Individualized treatment plan generation platform of cmvd combining cpet and four diagnoses

Abstract

The present application relates to the technical field of CMVD treatment scheme, in particular to a CMVD individualized treatment scheme generation platform fusing CPET and four diagnoses, comprising: collecting patient exercise cardiopulmonary function test data, traditional Chinese medicine four diagnosis information data and routine physiological and biochemical index data through a data acquisition module; an improved hierarchical clustering algorithm based on dynamic adjustment of feature distance according to traditional Chinese medicine syndrome factor weight is adopted in a feature fusion module to complete data fusion and feature extraction to obtain a comprehensive syndrome type feature vector; a strategy matching module maps and matches it with a CMVD traditional Chinese and western medicine diagnosis and treatment knowledge graph to generate a preliminary intervention strategy; a tolerance optimization module performs tolerance evaluation and intensity adaptation based on extreme exercise parameters; and a scheme generation module integrates related information to generate an individualized treatment scheme. The present application can conform to traditional Chinese medicine diagnosis and treatment logic, adapt to individual patient tolerance and physical differences, and improve diagnosis and treatment pertinence and adaptability.

Description

Technical Field

[0001] This invention relates to the field of CMVD treatment technology, and in particular to a platform for generating individualized CMVD treatment plans that integrates CPET and the four diagnostic methods. Background Technology

[0002] Treatment of CMVD (coronary microvascular disease) requires tailoring plans to individual patient differences. Currently, existing CMVD treatment plan generation technologies largely rely on single-dimensional diagnostic data, or are based solely on Western medical physiological and biochemical indicators, or simply on information from the four diagnostic methods of Traditional Chinese Medicine. Some technologies attempt to combine data from both Western and Traditional Chinese Medicine, but have not achieved effective integration of multi-dimensional data. Cardiopulmonary exercise testing (CPET) can reflect a patient's exercise tolerance and is of significant reference value for adjusting the intensity of treatment plans, yet it has not been effectively integrated into the process of generating individualized CMVD treatment plans.

[0003] Conventional data fusion methods fail to incorporate the characteristics of Traditional Chinese Medicine (TCM) diagnosis and treatment, and cannot adjust feature extraction logic according to the weights of TCM syndrome elements. This results in the fused syndrome characteristics not matching the patient's actual syndrome, affecting the targeted nature of treatment strategies. Furthermore, the current treatment plan generation process does not assess the patient's tolerance to the intervention strategy based on their extreme exercise parameters, nor does it integrate constitution identification results and the risk levels of physiological and biochemical indicators. It only generates generic TCM and Western medicine intervention strategies, leading to a mismatch between the plan and the patient's individual tolerance and physical condition, making it difficult to adapt to the differences in patients' conditions and physical circumstances. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a platform for generating individualized treatment plans for CMVD that integrates CPET and the four diagnostic methods.

[0005] To achieve the above objectives, the present invention employs the following technical solution: a CMVD individualized treatment plan generation platform integrating CPET and four diagnostic methods, comprising:

[0006] The data acquisition module collects a multidimensional health data set of CMVD patients, which includes exercise cardiopulmonary function test data, traditional Chinese medicine four diagnostic information data, and routine physiological and biochemical indicator data.

[0007] The feature fusion module uses an improved hierarchical clustering algorithm to perform data fusion and feature extraction on the multidimensional health data set to obtain the comprehensive syndrome feature vector of the patient. The improved hierarchical clustering algorithm dynamically adjusts the feature distance based on the weight of the TCM syndrome elements.

[0008] The strategy matching module maps and matches the comprehensive syndrome feature vector with the preset CMVD Chinese and Western medicine diagnosis and treatment knowledge graph to generate a preliminary set of Chinese and Western medicine integrated intervention strategies.

[0009] The tolerance optimization module, based on the extreme exercise parameters in the exercise cardiopulmonary function test data, performs tolerance assessment and intensity adaptation on the preliminary set of integrated traditional Chinese and Western medicine intervention strategies, and generates an optimized individualized intervention strategy.

[0010] The treatment plan generation module integrates the individualized intervention strategy, the constitution identification results from the four diagnostic methods of traditional Chinese medicine, and the risk level from the routine physiological and biochemical indicators to generate an individualized treatment plan that includes a traditional Chinese medicine prescription, an exercise prescription, lifestyle recommendations, and monitoring indicators.

[0011] As a further aspect of the present invention, the improved hierarchical clustering algorithm dynamically adjusts the feature distance based on the weights of TCM syndrome elements, including:

[0012] Construct an initial feature space, using each data item in the multidimensional health data set as a dimension of the initial feature space;

[0013] A basic weight is assigned to each dimension in the initial feature space, and the basic weight is determined based on the importance of the data represented by the dimension in the CMVD diagnosis and treatment guidelines;

[0014] Based on the TCM four diagnostic methods data, the distribution of TCM syndrome elements is extracted. The TCM syndrome elements include Qi deficiency, blood stasis, phlegm turbidity, Yin deficiency, and Yang deficiency.

[0015] Based on the distribution of the extracted TCM syndrome elements, the weights of the corresponding dimensions in the initial feature space are dynamically adjusted to form a dynamic weight feature space. The weights of dimensions with high correlation to the main syndrome elements are increased, while the weights of dimensions with low correlation are decreased.

[0016] In the dynamic weight feature space, the weighted Euclidean distance between any two patient data samples is calculated, and the weighted Euclidean distance is calculated using the weights in the dynamic weight feature space;

[0017] Based on the calculated weighted Euclidean distance, a bottom-up aggregation strategy is used for hierarchical clustering. In each aggregation, the two classes with the smallest inter-class distance are selected for merging, and the distance between the new class and other classes in the dynamic weight feature space is recalculated.

[0018] The inter-class distance is calculated using the Ward method with the goal of minimizing the increase in intra-class variance.

[0019] Iteratively execute the aggregation and distance recalculation process until all samples are aggregated into one class or the preset number of clusters is reached, and output the final clustering result;

[0020] From each category of the final clustering result, the feature values ​​of its center point are extracted to form a comprehensive syndrome feature vector representing the characteristics of the patient group in the corresponding category.

[0021] As a further aspect of the present invention, the comprehensive syndrome feature vector is mapped and matched with a preset CMVD (Chinese and Western medicine diagnosis and treatment knowledge graph) to generate a preliminary set of integrated Chinese and Western medicine intervention strategies, including:

[0022] The CMVD knowledge graph of Chinese and Western medicine diagnosis and treatment includes nodes for Chinese medicine syndrome types, nodes for Western medicine pathophysiology, nodes for Chinese medicine prescriptions, nodes for treatment methods, and the relationships between them.

[0023] The similarity between the comprehensive syndrome feature vector and the TCM syndrome node is calculated to identify the target TCM syndrome node with the highest matching degree with the patient;

[0024] In the CMVD knowledge graph of Chinese and Western medicine diagnosis and treatment, starting from the target TCM syndrome node, the Western medicine pathophysiological nodes directly associated with it are traversed to obtain the associated Western medicine pathophysiological state description.

[0025] Continue traversing the nodes of traditional Chinese medicine prescriptions and treatment methods that are connected to the target TCM syndrome node and the associated Western medicine pathophysiology node, to obtain a basic set of candidate TCM prescriptions and candidate treatment methods.

[0026] Based on the description of the Western medicine pathophysiological state, corresponding standard Western medicine intervention measures are retrieved from the medical guideline database to form a candidate set of Western medicine intervention measures;

[0027] The candidate sets of basic Chinese herbal formulas, treatment methods, and Western medicine interventions are combined, and the compatibility and synergistic relationships between the items within each combination are verified. Conflicting combinations are eliminated, and compatible combinations are retained to form the preliminary set of integrated Chinese and Western medicine intervention strategies.

