based on the standardized postoperative patient feature data set and the multi-dimensional feature weight matrix, the search space of the rehabilitation node prediction task is defined. The search space consists of the rehabilitation key stage node set N and the time attribute set T:

[0146] N={n1,n2,…,n k},1≤i≤k;

[0147] T={t1,t2,…,t k},1≤i≤k;

[0148] Among them, n i represents the i-th key stage node in the rehabilitation process, t i Represents the node n i The corresponding time attribute, k is the total number of nodes, and the construction and time attributes of the nodes are determined by the feature values ​​in the standardized postoperative patient feature data set and the weight distribution of the multi-dimensional feature weight matrix;

[0149] S42. Initialize the initial pheromone value of each path according to the node set N and the time attribute set T

[0150]

[0151] Among them, C ij For node n i To node n j The comprehensive cost of the path is:

[0152]

[0153] in, and For node n i and node n i Normalized value on feature k of the standardized postoperative patient feature dataset;

[0154] S43. Initialize the global parameters of the improved ant colony algorithm, including:

[0155] The pheromone volatility coefficient ρ, which ranges from 0<ρ<1, is used to control the rate of pheromone forgetting;

[0156] Path heuristic factor

[0157] Maximum number of iterations G max , used to limit the number of searches of the algorithm;

[0158] The number of ants A is used to control the number of ants participating in each round of path search;

[0159] S44. Combined with the characteristics of the rehabilitation node prediction task, a dynamic rule improvement mechanism is introduced to update the path cost based on real-time feedback

[0160]

[0161] Among them, λ is the adjustment coefficient, ΔC ij Indicates the change in path cost updated based on real-time data;

[0162] Dynamically update the path heuristic factor η based on the time attribute set T between nodes ij .

[0163] In this implementation, S5 specifically includes:

[0164] S51. Based on the rehabilitation key stage node set N, time attribute set T, and multi-dimensional feature weight matrix W multi And the comprehensive path cost C ij Constructing an initial recovery pathway:

[0165] P init ={n1→n2→…→n k};

[0166] S52. According to the complexity of the rehabilitation node prediction task, combined with the path comprehensive cost C ij And the dynamics of feature changes between nodes, initialize the number of ants A(t):

[0167]

[0168] Among them, A0 is the basic number of ants, κ is the adjustment coefficient, ΔC ij (t) represents node n i and n j The change in the comprehensive cost of the path at time t is:

[0169]

[0170] in, and Represents node n i and node n j Normalized value over time on feature k, m is the total number of features;

[0171] S53. According to the initialized dynamic ant number A(t) and path selection probability p ij Determine the path selection probability of each ant as:

[0172]

[0173] Among them, τ ij(t) is the path pheromone concentration at the current iteration time, η ij (t) is the heuristic factor of the path, α1 and β are the weight parameters of pheromone and heuristic factor respectively, N i For node n i The set of reachable adjacent nodes of ;

[0174] The dynamic path selection rule preferentially explores rehabilitation nodes with lower path comprehensive costs;

[0175] S54. In each round of iteration, the number of ants A(t+1) is adjusted according to the changes in the comprehensive hanging cost of the paths between nodes and the real-time feedback:

[0176]

[0177] At the same time, the ant colony distribution strategy is dynamically adjusted to make the ants concentrate in the path areas with large changes in comprehensive costs;

[0178] S55. Based on the path sequence generated by the dynamic path selection rule, the pheromone concentration τ on the path is updated after each round of iteration ij (t+1):

[0179] τ ij (t+1)=(1-ρ)·τ ij (t)+Δτ ij (t);

[0180] Among them, ρ is the pheromone volatility coefficient, Δτ ij (t) is the pheromone increment on the path;

[0181] S56. At the maximum number of iterations G max According to the comprehensive cost of the path and the concentration of pheromone in the path, the optimal rehabilitation path P with the lowest comprehensive cost and stability is selected. best :

[0182]

[0183] In this implementation, S6 specifically includes:

[0184] S61. Based on the optimal path P best The time attribute set generates the time distribution matrix T of the rehabilitation node dist :

[0185]

[0186] Each row of the matrix represents a rehabilitation node n i and the corresponding time t i The matching relationship;

[0187] S63. Extracting key stages of rehabilitation by combining standardized postoperative patient feature dataset and multi-dimensional feature weight matrix The function recovery target set G func :

