Traditional Chinese medicine physiotherapy effect feedback system and method

Abstract

The invention discloses a traditional Chinese medicine physiotherapy effect feedback system and method, and the method comprises the steps: obtaining the physiological, pulse, tongue and subjective feeling information of a patient through a multi-modal data collection and preprocessing module, and mapping the information into a high-dimensional and quantized traditional Chinese medicine feature vector through a multi-layer attention mechanism conversion network; a'physical therapy-syndrome-effect 'dynamic quantitative correlation model is constructed, a reinforcement learning and causal inference mechanism is built in the model, the causal influence of physical therapy intervention on traditional Chinese medicine syndromes can be quantified, dynamic evolution of the syndromes can be predicted, the syndromes can be converted into clinical effects, and therefore traditional Chinese medicine mechanism feedback which can be deeply explained is provided. On the basis, the system evaluates the comprehensive curative effect in real time through an original traditional Chinese medicine physical therapy dynamic curative effect index generation module, and provides highly personalized physical therapy scheme optimization suggestions based on a traditional Chinese medicine mechanism through an intelligent feedback and intervention recommendation module in combination with a traditional Chinese medicine knowledge graph and an individual effect file. And deep intelligent evaluation and real-time optimization of the traditional Chinese medicine physiotherapy effect are realized.

Description

Technical Field

[0001] This invention belongs to the fields of medical devices, artificial intelligence, and modernization of traditional Chinese medicine, and in particular relates to a deep intelligent evaluation and real-time personalized feedback technology for the effects of traditional Chinese medicine physiotherapy. It aims to break through the limitations of traditional evaluation and provide a quantitative, objective, and interpretable physiotherapy effect feedback system and method based on dynamic mapping of traditional Chinese medicine feature space. Background Technology

[0002] The evaluation of the efficacy of traditional Chinese medicine (TCM) physiotherapy has long faced challenges in achieving objectivity, standardization, and personalization. Current technologies largely remain at the level of:

[0003] Monitoring single physiological indicators, such as heart rate, blood pressure, and body temperature, lacks the holistic concept of syndrome differentiation and treatment in traditional Chinese medicine.

[0004] Simple multimodal data overlay: This involves simple fusion or parallel analysis of multi-source data, but fails to establish a deep, quantitative, and dynamic mapping relationship between the data and the core theories of traditional Chinese medicine (such as viscera, meridians, qi, blood, body fluids, and syndromes).

[0005] Rule-based expert systems lack self-learning and adaptive capabilities, making it difficult to handle complex and ever-changing clinical situations and individual differences.

[0006] Limited applications of machine learning: Most remain at the level of classification or regression prediction, failing to deeply explain the mechanisms of TCM that are "effective" or "ineffective".

[0007] Delayed or uninstructive feedback: Assessment results are usually post-event summaries, making it difficult to provide real-time guidance for adjusting the physical therapy plan, and the suggestions given are often too general.

[0008] For example, while existing technologies include systems that combine pulse diagnosis and tongue diagnosis for auxiliary diagnosis, their evaluation logic is typically as follows: the pulse diagnosis instrument identifies a "wiry pulse," the tongue diagnosis instrument identifies a "red tongue with yellow coating," and then these are matched according to rules to form a "liver fire excess" syndrome. This static "feature-syndrome" matching lacks a quantitative and mechanistic explanation of the dynamic evolution process of "feature-syndrome-efficacy" caused by physical therapy intervention. During physical therapy, the patient's physiological state and TCM syndromes are dynamically changing, and existing systems struggle to capture and quantify the contribution of these changes to the therapeutic effect. Summary of the Invention

[0009] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0010] The purpose of this invention is to provide a physiotherapy effect feedback system and method based on dynamic mapping of TCM feature space. By constructing a multi-dimensional TCM feature space and a dynamic quantitative correlation model of "physiotherapy-syndrome-effect", it can achieve in-depth interpretability, real-time quantification and personalized feedback on the effect of TCM physiotherapy, and can intelligently provide scheme optimization suggestions based on TCM mechanism.