[0028] As a further aspect of the present invention, based on the extreme exercise parameters in the cardiopulmonary exercise test data, the tolerance and intensity of the preliminary integrated traditional Chinese and Western medicine intervention strategy set are assessed to generate an optimized individualized intervention strategy, including:

[0029] Peak oxygen uptake, anaerobic threshold, and maximum heart rate were extracted from the exercise cardiopulmonary function test data.

[0030] Set an aerobic exercise intensity range expressed as a percentage of the peak oxygen uptake, an anaerobic exercise intensity threshold based on the power or heart rate corresponding to the anaerobic threshold, and use a percentage of the maximum heart rate as the upper limit of heart rate control.

[0031] For each strategy in the preliminary set of integrated traditional Chinese and Western medicine intervention strategies, the preset intensity of the exercise prescription recommendations contained therein is analyzed;

[0032] The preset intensity is compared with the aerobic exercise intensity range, the anaerobic exercise intensity threshold, and the heart rate control upper limit to determine whether the preset intensity is within a safe and effective range.

[0033] For strategies whose preset intensity exceeds the safe and effective range, the exercise intensity shall be adjusted to the median of the aerobic exercise intensity range or below the anaerobic exercise intensity threshold.

[0034] For a strategy with a preset intensity that includes drug intervention, the patient's tolerance to potential adverse drug reactions is assessed based on the cardiac function reserve level reflected by the peak oxygen uptake, and the starting drug dose or physiological indicators that need to be closely monitored are recommended accordingly.

[0035] After evaluating and adjusting all strategies, the optimized individualized intervention strategy is generated, labeled with the appropriate exercise intensity, precautions for drug use, and intensity basis.

[0036] As a further aspect of the present invention, the individualized intervention strategy, the constitution identification results from the four diagnostic methods of traditional Chinese medicine, and the risk levels from the routine physiological and biochemical indicator data are integrated to generate an individualized treatment plan that includes a traditional Chinese medicine prescription, an exercise prescription, lifestyle recommendations, and monitoring indicators, including:

[0037] Extract appropriate integrated traditional Chinese and Western medicine intervention measures from the optimized individualized intervention strategy;

[0038] The constitution identification results are extracted from the TCM four diagnostic information data. The constitution identification results are one or more of the following: balanced constitution, qi deficiency constitution, yang deficiency constitution, yin deficiency constitution, phlegm-dampness constitution, damp-heat constitution, blood stasis constitution, qi stagnation constitution, and special constitution.

[0039] Based on the constitution identification results, the traditional Chinese medicine prescriptions in the adapted Chinese and Western medicine intervention measures are fine-tuned. The fine-tuning includes increasing or decreasing the dosage of specific herbs according to the constitution, or adding herbs that are targeted at conditioning the constitution.

[0040] Based on the results of the constitution identification, we propose tendencies for the types of exercise prescriptions in the adapted Chinese and Western medicine intervention measures;

[0041] Extract cardiovascular risk levels from the aforementioned routine physiological and biochemical indicator data;

[0042] Based on the cardiovascular risk level, the monitoring frequency, target range of monitoring indicators, and emergency response triggering conditions of the treatment plan are set.

[0043] The finely adjusted traditional Chinese medicine prescription, the accompanying exercise prescription with type suggestions, the personalized lifestyle recommendations based on physical condition and risk level, and the set monitoring frequency, target range and emergency response trigger conditions are integrated into a structured individualized treatment plan.

[0044] As a further aspect of the present invention, based on the constitution identification results, the traditional Chinese medicine prescriptions in the adapted integrated traditional Chinese and Western medicine intervention measures are fine-tuned, including:

[0045] Establish a constitution-medicine property correspondence table, which describes the appropriate medicinal property bias for different constitution types;

[0046] Query the appropriate medicinal properties corresponding to the constitution identification results in the constitution-medicine property correspondence table;

[0047] Analyze the overall medicinal properties of the traditional Chinese medicine prescriptions in the adapted integrated traditional Chinese and Western medicine intervention measures;

[0048] Compare the overall medicinal properties of the Chinese herbal formula with the appropriate medicinal properties corresponding to the constitution identification results;

[0049] If there is a discrepancy between the overall medicinal properties of a traditional Chinese medicine prescription and the appropriate medicinal properties corresponding to the constitution identification results, one or more adjuvant herbs with corrective medicinal properties may be added to the prescription, or the dosage ratio of the principal and assistant herbs in the prescription may be adjusted so that the overall medicinal properties of the adjusted prescription are closer to the appropriate medicinal properties.

[0050] The process of adjusting traditional Chinese medicine prescriptions strictly follows the principle of incompatibilities in Chinese medicine combinations, and ensures that the core therapeutic function of the prescription remains unchanged.

[0051] As a further aspect of the present invention, in the dynamic weight feature space, the weighted Euclidean distance between any two patient data samples is calculated, wherein the calculation of the weighted Euclidean distance uses weights in the dynamic weight feature space, including:

[0052] Obtain the weights of all dimensions in the dynamic weight feature space to form a weight vector;

[0053] For any two patient data samples, extract their feature values ​​in all dimensions of the dynamic weight feature space to form their respective feature vectors;

[0054] Calculate the difference between two feature vectors along each dimension;

[0055] Squaring the differences in each dimension;

[0056] Multiply the difference in each dimension after squaring by the weight of the corresponding dimension in the weight vector;

[0057] The sum is obtained by summing the weighted squared differences across all dimensions;

[0058] The sum is square-rooted, and the result is used as the weighted Euclidean distance between the two patient data samples.

[0059] As a further aspect of the present invention, the basic set of candidate traditional Chinese medicine prescriptions, the set of candidate treatment methods, and the set of candidate Western medicine intervention measures are combined, and the compatibility and synergistic relationships between the items within each combination are verified. Conflicting combinations are eliminated, and compatible combinations are retained to form the preliminary set of integrated traditional Chinese and Western medicine intervention strategies, including:

[0060] All items in the basic set of candidate Chinese herbal formulas, the set of candidate treatment methods, and the set of candidate Western medicine interventions are combined to generate a series of prototype intervention strategies that combine Chinese and Western medicine.

[0061] Each of the aforementioned strategy prototypes is input into a preset compatibility check model, which includes a rule base of eighteen incompatibilities and nineteen antagonisms of traditional Chinese medicine, a knowledge base of interactions between traditional Chinese and Western medicine, and a knowledge base of contraindications for specific treatment methods.

[0062] The matching taboo verification model is used to check whether there are explicit matching taboos between different items in the strategy prototype. If there is a taboo relationship between any two items, the strategy prototype is determined to be a conflicting combination and is removed from the candidate set.

[0063] For the strategy prototype that passes the incompatibility check, it is input into a preset synergy evaluation model. The synergy evaluation model calculates the comprehensive synergy score of the internal items of the strategy prototype based on the similarity of drug action mechanism, the superposition effect of efficacy and the complementarity of action target.

[0064] Set a collaboration score threshold, and identify strategy prototypes with a comprehensive collaboration score below the threshold as combinations with poor compatibility and remove them;

[0065] All compatible strategy prototypes retained after the incompatibilities check and synergy assessment are compiled into the preliminary set of integrated traditional Chinese and Western medicine intervention strategies.

[0066] As a further aspect of the present invention, after generating the individualized treatment plan, further treatment feedback and plan iteration are performed, including:

[0067] During the period when the patient is implementing the individualized treatment plan, periodic follow-up data will be collected regularly. The periodic follow-up data includes simplified cardiopulmonary function indicators, changes in the four diagnostic methods, results of physiological and biochemical indicator retests, and the patient's subjective symptom scores.