[0188] G func ={g1,g2,…,g c};

[0189] Among them, g i Key stage of rehabilitation critical,i The functional recovery goal is to use the feature weight matrix W multi The weight values ​​in determine the functional recovery features that are prioritized at different stages;

[0190] S64. According to the optimal path P best The comprehensive cost C ij (t) and the characteristic change of real-time feedback Generate a dynamic adjustment suggestion matrix A for rehabilitation nodes dyn :

[0191]

[0192] Among them, a i Represents node n i Dynamic adjustment recommendations, including delaying, advancing or strengthening rehabilitation activities, are calculated as:

[0193]

[0194] Among them, δ min and δ max is a specific threshold used to determine whether the amount of feature change requires adjustment of the timing or content of the rehabilitation node;

[0195] S65. The time distribution matrix T of the rehabilitation node dist , Functional recovery target set G in key rehabilitation stages func and dynamically adjust the suggestion matrix A dyn A postoperative rehabilitation plan is generated comprehensively as the predicted results.

[0196] Embodiment 1:

[0197] Example In an orthopedic rehabilitation center, a 45-year-old female patient, Ms. Zhang, entered the rehabilitation stage after left knee replacement surgery. The doctor needed to develop a personalized rehabilitation node plan involving key functional goals such as pain relief, joint range of motion recovery, and gait stability. The rehabilitation center adopted an orthopedic postoperative rehabilitation node prediction method based on an improved ant colony algorithm to help dynamically plan rehabilitation nodes and adjust the rehabilitation plan in real time. The following is a detailed process of the actual application scenario:

[0198] On the first day after surgery, the system collected initial data from the patient's postoperative data monitoring device. Ms. Zhang's inflammatory marker CRP value was 35 mg / L, the average heart rate was 92 beats / minute, the knee joint range of motion was only 30°, and the gait stability score was 1 / 10. Image data analysis showed that there was moderate postoperative exudate around the knee joint. The system first preprocessed the data and calculated the feature weight matrix through the entropy weight method to determine that the current inflammatory indicator was the most important (weight of 40%), followed by joint range of motion (weight of 35%) and gait stability (weight of 25%).

[0199] Based on the above weight distribution, the improved ant colony algorithm began to make preliminary predictions of rehabilitation nodes. In the first round of path search, the system predicted that Ms. Zhang's pain relief node would be the 7th day after surgery, the node when the joint range of motion would recover to 90° would be the 21st day after surgery, and the node when her gait would be stable would be the 40th day after surgery. However, on the 3rd day after surgery, the system collected Ms. Zhang's CRP value in real time and it still remained at 34mg / L (normal should be less than 10mg / L), while the joint range of motion only increased to 35°, which was significantly lower than the system model's expectation (50°).

[0200] The system activated the dynamic adjustment mechanism, reallocated the pheromone concentration, and adjusted the comprehensive cost of the path. In the new prediction, the pain relief node was adjusted to the 10th day after surgery, and the joint range of motion recovery node was adjusted to the 25th day. At the same time, the system generated dynamic adjustment suggestions: delay the light weight training originally scheduled for the 5th day after surgery to the 7th day, and increase the number of cold compresses on the knee joint to accelerate the relief of inflammation.

[0201] On the 10th day after surgery, the system evaluated the patient's recovery status again. The data showed that Ms. Zhang's CRP value had dropped to 15 mg / L and the range of motion of her joints had increased to 65°, but her gait stability was still 1 / 10. Based on real-time data feedback, the system determined that the patient's inflammation relief was close to expectations, but her gait recovery was slow. The dynamic adjustment mechanism further optimized the path and recommended strengthening balance training, extending the interval between rehabilitation nodes, and replanning the gait stability node to the 45th day after surgery.

[0202] On the 21st day after surgery, Ms. Zhang's CRP value returned to the normal range (6mg / L), the range of motion of the joints reached 90°, and the gait stability improved to 3 / 10. The recovery process basically met the adjusted planning nodes. At this point, the system generated further rehabilitation goals for the subsequent stages, including improving gait stability to 7 / 10 on the 35th day after surgery and achieving normal walking ability on the 45th day after surgery.