[0011] Specifically, the present invention provides the following technical solution: a Traditional Chinese Medicine (TCM) physiotherapy effect feedback system, comprising the following components: a) a data processing device, including at least one processor and a memory, wherein the memory stores computer program instructions, and the processor executes the computer program instructions to achieve the functions of the following modules: b) a TCM feature space mapping module, the module including a multi-layer attention mechanism transformation network (MATN); the MATN is configured to: receive preprocessed physiological parameter data, pulse data, tongue data, and subjective feeling data, and through the modal encoder, cross-modal attention fusion layer, and TCM feature mapping layer within the MATN, process the received data to generate a high-dimensional, continuous, and quantized TCM feature vector V. TCM c) The TTSEDQM dynamic quantization correlation model consists of a series of interconnected neural networks stored in memory. The neural networks include: i. an initial syndrome identification and quantification network (PSQN), whose input is connected to the output of the TCM feature space mapping module to receive the TCM feature vector V. TCM,pre It outputs an initial syndrome state vector S. pre ii. A TIEN network for quantifying the effects of a physical therapy intervention, comprising a graph neural network (GNN) and a causal inference component, the input of which is configured to receive data A representing the current physical therapy intervention. t and the initial symptom state vector S pre It outputs a symptom influence vector ΔS(A) that quantifies the effect of physical therapy intervention on the syndrome. t iii. A syndrome dynamic evolution prediction network SDEN, whose input is configured to receive an initial syndrome state vector S. pre The influence vector of the syndrome ΔS(A) t It combines historical physiotherapy intervention data and outputs a real-time syndrome state vector S after physiotherapy. post ;iv. A clinical effect quantification network (CEQN) whose input is configured to receive a real-time symptom state vector S post Traditional Chinese Medicine Feature Vector V after Physiotherapy TCM,post It outputs a clinical effect vector E that quantifies the clinical effect. clinical d) The Dynamic Efficacy Index (DEI) generation module is configured to: generate the DEI based on a preset model, combined with the clinical effect vector E. clinical Initial syndrome state vector S pre and real-time syndrome state vector Spost The difference between the values ​​and the KL divergence between the TCM feature vector and a target health state vector are used to calculate and generate a dynamic therapeutic efficacy index (DEI) value.

[0012] As a preferred embodiment of the TCM physiotherapy effect feedback system described in this invention, it further includes: a multimodal data acquisition device, specifically including: a three-dimensional multi-point piezoelectric array pulse sensor for acquiring a three-dimensional pulse waveform array, and a multispectral intelligent tongue image acquisition device for acquiring multispectral images and constructing a three-dimensional tongue model.

[0013] As a preferred embodiment of the TCM physiotherapy effect feedback system described in this invention, it further includes an intelligent feedback and intervention recommendation module; specifically, it includes a recommendation engine; the recommendation engine is configured to: a) receive the strategy output by the reinforcement learning component in the dynamic quantitative association model TTSEDQM; b) query the TCM acupoint / intervention knowledge graph stored in the memory based on the strategy; c) generate a data signal containing optimization suggestions for the physiotherapy plan based on the query results.

[0014] As a preferred embodiment of the TCM physiotherapy effect feedback system described in this invention, the TCM acupoint / intervention knowledge graph stored in the memory is a data structure that includes entities such as acupoints, meridians, syndromes, and effects, as well as the relationships between them.

[0015] Additionally, this invention also provides the following technical solution: a method for feedback on the effect of traditional Chinese medicine physiotherapy, which is executed on a computing device including a processor and a memory, comprising the following steps: a) receiving physiological parameter data, pulse data, tongue data, and subjective feeling data collected and preprocessed by multimodal sensors; b) processing the data received in step a) through a multi-layer attention mechanism transformation network (MATN) to generate a high-dimensional, continuous, and quantified traditional Chinese medicine feature vector V. TCM c) The TCM feature vector V generated in step b) TCM The input is processed into a dynamic quantization association model (TTSEDQM). This step specifically includes: i. Generating an initial symptom state vector S through an initial symptom identification and quantization network (PSQN) within the TTSEDQM. pre ii. Using the TIEN network, a physical therapy intervention effect quantification network within the TTSEDQM, based on data A representing the current physical therapy intervention... t and the initial symptom state vector S pre Generate a syndrome influence vector ΔS(A) t iii. Using the SDEN network, a symptom dynamic evolution prediction network in TTSEDQM, based on the initial symptom state vector S pre and the influence vector of symptoms ΔS(A) tGenerate a real-time syndrome state vector S after physiotherapy. post iv. Using the Clinical Effect Quantification Network (CEQN) within the TTSEDQM, based on the real-time syndrome state vector S post Traditional Chinese Medicine Feature Vector V after Physiotherapy TCM,post Generate a clinical effect vector E clinical d) Based on the pre-set model, combined with the clinical effect vector E generated in step c)iv, clinical Initial syndrome state vector S pre With the real-time symptom state vector S post The difference between the values ​​of the TCM feature vector and the KL divergence between the TCM feature vector and a target health state vector are used to calculate a dynamic therapeutic efficacy index (DEI) value.