[0068] The periodic follow-up data is compared with the baseline data before the implementation of the individualized treatment plan to generate a multidimensional efficacy assessment vector;

[0069] The multidimensional efficacy assessment vector is input into a protocol optimization model, which outputs adjustment suggestions for each intervention in the current treatment protocol. The adjustment suggestions include adding, reducing, maintaining or replacing a certain measure, as well as the intensity or frequency of the adjustment measure.

[0070] Based on the aforementioned adjustment recommendations, the individualized treatment plan is revised to generate a new version of the individualized treatment plan;

[0071] The new individualized treatment plan, the basis for adjustment, and the follow-up data will be stored together in the patient's medical record and used as input for the next round of plan iteration.

[0072] As a further aspect of the present invention, the multidimensional efficacy assessment vector is input into a treatment plan optimization model, and the treatment plan optimization model outputs adjustment suggestions for various interventions in the current treatment plan, including:

[0073] The optimization model described is a model built on a reinforcement learning framework, which treats the treatment plan as a combination of actions taken by the agent and regards the improvement of the patient's health status as a reward.

[0074] The multidimensional efficacy assessment vector is encoded as a state vector, which describes the change in the patient’s health status relative to the baseline at the end of the current treatment cycle.

[0075] The currently executed individualized treatment plan is encoded as an action vector;

[0076] The immediate reward value corresponding to the current state vector is calculated according to a preset reward function, which comprehensively considers the degree of improvement of physiological indicators, the degree of symptom relief, and the safety indicators.

[0077] The policy network and value network in the scheme optimization model are updated using the state vector, the action vector, and the instant reward value.

[0078] The updated strategy network outputs the probability distribution of adjustment suggestions for each intervention in the next treatment cycle, and selects the adjustment suggestion with the highest probability as the output.

[0079] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0080] An improved hierarchical clustering algorithm was employed to fuse and extract features from a multidimensional health dataset of CMVD patients. This dataset included cardiopulmonary exercise test data, TCM diagnostic information (four diagnostic methods), and routine physiological and biochemical indicators. The improved algorithm dynamically adjusted the feature distance based on the weights of TCM syndrome elements, resulting in a comprehensive syndrome feature vector for each patient. This approach aligns the data fusion process with TCM diagnostic logic, ensuring feature extraction more closely reflects the patient's actual syndrome condition. It avoids the syndrome feature bias caused by conventional data fusion methods that fail to consider syndrome element weights, allowing the comprehensive syndrome feature vector to better reflect the patient's true condition and enabling more targeted mapping and matching of subsequent treatment strategies.

[0081] Based on the extreme exercise parameters from cardiopulmonary exercise testing data, the tolerance and intensity of the initial set of integrated traditional Chinese and Western medicine intervention strategies are assessed and adapted to generate an optimized individualized intervention strategy. Simultaneously, the system integrates constitution identification results from the four diagnostic methods of traditional Chinese medicine, as well as risk levels from routine physiological and biochemical indicators, to generate an individualized treatment plan that includes traditional Chinese medicine prescriptions, exercise prescriptions, lifestyle recommendations, and monitoring indicators. This approach ensures that the intensity of the intervention strategy is matched to the patient's exercise tolerance, avoiding mismatches between the intervention intensity and the patient's physical condition. Furthermore, by combining constitution identification results and risk levels, the treatment plan encompasses multi-dimensional intervention content, tailored to the individual patient's constitution and disease risk, distinguishing it from conventional universal intervention strategies and making the plan more adaptable to individual patient differences. Attached Figure Description

[0082] Figure 1 This is a timing diagram of the CMVD individualized treatment plan generation platform that integrates CPET and the four diagnostic methods described in this invention.

[0083] Figure 2 A flowchart for generating a preliminary set of intervention strategies by mapping and matching feature vectors of comprehensive syndrome types;

[0084] Figure 3 A flowchart for generating individualized intervention strategies based on extreme motion parameters. Detailed Implementation

[0085] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0086] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0087] See Figure 1 This invention provides a platform for generating personalized treatment plans for coronary microvascular diseases that integrates cardiopulmonary exercise testing and the four diagnostic methods of traditional Chinese medicine. The overall implementation scheme is as follows:

[0088] The data acquisition module is responsible for collecting multi-dimensional health information from patients with coronary microvascular disease, forming a multi-dimensional health dataset. This dataset includes exercise cardiopulmonary function test data, traditional Chinese medicine (TCM) diagnostic information, and routine physiological and biochemical indicators. The feature fusion module receives the aforementioned multi-dimensional health dataset and employs an improved hierarchical clustering algorithm to fuse data and extract features. This algorithm dynamically adjusts the feature distance based on the weights of TCM syndrome elements, ultimately outputting a comprehensive syndrome feature vector that characterizes the patient's overall condition. The strategy matching module has a pre-set CMVD (Coronary Ventilation Disease) TCM and Western medicine diagnostic knowledge graph. This module maps and matches the comprehensive syndrome feature vector with the knowledge graph to generate a preliminary set of integrated TCM and Western medicine intervention strategies. The tolerance optimization module uses the extreme exercise parameters from the exercise cardiopulmonary function test data to assess the tolerance and intensity of each strategy in the preliminary strategy set, filtering out unsafe or unsuitable strategies and adjusting the intervention intensity to generate an optimized individualized intervention strategy. The final step of the treatment plan generation module is to integrate the optimized individualized intervention strategy, the constitution identification results from the four diagnostic methods of traditional Chinese medicine, and the risk levels from routine physiological and biochemical indicators to generate a structured, complete individualized treatment plan that includes Chinese medicine prescriptions, exercise prescriptions, lifestyle recommendations, and monitoring indicators for clinicians to refer to and implement.

[0089] In one embodiment of the present invention, the improved hierarchical clustering algorithm dynamically adjusts the feature distance based on the weights of TCM syndrome elements. The algorithm first constructs an initial feature space, with each data item in the multidimensional health dataset serving as a dimension of this space. A basic weight is assigned to each dimension in the initial feature space, determined based on the importance of the data represented by that dimension in the guidelines for the diagnosis and treatment of coronary microvascular disease. Based on the four diagnostic methods of TCM, the distribution of TCM syndrome elements is extracted, including Qi deficiency, blood stasis, phlegm turbidity, Yin deficiency, and Yang deficiency. According to the extracted distribution of TCM syndrome elements, the weights of the corresponding dimensions in the initial feature space are dynamically adjusted to form a dynamic weighted feature space. In this space, the weights of dimensions with high correlation to the main syndrome elements are increased, while the weights of dimensions with low correlation are decreased. In the dynamic weighted feature space, the weighted Euclidean distance between any two patient data samples needs to be calculated. Specifically, the weights of all dimensions in the dynamic weighted feature space are obtained to form a weight vector. For any two patient data samples, their feature values ​​on all dimensions in the dynamic weighted feature space are extracted to form their respective feature vectors. Calculate the difference between two feature vectors across all dimensions. Square the difference across each dimension. Multiply the squared difference across each dimension by the corresponding weight in the weight vector. Sum the weighted squared differences across all dimensions to obtain a total. Calculate the square root of this sum; the result is the weighted Euclidean distance between the two patient data samples. Based on the calculated weighted Euclidean distance, a bottom-up aggregation strategy is used for hierarchical clustering. In each aggregation, the two classes with the smallest inter-class distance are merged. The inter-class distance is calculated using the Ward method, aiming to minimize the increase in intra-class variance. The distance between the new class and other classes in the dynamic weight feature space is recalculated. Iterate through the aggregation and distance recalculation process until all samples are aggregated into one class or the preset number of clusters is reached, and output the final clustering result. From each class in the final clustering result, extract the feature values ​​of its centroid to form a comprehensive syndrome feature vector representing the characteristics of the corresponding patient group.