[0203] To verify the effectiveness of the method, the hospital compared the data of 50 patients who used the traditional method and the improved ant colony algorithm for rehabilitation planning during the same period. The results are shown in Table 1:

[0204] Table 1 Comparison results of rehabilitation node prediction between improved ant colony algorithm and traditional method

[0205]

[0206] It can be seen from Table 1 above that for patients using the improved ant colony algorithm, the average prediction error of the rehabilitation nodes was ±2.3 days, while that of the traditional method was ±7.8 days. In terms of gait stability recovery, the average score of patients in the improved ant colony algorithm group on the 45th day was 8.2 / 10, while that of the traditional method group was 6.5 / 10. In terms of dynamic adjustment capability, the average adjustment response time of the improved ant colony algorithm group was 15 minutes, while the traditional method could not make real-time adjustments.

[0207] The examples fully demonstrate the significant advantages of the present invention in dynamically adjusting rehabilitation nodes and optimizing personalized rehabilitation plans. It not only effectively solves the problems of low prediction accuracy and inability to make real-time adjustments in traditional methods, but also significantly improves patients' rehabilitation efficiency and satisfaction.

[0208] The present invention dynamically adjusts the number of ants and optimizes the ant colony distribution strategy in real time according to the dynamic changes of the comprehensive cost of the path between nodes. At key nodes in the rehabilitation process, the number of ants will be adaptively increased according to the complexity of the path and real-time feature changes, so as to concentrate resources to explore nodes of high importance. In the stage of relatively stable recovery, the number of ants will be appropriately reduced to improve computing efficiency. Compared with the traditional fixed-scale ant colony algorithm, the present invention can utilize computing resources more efficiently and significantly improve the adaptability to complex rehabilitation scenarios.

[0209] The present invention adopts a spatiotemporal joint weight calculation mechanism based on multi-scale discrete degree and dynamic weighted information gain to construct a multi-dimensional feature weight matrix. The weight matrix comprehensively considers the time series changes and global importance of the patient's postoperative pain level characteristics, limb function recovery progress characteristics and motor ability assessment data characteristics, and accurately captures the key feature changes in the rehabilitation process. Compared with the traditional static weight allocation method, the spatiotemporal joint optimization mechanism of the present invention makes the rehabilitation node prediction more prominent in terms of temporal dynamics and personalized feature extraction.

[0210] The present invention combines real-time feedback data in the process of path optimization, and realizes dynamic adjustment of the prediction results of rehabilitation nodes by adjusting the comprehensive cost of the path and the pheromone volatility coefficient. For the fluctuations of indicators (changes in pain level or delayed functional recovery) that occur during the patient's postoperative recovery process, the time arrangement and goal setting of the rehabilitation nodes can be optimized through the real-time updated adjustment suggestion matrix. The dynamic feedback mechanism can significantly reduce the risk of deviation from the rehabilitation rhythm, ensuring that the prediction results are always consistent with the patient's actual recovery status.

[0211] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for predicting orthopedic postoperative rehabilitation nodes based on an improved ant colony algorithm, characterized in that: The steps include: S1. Obtain multi-dimensional data of postoperative patients, extract postoperative pain level characteristics, limb function recovery progress characteristics and motor ability assessment data characteristics of postoperative patients; S2. Perform data cleaning, outlier processing, data standardization, and missing value filling for postoperative pain level characteristics, limb function recovery progress characteristics, and motor ability assessment data characteristics to generate a standardized postoperative patient characteristic data set; S3. Use the entropy weight method to assign weights to the importance of postoperative pain level features, limb function recovery progress features, and motor ability assessment data features in the standardized postoperative patient feature data set, and generate a weight matrix containing the weights of each feature; S4. Initialize the parameters of the improved ant colony algorithm based on the standardized feature dataset and weight matrix of postoperative patients; S5. Use improved ant colony algorithm to search the path of postoperative rehabilitation nodes; S6. The result of the improved ant colony algorithm path search is used as the prediction result of the postoperative rehabilitation node; S7. Generate a postoperative rehabilitation plan for the patient based on the prediction results, including the functional recovery goals of each rehabilitation node, node schedule, and related rehabilitation training recommendations.