[0016] As a preferred embodiment of the TCM physiotherapy effect feedback method described in this invention, the cross-modal attention fusion layer of the Multilayer Attention Mechanism Transformation Network (MATN) employs self-attention and cross-attention mechanisms to determine the fusion weights between different modal data features.

[0017] As a preferred embodiment of the TCM physiotherapy effect feedback method described in this invention, the initial syndrome identification and quantification network PSQN is a neural network based on the Transformer architecture.

[0018] As a preferred embodiment of the TCM physiotherapy effect feedback method described in this invention, the syndrome dynamic evolution prediction network SDEN is a recurrent neural network (RNN) or Transformer network with an attention mechanism.

[0019] As a preferred embodiment of the TCM physiotherapy effect feedback method described in this invention, the preset model upon which the dynamic efficacy index (DEI) generation module is based is: DEI t =w1·E clinical,t overall +w2·(∑ j＝1 L max(0,S pre,j -S post,j )·β j )+w3·KL_Divergence(V TCM,t V TCM,target )·γ t Among them, E clinical,t overall S is the weighted average of the clinical effect vectors. pre,j and S post,j These are the quantitative values ​​of the j-th symptom before and after physiotherapy, β j V represents the importance weight of the j-th symptom. TCM,t V represents the current feature vector of Traditional Chinese Medicine.TCM,target Let w1, w2, w3, γ be the preset target health state vector, and KL_Divergence be the Kullback-Leibler divergence between the two. t These are the preset weighting coefficients.

[0020] As a preferred embodiment of the TCM physiotherapy effect feedback method described in this invention, the method further includes the following steps: a) receiving a strategy output by the reinforcement learning component in the TTSEDQM dynamic quantitative association model; b) querying a preset TCM acupoint / intervention knowledge graph based on the strategy; c) generating and outputting a data signal containing optimization suggestions for the physiotherapy plan based on the query results.

[0021] This invention provides a feedback system and method for the effects of traditional Chinese medicine physiotherapy, which has the following beneficial effects:

[0022] 1. It achieves a quantitative, dynamic, and interpretable correlation from "phenomenon" to "evolution of TCM syndrome" and then to "clinical effect", rather than a simple judgment of efficacy, which greatly enhances the scientific nature and persuasiveness of the system.

[0023] 2. It breaks through the limitations of traditional multimodal data fusion, constructs a high-dimensional quantitative feature space with TCM connotations, and captures the dynamic impact of physical therapy intervention on the body's deep TCM state through a complex dynamic correlation model.

[0024] 3. It integrates multi-dimensional information, especially the quantitative value of symptom improvement, and provides a more comprehensive real-time efficacy evaluation indicator that is more in line with the holistic view of traditional Chinese medicine.

[0025] 4. It can intelligently recommend adjustments to physiotherapy plans based on TCM mechanisms according to individual TCM characteristics, syndrome dynamics, and historical responses, achieving true precision medicine and personalized intervention. Detailed Implementation

[0026] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0027] The purpose of this invention is to provide a physiotherapy effect feedback system based on dynamic mapping of TCM feature space. By constructing a multi-dimensional TCM feature space and a dynamic quantitative correlation model of "physiotherapy-syndrome-effect", it can achieve in-depth interpretability, real-time quantification and personalized feedback on the effect of TCM physiotherapy, and can intelligently provide scheme optimization suggestions based on TCM mechanism.