[0090] In its implementation, the improved hierarchical clustering algorithm dynamically adjusts the feature distance based on the weights of TCM syndrome elements. The algorithm first constructs an initial feature space, and the exercise cardiopulmonary function test data, TCM four diagnostic information data, and routine physiological and biochemical index data in the multidimensional health data set are all mapped to dimensions in the initial feature space. In its implementation, a basic weight is assigned to each dimension in the initial feature space. The basic weight is determined according to the importance of the data represented by the dimension in the coronary microvascular disease diagnosis and treatment guidelines. For example, the peak oxygen uptake dimension may be assigned a higher basic weight, while the routine blood pressure dimension may be assigned a medium basic weight. In some embodiments, the distribution of TCM syndrome elements is extracted based on TCM four diagnostic methods data. These TCM syndrome elements include Qi deficiency, blood stasis, phlegm turbidity, Yin deficiency, and Yang deficiency. The weights of corresponding dimensions in the initial feature space are dynamically adjusted according to the extracted distribution of TCM syndrome elements, forming a dynamic weighted feature space. In specific implementations, the weights of dimensions with high correlation to the main syndrome elements are increased, while the weights of dimensions with low correlation are decreased. For example, if a patient's TCM syndrome element distribution is predominantly Qi deficiency, the weights of dimensions related to cardiopulmonary function are increased, while the weights of dimensions related to phlegm turbidity are decreased. Optionally, the weight adjustment in the dynamic weighted feature space is based on a preset syndrome element-dimension correlation mapping table, which defines the correlation strength between different syndrome elements and each dimension in the multidimensional health data set.

[0091] In practice, the weighted Euclidean distance between any two patient data samples is calculated in the dynamic weighted feature space. The weighted Euclidean distance is calculated using the weights in the dynamic weighted feature space. The weights of all dimensions in the dynamic weighted feature space are used to construct a weight vector. For any two patient data samples, their feature values ​​in all dimensions of the dynamic weighted feature space are extracted to form their respective feature vectors. The formula for calculating the weighted Euclidean distance is as follows:

[0092]

[0093] in: This represents the weighted Euclidean distance between patient data sample A and patient data sample B. This represents the total number of dimensions in the dynamic weight feature space. Represents the first in the dynamic weight feature space Weights of each dimension This indicates that patient data sample A is in the first... Feature values ​​in each dimension This indicates that patient data sample B is in the... The formula defines eigenvalues ​​across several dimensions. This ensures that, in distance calculations, differences in features across higher-weighted dimensions contribute more significantly to the overall distance, thus reflecting the importance of TCM syndrome elements. In some embodiments, the eigenvalues ​​are standardized before calculation to eliminate the influence of differences in the dimensions of different dimensions.

[0094] Based on the calculated weighted Euclidean distance, a bottom-up aggregation strategy is used for hierarchical clustering. In each aggregation, the two classes with the smallest inter-class distance are merged. The inter-class distance is calculated using the Ward method, which aims to minimize the increase in intra-class variance. The distance between the new class and other classes in the dynamic weighted feature space is recalculated. This aggregation and distance recalculation process is iteratively executed until all samples are aggregated into one class or a preset number of clusters is reached, at which point the final clustering result is output. Optionally, the preset number of clusters is determined based on clinical needs or algorithm evaluation metrics such as the silhouette coefficient. From each class in the final clustering result, the feature values ​​of its centroids are extracted. The centroids are obtained by calculating the mean of the feature values ​​of all samples within the class, forming a comprehensive syndrome feature vector representing the characteristics of the corresponding patient group. This comprehensive syndrome feature vector can be understood as integrating dynamically weighted information from a multi-dimensional health data set.

[0095] In one embodiment of the present invention, the CMVD (Combined Traditional Chinese Medicine and Western Medicine Diagnosis and Treatment Knowledge Graph) includes nodes for TCM syndrome types, nodes for Western medicine pathophysiology, nodes for TCM prescriptions, nodes for treatment methods, and the relationships between them. See also... Figure 2The similarity calculation is performed between the comprehensive syndrome feature vector and the TCM syndrome nodes in the knowledge graph to identify the target TCM syndrome node with the highest matching degree with the patient. Starting from the target TCM syndrome node in the knowledge graph, the system traverses the directly associated Western medicine pathophysiological nodes to obtain the associated Western medicine pathophysiological state descriptions. The system continues to traverse the TCM formula nodes and treatment method nodes connected to both the target TCM syndrome node and the associated Western medicine pathophysiological nodes, obtaining a basic set of TCM formula candidates and treatment method candidates. Based on the Western medicine pathophysiological state descriptions, corresponding standard Western medicine intervention measures are retrieved from the medical guideline database to form a Western medicine intervention measure candidate set. All items in the basic TCM formula candidate set, treatment method candidate set, and Western medicine intervention measure candidate set are fully combined to generate a series of TCM-Western medicine integrated intervention strategy prototypes. Each strategy prototype is input into a preset incompatibility verification model, which includes a TCM 18-antagonism and 19-antagonism rule base, a TCM-Western medicine interaction knowledge base, and a specific treatment method contraindication knowledge base. A drug incompatibility verification model is used to check whether there are explicit drug incompatibilities between different items in the strategy prototype. If any two items are incompatible, the strategy prototype is determined to be a conflicting combination and is removed from the candidate set. For strategy prototypes that pass the drug incompatibility verification, they are input into a pre-defined synergy evaluation model. The synergy evaluation model calculates a comprehensive synergy score for the items within the strategy prototype based on the similarity of drug mechanisms of action, the cumulative effect of therapeutic efficacy, and the complementarity of drug targets. A synergy score threshold is set; strategy prototypes with a comprehensive synergy score below this threshold are determined to be combinations with poor compatibility and are removed. All compatible strategy prototypes retained after drug incompatibility verification and synergy evaluation are compiled into a preliminary set of integrated traditional Chinese and Western medicine intervention strategies.

[0096] In specific implementation, the strategy matching module maps and matches the comprehensive syndrome feature vector with a pre-defined CMVD (Chinese and Western Medicine Diagnosis and Treatment Knowledge Graph). The CMVD knowledge graph includes TCM syndrome nodes, Western medicine pathophysiology nodes, TCM formula nodes, treatment method nodes, and their interrelationships. In practice, the comprehensive syndrome feature vector is compared with the TCM syndrome nodes in the knowledge graph, and the cosine similarity is calculated between the comprehensive syndrome feature vector and the feature vector of each TCM syndrome node to identify the target TCM syndrome node with the highest match to the patient. In some embodiments, starting from the target TCM syndrome node, the CMVD knowledge graph traverses the directly associated Western medicine pathophysiology nodes to obtain the associated Western medicine pathophysiology state description. This process continues by traversing the TCM formula nodes and treatment method nodes connected to both the target TCM syndrome node and the associated Western medicine pathophysiology nodes, obtaining a basic set of TCM formula candidates and treatment method candidates. This traversal process is understood to be based on predefined entity relationship edges in the knowledge graph. Based on the description of the pathophysiological state in Western medicine, corresponding standard Western medicine intervention measures are retrieved from the medical guideline database to form a candidate set of Western medicine intervention measures. The medical guideline database stores intervention recommendations from authoritative clinical guidelines in a structured manner.

[0097] In practice, all items within the basic candidate sets of traditional Chinese medicine (TCM) formulas, treatment methods, and Western medicine interventions are fully combined to generate a series of TCM-Western medicine integrated intervention strategy prototypes. Each prototype includes one TCM formula option, one treatment method option, and one Western medicine intervention option. Optionally, the full combination process generates all possible combinations using a loop iterative algorithm. Each TCM-Western medicine integrated intervention strategy prototype is input into a pre-defined incompatibility verification model, which includes a TCM 18-antagonism and 19-antagonism rule base, a TCM-Western medicine interaction knowledge base, and a specific treatment method contraindication knowledge base. The incompatibility verification model checks whether there are explicit incompatibilities between different items in the TCM-Western medicine integrated intervention strategy prototype. If any two items have an incompatibility relationship, the TCM-Western medicine integrated intervention strategy prototype is determined to be a conflicting combination and is removed from the candidate set. In some embodiments, contraindicated combinations include the simultaneous use of a traditional Chinese medicine formula containing "Fuzi" and a traditional Chinese medicine formula containing "Banxia", or the combined use of the traditional Chinese medicine "Danshen" and the Western medicine "warfarin", or "acupuncture therapy" for patients with bleeding tendencies.