2. The orthopedic postoperative rehabilitation node prediction method based on improved ant colony algorithm according to claim 1, characterized in that: The S1 specifically includes: S11. Collect multi-dimensional data of the postoperative patient through the postoperative patient monitoring system, the multi-dimensional data including: Postoperative physiological index monitoring data, including heart rate, blood oxygen saturation, muscle electrical activity level, and inflammatory marker level; Postoperative imaging data, wherein the postoperative imaging data is medical imaging of a specific part of a patient after surgery; Postoperative activity assessment data, including range of motion of joints, gait stability score and movement balance ability; S12. The postoperative pain level feature P is extracted from the postoperative physiological index monitoring data by using the feature analysis method, the key areas in the image are identified and segmented by using the image analysis algorithm for the postoperative imaging data, and the limb function recovery progress feature F is calculated, and the motion data analysis method is used to calculate the motion ability assessment data feature M for the postoperative activity ability assessment data; S13. Based on the extracted postoperative pain level feature P, limb function recovery progress feature F and motor ability assessment data feature M of postoperative patients, a postoperative patient feature dataset D is constructed: D={P,F,M}.

3. The orthopedic postoperative rehabilitation node prediction method based on improved ant colony algorithm according to claim 1, characterized in that: The S2 specifically includes: S21. Perform data cleaning on the postoperative patient feature data set D to remove obvious outliers in the postoperative pain level feature P, limb function recovery progress feature F, and motor ability assessment data feature M. The determination basis of the outliers is: Outliers = {x i ∣x i >Q3+1.5·IQR or x i <Q1-1.5·IQR}; Among them, x i is the characteristic data point, Q1 and Q3 are the first quartile and the third quartile of the data respectively, and IQR is the interquartile range; S22. Perform replacement processing on the outliers detected in the cleaned postoperative patient characteristic data set, use the mean interpolation method to process the outliers of the postoperative pain level characteristics of the postoperative patients, and use the linear interpolation method to process the outliers of the limb function recovery progress characteristics and the motor ability assessment data characteristics; S23. The Z-score standardization method was used to standardize the postoperative pain level characteristics, limb function recovery progress characteristics and motor ability assessment data characteristics of the postoperative patients in the postoperative patient characteristic data set; S24. Perform missing value detection and completion on the standardized postoperative patient characteristic data set, detect the location of missing values ​​in the postoperative pain level characteristics, limb function recovery progress characteristics and motor ability assessment data characteristics of postoperative patients, and use the nearest neighbor interpolation method to complete the missing values; S25. Integrate the postoperative patient characteristic dataset after data cleaning, outlier processing, data standardization and missing value filling into a standardized postoperative patient characteristic dataset: D std ={P std ,F std ,M std }。 4. The orthopedic postoperative rehabilitation node prediction method based on improved ant colony algorithm according to claim 1, characterized in that: The S3 specifically includes: S31. Standardized postoperative patient characteristics dataset D std The multi-scale discreteness of each feature in is analyzed, and the multi-scale discreteness is calculated by combining the local and global distribution information to calculate the multi-scale discreteness of feature i for: in, is the local discreteness of the feature, is the global discreteness of the feature, α is the weight coefficient of the local and global discreteness; S32. Calculate dynamic weighted information gain based on the change of feature importance during the dynamic rehabilitation process of postoperative patients Among them, G i is the static information gain value of the feature, ω i (t) is the time-related feature weight change factor, which is calculated based on the gradient change of the historical data of the rehabilitation node: in, is the recovery loss function at a specific time point t, x i is the eigenvalue, ω i (t) represents the influence of the feature on the rehabilitation outcome at time t; S33. Combining multi-scale discreteness and dynamic weighted information gain Calculating spatiotemporal joint weights Where m is the total number of features, Indicates the multi-scale discreteness of feature i, Represents the dynamic weighted information gain of feature i; S34. Based on spatiotemporal joint weight and Construct a multi-dimensional feature weight matrix W multi :