[0028] Specifically, the present invention provides a feedback system for the effect of traditional Chinese medicine physiotherapy, comprising the following components:

[0029] a) A data processing apparatus, comprising at least one processor and a memory, wherein the memory stores computer program instructions, and the processor executes the computer program instructions to perform the functions of the following modules:

[0030] b) Traditional Chinese Medicine Feature Space Mapping Module: This module includes a multi-layer attention mechanism transformation network (MATN). The MATN is configured to receive preprocessed physiological parameter data, pulse data, tongue data, and subjective feeling data, and process the received data to generate a high-dimensional, continuous, and quantized traditional Chinese medicine feature vector V through a modal encoder, a cross-modal attention fusion layer, and a traditional Chinese medicine feature mapping layer within the MATN. TCM ;

[0031] c) The Dynamic Quantization Relation Model (TTSEDQM) consists of a series of interconnected neural networks stored in memory. These neural networks include:

[0032] i. An initial syndrome identification and quantification network (PSQN) is established, with its input connected to the output of a traditional Chinese medicine feature space mapping module to receive the traditional Chinese medicine feature vector V. TCM,pre It outputs an initial syndrome state vector S. pre ;

[0033] ii. A TIEN network for quantifying the effects of a physical therapy intervention, comprising a graph neural network (GNN) and a causal inference component, the input of which is configured to receive data A representing the current physical therapy intervention. t and the initial symptom state vector S pre It outputs a symptom influence vector ΔS(A) that quantifies the effect of physical therapy intervention on the syndrome. t );

[0034] iii. A syndrome dynamic evolution prediction network SDEN, whose input is configured to receive an initial syndrome state vector S pre The influence vector of the syndrome ΔS(A) t It combines historical physiotherapy intervention data and outputs a real-time syndrome state vector S after physiotherapy. post ;

[0035] iv. A clinical effect quantification network (CEQN) whose input is configured to receive a real-time symptom state vector S. post Traditional Chinese Medicine Feature Vector V after Physiotherapy TCM,post It outputs a clinical effect vector E that quantifies the clinical effect. clinical ;

[0036] d) The Dynamic Efficacy Index (DEI) generation module is configured to: generate the DEI based on a preset model, combined with the clinical effect vector E. clinical Initial syndrome state vector S pre and real-time syndrome state vector S post The difference between the values ​​and the KL divergence between the TCM feature vector and a target health state vector are used to calculate and generate a dynamic therapeutic efficacy index (DEI) value.

[0037] Furthermore, it also includes: a multimodal data acquisition device, specifically including: a three-dimensional multi-point piezoelectric array pulse sensor for acquiring a three-dimensional pulse waveform array, and a multispectral intelligent tongue image acquisition device for acquiring multispectral images and constructing a three-dimensional tongue model.

[0038] Furthermore, it also includes an intelligent feedback and intervention recommendation module;

[0039] Specifically, this includes a recommendation engine; the recommendation engine is configured as follows:

[0040] a) Receive the policy output by the reinforcement learning component in the TTSEDQM dynamic quantization association model;

[0041] b) Based on the strategy, query the TCM acupoint / intervention knowledge graph stored in memory;

[0042] c) Based on the query results, generate a data signal containing suggestions for optimizing the physiotherapy plan.

[0043] Specifically, the TCM acupoint / intervention knowledge graph stored in the memory is a data structure that includes entities such as acupoints, meridians, syndromes, and effects, as well as the relationships between them.

[0044] It should be noted that:

[0045] 1. Multimodal data acquisition and preprocessing module

[0046] ① Smart wearable physiological sensor array:

[0047] High-precision skin temperature sensor (Temp_Sensor_Array): A multi-point array deployed at specific acupoint groups (such as back-shu points, front-mu points, and limb source points) to monitor subtle changes in local temperature. Parameters: Spatial resolution (R spatial_temp Time resolution (R) temporal_temp Temperature accuracy (ΔT). The goal is to capture the effect of "thermosensitive acupoints" or local changes in Qi and blood.

[0048] High-sensitivity electromyography (sEMG_Sensor_Array): Multi-channel, targeting muscle groups involved in physical therapy. Parameter: Number of channels (N) channel_emg ), bandwidth (B emgCommon-mode rejection ratio (CMRR) emg Used to assess muscle tension and fatigue.

[0049] High-precision HRV and skin conductance sensor (HRV_GSR_Sensor): Integrated design for monitoring autonomic nervous system activity and emotional state. Parameters: Sampling frequency (f sample_hrv f sample_gsr ), resolution (Δ hrv Δ gsr ).