[0098] For the prototype of the integrated traditional Chinese and Western medicine intervention strategy that has passed the drug incompatibility verification, it is input into a pre-set synergy assessment model. The synergy assessment model calculates the comprehensive synergy score of the internal items of the prototype based on the similarity of drug action mechanisms, the cumulative effect of efficacy, and the complementarity of action targets. The formula for calculating the comprehensive synergy score is expressed as follows:

[0099]

[0100] in: Indicates the comprehensive collaborative score. Sub-scores based on similarity of mechanisms of action This represents a sub-score based on the cumulative effect of therapeutic efficacy. This represents a sub-score based on target complementarity. These represent the weight coefficients of the corresponding sub-ratings, and It is understandable that the mechanism-of-action similarity sub-score assesses the consistency of different interventions acting on the same pathological pathway; the efficacy additive effect sub-score assesses whether different interventions have an additive effect in improving the same symptom; and the target complementarity sub-score assesses whether different interventions act on different key targets in the disease network. A synergy score threshold is set to identify and eliminate prototypes of integrated traditional Chinese and Western medicine intervention strategies with a comprehensive synergy score below the threshold as poorly compatible combinations. Optionally, the synergy score threshold is determined through data analysis of historically effective treatment protocols. All compatible prototypes of integrated traditional Chinese and Western medicine intervention strategies retained after incompatibility verification and synergy assessment are compiled into a preliminary set of integrated traditional Chinese and Western medicine intervention strategies.

[0101] In one embodiment of the present invention, see [reference] Figure 3 Peak oxygen uptake, anaerobic threshold, and maximum heart rate are extracted from exercise cardiopulmonary function test data. Aerobic exercise intensity zones are defined as a percentage of peak oxygen uptake, anaerobic exercise intensity thresholds are based on the power or heart rate corresponding to the anaerobic threshold, and the upper limit of heart rate control is defined as a percentage of maximum heart rate. For each strategy in the initial set of integrated traditional Chinese and Western medicine intervention strategies, the preset intensity of the exercise prescription recommendations is analyzed. The preset intensity is compared with the aerobic exercise intensity zone, anaerobic exercise intensity threshold, and upper limit of heart rate control to determine whether the preset intensity is within a safe and effective range. For strategies with preset intensities exceeding the safe and effective range, their exercise intensity is adjusted to the median of the aerobic exercise intensity zone or below the anaerobic exercise intensity threshold. For strategies with preset intensities involving drug intervention, the patient's tolerance to potential adverse drug reactions is assessed based on the cardiac functional reserve level reflected by peak oxygen uptake, and the initial drug dose or physiological indicators requiring close monitoring are recommended accordingly. After completing the evaluation and adjustment of all strategies, an optimized individualized intervention strategy is generated, labeled with the appropriate exercise intensity, drug usage precautions, and intensity justification.

[0102] In practical implementation, the tolerance optimization module assesses the tolerance and intensity of the initial integrated traditional Chinese and Western medicine intervention strategy set based on the extreme exercise parameters in the cardiopulmonary exercise test data. Peak oxygen uptake, anaerobic threshold, and maximum heart rate are extracted from the cardiopulmonary exercise test data; these parameters serve as objective bases for evaluating patients' exercise tolerance and cardiac reserve. In some embodiments, aerobic exercise intensity ranges are defined as a percentage of peak oxygen uptake, anaerobic exercise intensity thresholds are defined based on the power or heart rate corresponding to the anaerobic threshold, and the upper limit of heart rate control is defined as a percentage of maximum heart rate. For each strategy in the initial integrated traditional Chinese and Western medicine intervention strategy set, the preset intensity of its included exercise prescription recommendations is analyzed. The preset intensity may be expressed in the form of exercise metabolic equivalent, target heart rate range, or perceived fatigue rating. It is understood that comparing the preset intensity with the aerobic exercise intensity range, anaerobic exercise intensity threshold, and upper limit of heart rate control determines whether the preset intensity is within a safe and effective range. This process requires unifying the different forms of intensity expression into values ​​that can be directly compared with the aforementioned ranges and thresholds. Optionally, intensity comparisons can be performed using a preset conversion rule table, such as converting exercise metabolic equivalents into a percentage of the corresponding peak oxygen uptake.

[0103] In practice, for strategies where the preset intensity exceeds the safe and effective range, the exercise intensity is adjusted to the median of the aerobic exercise intensity range or below the anaerobic exercise intensity threshold. The magnitude of the intensity adjustment is calculated linearly based on the degree to which it exceeds the safe range. The adjusted exercise intensity calculation formula is expressed as follows:

[0104]

[0105] in: This indicates the adjusted exercise intensity. This represents the preset motion intensity parsed from the strategy. Indicates the lower limit of the safe and effective range. This indicates the upper limit of the safe and effective range. This represents an adjustment factor set based on clinical experience. For strategies with a preset intensity that include pharmacological intervention, the patient's tolerance to potential adverse drug reactions is assessed based on the cardiac functional reserve level reflected by peak oxygen uptake, and the starting drug dose or physiological indicators requiring close monitoring are recommended accordingly. In some embodiments, the correspondence between cardiac functional reserve level classifications and recommended drug dosage adjustments is shown in Table 1.

[0106] Table 1: Reference Table for Adjustment of Cardiac Functional Reserve Level and Drug Intervention

[0107]

[0108] Optionally, after evaluating and adjusting all strategies, an optimized individualized intervention strategy is generated. This optimized individualized intervention strategy includes information on the appropriate exercise intensity, precautions for medication use, and the basis for the intensity. The intensity basis references specific extreme exercise parameter values ​​and comparative safety ranges. It can be understood that the tolerance optimization module ensures, through the above process, that the exercise load of the intervention strategy matches the patient's objective physiological limits, while simultaneously providing early warnings regarding the safety of medication use based on quantitative functional assessments.

[0109] In one embodiment of the present invention, adapted TCM and Western medicine intervention measures are extracted from the optimized individualized intervention strategy. Constitution identification results are extracted from TCM four diagnostic methods data, with the results identified as one or more of the following: balanced constitution, qi deficiency constitution, yang deficiency constitution, yin deficiency constitution, phlegm-dampness constitution, damp-heat constitution, blood stasis constitution, qi stagnation constitution, and special constitution. Cardiovascular risk levels are extracted from routine physiological and biochemical indicator data. Based on the constitution identification results, the TCM formulas in the adapted TCM and Western medicine intervention measures are fine-tuned. A constitution-medicine property correspondence table is established, describing the appropriate medicinal property biases for different constitution types. The appropriate medicinal properties corresponding to the constitution identification results in the constitution-medicine property correspondence table are queried. The overall medicinal property bias of the TCM formulas in the adapted TCM and Western medicine intervention measures is analyzed. The overall medicinal property bias of the TCM formulas is compared with the appropriate medicinal properties corresponding to the constitution identification results. If the overall medicinal properties of a traditional Chinese medicine (TCM) formula deviate from the appropriate medicinal properties corresponding to the constitution identification results, one or more adjuvant herbs with corrective properties are added to the formula, or the dosage ratio of the principal and assistant herbs is adjusted to make the overall medicinal properties of the adjusted formula closer to the appropriate medicinal properties. The adjustment process of the TCM formula strictly follows the principles of TCM compatibility contraindications and ensures that the core therapeutic function of the formula remains unchanged. Based on the constitution identification results, a preference is given to the type of exercise prescription in the adapted TCM and Western medicine intervention measures. According to the cardiovascular risk level, the monitoring frequency, target range of monitoring indicators, and emergency response trigger conditions are set for the treatment plan. The fine-tuned TCM formula, the exercise prescription with type suggestions, the personalized lifestyle recommendations combined with constitution and risk level, and the set monitoring frequency, target range, and emergency response trigger conditions are integrated into a structured individualized treatment plan.