5. The orthopedic postoperative rehabilitation node prediction method based on improved ant colony algorithm according to claim 1, characterized in that: The S4 specifically includes: S41. Based on the standardized postoperative patient feature data set and the multi-dimensional feature weight matrix, the search space of the rehabilitation node prediction task is defined. The search space consists of the rehabilitation key stage node set N and the time attribute set T: N={n1,n2,…,n k },1≤i≤k; T={t1,t2,…,t k },1≤i≤k; Among them, n i represents the i-th key stage node in the rehabilitation process, t i Represents the node n i The corresponding time attribute, k is the total number of nodes, and the construction and time attributes of the nodes are determined by the feature values ​​in the standardized postoperative patient feature data set and the weight distribution of the multi-dimensional feature weight matrix; S42. Initialize the initial pheromone value of each path according to the node set N and the time attribute set T Among them, C ij For node n i To node n j The comprehensive cost of the path is: in, and For node n i and node n i Normalized value on feature k of the standardized postoperative patient feature dataset; S43. Initialize the global parameters of the improved ant colony algorithm, including: The pheromone volatility coefficient ρ, which ranges from 0<ρ<1, is used to control the rate of pheromone forgetting; Path heuristic factor Maximum number of iterations G max , used to limit the number of searches of the algorithm; The number of ants A is used to control the number of ants participating in each round of path search; S44. Combined with the characteristics of the rehabilitation node prediction task, a dynamic rule improvement mechanism is introduced to update the path cost based on real-time feedback Among them, λ is the adjustment coefficient, ΔC ij Indicates the change in path cost updated based on real-time data; Dynamically update the path heuristic factor η based on the time attribute set T between nodes ij .

6. The orthopedic postoperative rehabilitation node prediction method based on improved ant colony algorithm according to claim 1, characterized in that: The S5 specifically includes: S51. Based on the rehabilitation key stage node set N, time attribute set T, and multi-dimensional feature weight matrix W multi And the comprehensive path cost C ij Constructing an initial recovery pathway: P init ={n1→n2→…→n k }; S52. According to the complexity of the rehabilitation node prediction task, combined with the path comprehensive cost C ij And the dynamics of feature changes between nodes, initialize the number of ants A(t): Among them, A0 is the basic number of ants, κ is the adjustment coefficient, ΔC ij (t) represents node n i and n j The change in the comprehensive cost of the path at time t is: in, and Represents node n i and node n j Normalized value over time on feature k, m is the total number of features; S53. According to the initialized dynamic ant number A(t) and path selection probability p ij Determine the path selection probability of each ant as: Among them, τ ij (t) is the path pheromone concentration at the current iteration time, η ij (t) is the heuristic factor of the path, α1 and β are the weight parameters of pheromone and heuristic factor respectively, N i For node n i The set of reachable adjacent nodes of ; The dynamic path selection rule preferentially explores rehabilitation nodes with lower path comprehensive costs; S54. In each round of iteration, the number of ants A(t+1) is adjusted according to the changes in the comprehensive hanging cost of the paths between nodes and the real-time feedback: At the same time, the ant colony distribution strategy is dynamically adjusted to make the ants concentrate in the path areas with large changes in comprehensive costs; S55. Based on the path sequence generated by the dynamic path selection rule, the pheromone concentration τ on the path is updated after each round of iteration ij (t+1): t ij (t+1)=(1-ρ)·τ ij (t)+Δτ ij (t); Among them, ρ is the pheromone volatility coefficient, Δτ ij (t) is the pheromone increment on the path; S56. At the maximum number of iterations G max According to the comprehensive cost of the path and the concentration of pheromone in the path, the optimal rehabilitation path P with the lowest comprehensive cost and stability is selected. best :

7. The orthopedic postoperative rehabilitation node prediction method based on improved ant colony algorithm according to claim 1, characterized in that: The S6 specifically includes: S61. Based on the optimal path P best The time attribute set generates the time distribution matrix T of the rehabilitation node dist : Each row of the matrix represents a rehabilitation node n. i and the corresponding time t i The matching relationship; S63. Extracting key stages of rehabilitation by combining standardized postoperative patient feature dataset and multi-dimensional feature weight matrix The function recovery target set G func : G func ={g1,g2,…,g c }; Among them, g i Key stage of rehabilitation critical,i The functional recovery goal is to use the feature weight matrix W multi The weight values ​​in determine the functional recovery features that are prioritized at different stages; S64. According to the optimal path P best The comprehensive cost C ij (t) and the characteristic change of real-time feedback Generate a dynamic adjustment suggestion matrix A for rehabilitation nodes dyn : Among them, a i Represents node n i Dynamic adjustment recommendations, including delaying, advancing or strengthening rehabilitation activities, are calculated as: Among them, δ min and δ max is a specific threshold used to determine whether the amount of feature change requires adjustment of the timing or content of the rehabilitation node; S65. The time distribution matrix T of the rehabilitation node dist , Functional recovery target set G in key rehabilitation stages func and dynamically adjust the suggestion matrix A dyn A postoperative rehabilitation plan is generated comprehensively as the predicted results.

Images