[0110] In practice, the treatment plan generation module integrates individualized intervention strategies, constitution identification results from TCM diagnostic information, and risk levels from routine physiological and biochemical indicators to generate individualized treatment plans. It extracts appropriate TCM and Western medicine interventions from the optimized individualized intervention strategies, including specific intervention methods adjusted based on tolerability assessments. Specifically, constitution identification results are extracted from the TCM diagnostic information, identifying one or more of the following constitutions: balanced, qi-deficient, yang-deficient, yin-deficient, phlegm-dampness, damp-heat, blood stasis, qi stagnation, and special constitution. Cardiovascular risk levels are extracted from routine physiological and biochemical indicators, categorized according to predefined stratification standards based on indicators such as blood lipids, blood glucose, and blood pressure. In essence, constitution identification results and cardiovascular risk levels provide supplementary information on the patient's homeostatic characteristics and pathological risk levels.

[0111] Based on the results of constitution identification, the traditional Chinese medicine prescriptions in the adapted integrated traditional Chinese and Western medicine interventions were fine-tuned, and a constitution-medicine property correspondence table was established to describe the appropriate medicinal property biases for different constitution types. See Table 2 for the constitution-medicine property correspondence table.

[0112] Table 2: Correspondence between Body Constitution Type and Suitable Medicinal Properties

[0113]

[0114] In practice, the appropriate medicinal properties corresponding to the constitution identification results are queried in the constitution-medicine property correspondence table. The overall medicinal property bias of the Chinese herbal medicine formula in the adapted TCM-Western medicine intervention measures is analyzed. The overall medicinal property bias of the Chinese herbal medicine formula is obtained by calculating the weighted average of the medicinal properties of each herb in the formula. The overall medicinal property bias of the Chinese herbal medicine formula is compared with the appropriate medicinal properties corresponding to the constitution identification results. If there is a deviation between the overall medicinal property bias of the Chinese herbal medicine formula and the appropriate medicinal properties corresponding to the constitution identification results, one or more adjuvant herbs with corrective medicinal properties are added to the Chinese herbal medicine formula, or the dosage ratio of the principal and assistant herbs in the formula is adjusted to make the overall medicinal properties of the adjusted formula closer to the appropriate medicinal properties. The deviation score between the overall medicinal property bias of the Chinese herbal medicine formula and the appropriate medicinal properties can be evaluated using the following formula:

[0115]

[0116] in: This represents the deviation score between the overall medicinal properties of a traditional Chinese medicine prescription and the medicinal properties suitable for the individual's constitution. This indicates the number of dimensions of the medicinal properties being examined. This indicates that the traditional Chinese medicine prescription is in the first... A comprehensive score based on each drug's properties. The appropriate medicinal properties corresponding to the constitution identification results are indicated in the [number]th [item]. Scoring based on individual drug properties Indicates the first The adjustment weights for each medicinal property dimension. In some embodiments, the adjustment process of traditional Chinese medicine prescriptions strictly follows the principles of compatibility and contraindications of traditional Chinese medicine combinations, and ensures that the core therapeutic function of the prescription remains unchanged. Optionally, the added adjuvant medicines must have medicinal properties that directly target the constitution imbalance, and the dosage is usually less than that of the assistant medicines.

[0117] Based on the constitution identification results, the exercise prescriptions in the adapted TCM and Western medicine interventions are given preferential suggestions. For example, for those with Qi deficiency or Yang deficiency constitutions, gentle exercises such as Baduanjin and Tai Chi are recommended, while for those with damp-heat or phlegm-dampness constitutions, more intense activities such as brisk walking and swimming are recommended. The monitoring frequency, target range of monitoring indicators, and emergency response trigger conditions are set according to the cardiovascular risk level. Optionally, when the cardiovascular risk level is high, the monitoring frequency is set to once a week, the target range is more stringent, and the emergency response trigger condition is immediate medical attention if specific symptoms appear. In some embodiments, the fine-tuned TCM prescription, the exercise prescription with type suggestions, the personalized lifestyle recommendations combined with constitution and risk level, and the set monitoring frequency, target range, and emergency response trigger conditions are integrated into a structured individualized treatment plan. It is understood that the structured individualized treatment plan is output as a standardized document template, containing clear chapters and items to facilitate clinical implementation.

[0118] In one embodiment of the present invention, during the implementation of a personalized treatment plan for a patient, periodic follow-up data is collected periodically. This periodic follow-up data includes simplified cardiopulmonary function indicators, changes in the four diagnostic methods (inspection, auscultation and olfaction, palpation, and olfaction), retest results of physiological and biochemical indicators, and the patient's subjective symptom scores. The periodic follow-up data is compared with baseline data before the implementation of the personalized treatment plan to generate a multidimensional efficacy evaluation vector. The plan optimization model is a model built on a reinforcement learning framework, which treats the treatment plan as a combination of actions taken by an agent and the improvement of the patient's health status as a reward. The multidimensional efficacy evaluation vector is encoded as a state vector, which describes the change in the patient's health status relative to the baseline after the current treatment cycle. The currently implemented personalized treatment plan is encoded as an action vector. An immediate reward value corresponding to the current state vector is calculated according to a preset reward function, which comprehensively considers the degree of improvement in physiological indicators, the degree of symptom relief, and safety indicators. The policy network and value network in the plan optimization model are updated using the state vector, action vector, and immediate reward value. The updated policy network outputs the probability distribution of adjustment suggestions for each intervention measure in the next treatment cycle, and the adjustment suggestion with the highest probability is selected as the output. Based on the adjustment recommendations, the individualized treatment plan was revised, generating a new version. The new individualized treatment plan, the basis for the adjustments, and the follow-up data were jointly stored in the patient's medical record and used as input for the next round of plan iteration.

[0119] In practice, after generating a personalized treatment plan, further treatment feedback and plan iteration are conducted. Periodic follow-up data is collected regularly during the patient's implementation of the personalized treatment plan. This periodic follow-up data includes simplified cardiopulmonary function indicators, changes in the four diagnostic methods (inspection, auscultation and olfaction, palpation, and olfaction), retest results of physiological and biochemical indicators, and the patient's subjective symptom scores. In practice, the periodic follow-up data is compared with baseline data before the implementation of the personalized treatment plan to generate a multidimensional efficacy evaluation vector. Each dimension in the multidimensional efficacy evaluation vector represents the change in a health indicator relative to the baseline. The plan optimization model is built based on a reinforcement learning framework. The plan optimization model treats the treatment plan as a combination of actions taken by an agent and considers the improvement of the patient's health status as a reward. It can be understood that the plan optimization model learns optimization strategies through continuous interaction with the environment (i.e., the patient's response to treatment).

[0120] In practice, the multidimensional efficacy assessment vector is encoded as a state vector. The state vector describes the change in the patient's health status relative to the baseline after the current treatment cycle ends, and serves as the state input to the reinforcement learning model. The currently executed individualized treatment plan is encoded as an action vector, which includes parameterized representations such as the type, dosage, and frequency of treatment measures. The immediate reward value corresponding to the current state vector is calculated based on a preset reward function, which comprehensively considers the improvement of physiological indicators, symptom relief, and safety indicators. The formula for calculating the immediate reward value is as follows:

[0121]

[0122] in: This represents the immediate reward value calculated at the end of the current treatment cycle t. This indicates the number of positive efficacy indicators. Indicates the first The improvement weight of each positive efficacy indicator, Indicates the first The change in a positive efficacy indicator relative to baseline. The number of indicators representing safety or negative metrics. Indicates the first The penalty weight for each negative indicator, Indicates the first The degree or frequency of occurrence of a negative indicator. In some embodiments, the weighting coefficients in the reward function are set based on clinical expert consensus.

[0123] The policy network and value network in the scheme optimization model are updated using state vectors, action vectors, and immediate reward values. The policy network generates the probability distribution of actions, and the value network evaluates the value of states. The update process employs a policy gradient algorithm or an actor-critic algorithm. The updated policy network outputs the probability distribution of adjustment suggestions for each intervention in the next treatment cycle. These suggestions include adding, reducing, maintaining, or replacing an intervention, as well as the intensity or frequency of the adjustment. In some embodiments, the adjustment suggestion with the highest probability is selected as the output. Optionally, the individualized treatment plan is revised based on the adjustment suggestions to generate a new version of the individualized treatment plan. The revision process adjusts specific parameters or components of the original plan while adhering to clinical safety guidelines. It is understood that the new individualized treatment plan, the basis for the adjustments, and the current follow-up data are jointly stored in the patient's medical record and used as input for the next round of plan iteration, thereby achieving a dynamic and personalized optimization cycle of the treatment plan.

[0124] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A platform for generating individualized treatment plans for CMVD that integrates CPET and the four diagnostic methods, characterized in that: include: The data acquisition module collects a multidimensional health data set of CMVD patients, which includes exercise cardiopulmonary function test data, traditional Chinese medicine four diagnostic information data, and routine physiological and biochemical indicator data. The feature fusion module uses an improved hierarchical clustering algorithm to perform data fusion and feature extraction on the multidimensional health data set to obtain the comprehensive syndrome feature vector of the patient. The improved hierarchical clustering algorithm dynamically adjusts the feature distance based on the weight of the TCM syndrome elements. The strategy matching module maps and matches the comprehensive syndrome feature vector with the preset CMVD Chinese and Western medicine diagnosis and treatment knowledge graph to generate a preliminary set of Chinese and Western medicine integrated intervention strategies. The tolerance optimization module, based on the extreme exercise parameters in the exercise cardiopulmonary function test data, performs tolerance assessment and intensity adaptation on the preliminary set of integrated traditional Chinese and Western medicine intervention strategies, and generates an optimized individualized intervention strategy. The treatment plan generation module integrates the individualized intervention strategy, the constitution identification results in the TCM four diagnostic methods information data, and the risk level in the routine physiological and biochemical indicator data to generate an individualized treatment plan that includes TCM prescriptions, exercise prescriptions, lifestyle recommendations, and monitoring indicators. The improved hierarchical clustering algorithm dynamically adjusts the feature distance based on the weights of TCM syndrome elements, including: Construct an initial feature space, using each data item in the multidimensional health data set as a dimension of the initial feature space; A basic weight is assigned to each dimension in the initial feature space, and the basic weight is determined based on the importance of the data represented by the dimension in the CMVD diagnosis and treatment guidelines; Based on the TCM four diagnostic methods data, the distribution of TCM syndrome elements is extracted. The TCM syndrome elements include Qi deficiency, blood stasis, phlegm turbidity, Yin deficiency, and Yang deficiency. Based on the distribution of the extracted TCM syndrome elements, the weights of the corresponding dimensions in the initial feature space are dynamically adjusted to form a dynamic weight feature space. The weights of dimensions with high correlation to the main syndrome elements are increased, while the weights of dimensions with low correlation are decreased. In the dynamic weight feature space, the weighted Euclidean distance between any two patient data samples is calculated, and the weighted Euclidean distance is calculated using the weights in the dynamic weight feature space. Based on the calculated weighted Euclidean distance, a bottom-up aggregation strategy is used for hierarchical clustering. In each aggregation, the two classes with the smallest inter-class distance are selected for merging, and the distance between the new class and other classes in the dynamic weight feature space is recalculated. The inter-class distance is calculated using the Ward method, with the goal of minimizing the increase in intra-class variance. Iteratively execute the aggregation and distance recalculation process until all samples are aggregated into one class or the preset number of clusters is reached, and output the final clustering result; From each category of the final clustering result, the feature values ​​of its center point are extracted to form a comprehensive syndrome feature vector representing the characteristics of the patient group in the corresponding category.

2. The CMVD individualized treatment plan generation platform integrating CPET and four diagnostic methods as described in claim 1, characterized in that, The comprehensive syndrome feature vector is mapped and matched with a pre-defined CMVD (Chinese and Western medicine) diagnostic and treatment knowledge graph to generate a preliminary set of integrated Chinese and Western medicine intervention strategies, including: The CMVD knowledge graph of Chinese and Western medicine diagnosis and treatment includes nodes for Chinese medicine syndrome types, nodes for Western medicine pathophysiology, nodes for Chinese medicine prescriptions, nodes for treatment methods, and the relationships between them. The similarity between the comprehensive syndrome feature vector and the TCM syndrome node is calculated to identify the target TCM syndrome node with the highest matching degree with the patient; In the CMVD knowledge graph of Chinese and Western medicine diagnosis and treatment, starting from the target TCM syndrome node, the Western medicine pathophysiological nodes directly associated with it are traversed to obtain the associated Western medicine pathophysiological state description. Continue traversing the nodes of traditional Chinese medicine prescriptions and treatment methods that are connected to the target TCM syndrome node and the associated Western medicine pathophysiology node, to obtain a basic set of candidate TCM prescriptions and candidate treatment methods. Based on the description of the Western medicine pathophysiological state, corresponding standard Western medicine intervention measures are retrieved from the medical guideline database to form a candidate set of Western medicine intervention measures; The candidate sets of basic Chinese herbal formulas, treatment methods, and Western medicine interventions are combined, and the compatibility and synergistic relationships between the items within each combination are verified. Conflicting combinations are eliminated, and compatible combinations are retained to form the preliminary set of integrated Chinese and Western medicine intervention strategies.

3. The CMVD individualized treatment plan generation platform integrating CPET and four diagnostic methods as described in claim 2, characterized in that, Based on the extreme exercise parameters in the aforementioned cardiopulmonary exercise test data, the tolerance and intensity of the preliminary integrated traditional Chinese and Western medicine intervention strategy set are assessed to generate an optimized individualized intervention strategy, including: Peak oxygen uptake, anaerobic threshold, and maximum heart rate were extracted from the exercise cardiopulmonary function test data. Set an aerobic exercise intensity range expressed as a percentage of the peak oxygen uptake, an anaerobic exercise intensity threshold based on the power or heart rate corresponding to the anaerobic threshold, and use a percentage of the maximum heart rate as the upper limit of heart rate control. For each strategy in the preliminary set of integrated traditional Chinese and Western medicine intervention strategies, the preset intensity of the exercise prescription recommendations contained therein is analyzed; The preset intensity is compared with the aerobic exercise intensity range, the anaerobic exercise intensity threshold, and the heart rate control upper limit to determine whether the preset intensity is within a safe and effective range. For strategies whose preset intensity exceeds the safe and effective range, the exercise intensity shall be adjusted to the median of the aerobic exercise intensity range or below the anaerobic exercise intensity threshold. For a strategy with a preset intensity that includes drug intervention, the patient's tolerance to potential adverse drug reactions is assessed based on the cardiac function reserve level reflected by the peak oxygen uptake, and the starting drug dose or physiological indicators that need to be closely monitored are recommended accordingly. After evaluating and adjusting all strategies, the optimized individualized intervention strategy is generated, labeled with the appropriate exercise intensity, precautions for drug use, and intensity basis.

4. The CMVD individualized treatment plan generation platform integrating CPET and four diagnostic methods as described in claim 3, is characterized in that, Integrating the individualized intervention strategy, the constitution identification results from the four diagnostic methods of traditional Chinese medicine, and the risk levels from the routine physiological and biochemical indicators, an individualized treatment plan is generated, including a traditional Chinese medicine prescription, an exercise prescription, lifestyle recommendations, and monitoring indicators. Extract appropriate integrated traditional Chinese and Western medicine intervention measures from the optimized individualized intervention strategy; The constitution identification results are extracted from the TCM four diagnostic information data. The constitution identification results are one or more of the following: balanced constitution, qi deficiency constitution, yang deficiency constitution, yin deficiency constitution, phlegm-dampness constitution, damp-heat constitution, blood stasis constitution, qi stagnation constitution, and special constitution. Based on the constitution identification results, the traditional Chinese medicine prescriptions in the adapted Chinese and Western medicine intervention measures are fine-tuned. The fine-tuning includes increasing or decreasing the dosage of specific herbs according to the constitution, or adding herbs that are targeted at conditioning the constitution. Based on the results of the constitution identification, we propose tendencies for the types of exercise prescriptions in the adapted Chinese and Western medicine intervention measures; Extract cardiovascular risk levels from the aforementioned routine physiological and biochemical indicator data; Based on the cardiovascular risk level, the monitoring frequency, target range of monitoring indicators, and emergency response triggering conditions of the treatment plan are set. The finely adjusted traditional Chinese medicine prescription, the accompanying exercise prescription with type suggestions, the personalized lifestyle recommendations based on physical condition and risk level, and the set monitoring frequency, target range and emergency response trigger conditions are integrated into a structured individualized treatment plan.

5. The CMVD individualized treatment plan generation platform integrating CPET and four diagnostic methods as described in claim 4, characterized in that, Based on the constitution identification results, the traditional Chinese medicine prescriptions in the adapted integrated traditional Chinese and Western medicine intervention measures are fine-tuned, including: Establish a constitution-medicine property correspondence table, which describes the appropriate medicinal property bias for different constitution types; Query the appropriate medicinal properties corresponding to the constitution identification results in the constitution-medicine property correspondence table; Analyze the overall medicinal properties of the traditional Chinese medicine prescriptions in the adapted integrated traditional Chinese and Western medicine intervention measures; Compare the overall medicinal properties of the Chinese herbal formula with the appropriate medicinal properties corresponding to the constitution identification results; If there is a discrepancy between the overall medicinal properties of a traditional Chinese medicine prescription and the appropriate medicinal properties corresponding to the constitution identification results, one or more adjuvant herbs with corrective medicinal properties may be added to the prescription, or the dosage ratio of the principal and assistant herbs in the prescription may be adjusted so that the overall medicinal properties of the adjusted prescription are closer to the appropriate medicinal properties. The process of adjusting traditional Chinese medicine prescriptions strictly follows the principle of incompatibilities in Chinese medicine combinations, and ensures that the core therapeutic function of the prescription remains unchanged.

6. The CMVD individualized treatment plan generation platform integrating CPET and four diagnostic methods as described in claim 5, characterized in that, In the dynamic weighted feature space, the weighted Euclidean distance between any two patient data samples is calculated. The calculation of the weighted Euclidean distance uses the weights in the dynamic weighted feature space, including: Obtain the weights of all dimensions in the dynamic weight feature space to form a weight vector; For any two patient data samples, extract their feature values ​​in all dimensions of the dynamic weight feature space to form their respective feature vectors; Calculate the difference between two feature vectors along each dimension; Squaring the differences in each dimension; Multiply the difference in each dimension after squaring by the weight of the corresponding dimension in the weight vector; The sum is obtained by summing the weighted squared differences across all dimensions; The sum is square-rooted, and the result is used as the weighted Euclidean distance between the two patient data samples.

7. The CMVD individualized treatment plan generation platform integrating CPET and four diagnostic methods as described in claim 6, characterized in that, The candidate sets of basic traditional Chinese medicine formulas, treatment methods, and Western medicine interventions are combined, and the compatibility and synergistic relationships between items within each combination are verified. Conflicting combinations are eliminated, and compatible combinations are retained to form the preliminary set of integrated traditional Chinese and Western medicine intervention strategies, including: All items in the basic set of candidate Chinese herbal formulas, the set of candidate treatment methods, and the set of candidate Western medicine interventions are combined to generate a series of prototype intervention strategies that combine Chinese and Western medicine. Each of the aforementioned strategy prototypes is input into a preset compatibility check model, which includes a rule base of eighteen incompatibilities and nineteen antagonisms of traditional Chinese medicine, a knowledge base of interactions between traditional Chinese and Western medicine, and a knowledge base of contraindications for specific treatment methods. The matching taboo verification model is used to check whether there are explicit matching taboos between different items in the strategy prototype. If there is a taboo relationship between any two items, the strategy prototype is determined to be a conflicting combination and is removed from the candidate set. For the strategy prototype that passes the incompatibility check, it is input into a preset synergy evaluation model. The synergy evaluation model calculates the comprehensive synergy score of the internal items of the strategy prototype based on the similarity of drug action mechanism, the superposition effect of efficacy and the complementarity of action target. Set a collaboration score threshold, and identify strategy prototypes with a comprehensive collaboration score below the threshold as combinations with poor compatibility and remove them; All compatible strategy prototypes retained after the incompatibilities check and synergy assessment are compiled into the preliminary set of integrated traditional Chinese and Western medicine intervention strategies.

8. The CMVD individualized treatment plan generation platform integrating CPET and four diagnostic methods as described in claim 7, characterized in that, After generating the individualized treatment plan, further treatment feedback and plan iteration are performed, including: During the period when the patient is implementing the individualized treatment plan, periodic follow-up data will be collected regularly. The periodic follow-up data includes simplified cardiopulmonary function indicators, changes in the four diagnostic methods, results of physiological and biochemical indicator retests, and the patient's subjective symptom scores. The periodic follow-up data is compared with the baseline data before the implementation of the individualized treatment plan to generate a multidimensional efficacy assessment vector; The multidimensional efficacy assessment vector is input into a protocol optimization model, which outputs adjustment suggestions for each intervention in the current treatment protocol. The adjustment suggestions include adding, reducing, maintaining or replacing a certain measure, as well as the intensity or frequency of the adjustment measure. Based on the aforementioned adjustment recommendations, the individualized treatment plan is revised to generate a new version of the individualized treatment plan; The new individualized treatment plan, the basis for adjustment, and the follow-up data will be stored together in the patient's medical record and used as input for the next round of plan iteration.

9. The CMVD individualized treatment plan generation platform integrating CPET and four diagnostic methods as described in claim 8, characterized in that, The multidimensional efficacy assessment vector is input into a treatment plan optimization model, which outputs adjustment suggestions for various interventions in the current treatment plan, including: The optimization model described is a model built on a reinforcement learning framework, which treats the treatment plan as a combination of actions taken by the agent and regards the improvement of the patient's health status as a reward. The multidimensional efficacy assessment vector is encoded as a state vector, which describes the change in the patient’s health status relative to the baseline at the end of the current treatment cycle. The currently executed individualized treatment plan is encoded as an action vector; The immediate reward value corresponding to the current state vector is calculated according to a preset reward function, which comprehensively considers the degree of improvement of physiological indicators, the degree of symptom relief, and the safety indicators. The policy network and value network in the scheme optimization model are updated using the state vector, the action vector, and the instant reward value. The updated strategy network outputs the probability distribution of adjustment suggestions for each intervention in the next treatment cycle, and selects the adjustment suggestion with the highest probability as the output.

Images