[0050] Microcirculation Monitor: Employs laser Doppler or video microscopy technology for non-invasive monitoring of microvascular blood perfusion, reflecting local blood and gas circulation. Parameters: Measurement depth (D) mc Spatial resolution (R) spatial_mc ).

[0051] ② High-precision pulse sensor (Pulse_Sensor_Advanced):

[0052] Three-dimensional multi-point piezoelectric array sensor: Breaking through the traditional single-point or one-dimensional pulse diagnosis, it can simultaneously collect pressure waveforms at multiple points (such as 9 o'clock or 12 o'clock) in the three positions of Cun, Guan, and Chi, and can also sense depth information.

[0053] Parameters: Array size (M×N sensor units), pressure range per unit (P) range_unit ), depth resolution (D res_pulse ), waveform sampling rate (f sample_pulse ).

[0054] Preprocessing: signal denoising, baseline drift correction, time-frequency analysis (wavelet transform, Hilbert-Huang transform) to extract multi-scale time-frequency features of the pulse wave.

[0055] ③ Intelligent Tongue Imaging Device (Tongue_Imaging_Device_Pro):

[0056] Multispectral image acquisition: In addition to visible light, near-infrared and ultraviolet bands are added to capture deep physiological information of the tongue and tongue coating (such as the ratio of oxyhemoglobin to deoxyhemoglobin, reflecting blood circulation and metabolic status).

[0057] 3D tongue modeling: Reconstructing the three-dimensional shape of the tongue through structured light or multi-view images, and accurately analyzing the size, thickness, and crack depth of the tongue.

[0058] Parameters: Visible light resolution (R vis ), number of spectral bands (N) band ), 3D reconstruction accuracy (Δ 3D).

[0059] ④ Structured subjective experience input:

[0060] Adaptive questionnaire system: The questionnaire content is dynamically adjusted according to the patient's condition and physical therapy plan to avoid fatigue.

[0061] Natural Language Processing (NLP) module: Performs semantic analysis and entity extraction on the patient's verbal symptoms and feelings, quantifying them into standardized indicators such as "pain level", "fatigue", and "mental state".

[0062] ⑤ Data cleaning, calibration, and normalization: Standard signal processing and data engineering procedures to ensure data quality and consistency.

[0063] 2. Traditional Chinese Medicine Feature Spatial Mapping Module

[0064] This module is one of the key components of this invention. It transforms heterogeneous raw data into a unified, high-dimensional quantized feature vector with TCM interpretive significance, thus constructing a "TCM feature space".

[0065] Module objective: To integrate raw physiological, pulse, tongue, and subjective feeling data into D... raw ={D phy D pulse D tongue D subj Mapped to TCM feature vector V TCM ∈R K , where K is the dimension of the TCM feature space.

[0066] Core network: Multi-Attention Transformation Network (MATN).

[0067] Input: Preprocessed features of each modality.

[0068] structure:

[0069] Modality Encoders: Each modality (physiological, pulse, tongue, subjective) has an independent deep learning encoder (e.g., CNN for tongue, RNN / Transformer for pulse time series data, MLP for physiological parameters), which encodes its raw features into a low-dimensional embedding vector D. phy D pulse D tongue D subj .

[0070] Cross-Modality Attention Fusion Layer:

[0071] Objective: To capture the interactions and contribution weights among different modal features.

[0072] Mechanism: It employs self-attention and cross-attention mechanisms. For example, tongue color may be influenced by pulse, and MATN learns this association.

[0073] Output: The fused feature vector E fused =Attention(D phy D pulse D tongue D subj ).

[0074] The Traditional Chinese Medicine Feature Mapping Layer (TCM Feature Mapping Layer) is a specially designed deep neural network layer whose output dimension K corresponds to various dimensions of the TCM feature space. These dimensions are formed during training through a combination of supervised learning (TCM practitioners' "TCM state quantification" annotation of a large number of cases) and unsupervised learning (clustering, dimensionality reduction).

[0075] Output: V TCM =ReLU(W TCM ·E fused +b TCM ).

[0076] V TCM Dimensional example:

[0077] Temperature (hot or cold): v TCM,1

[0078] Degree of realism / illusion: v TCM,2

[0079] Degree of Qi stagnation: v TCM,3

[0080] Blood stasis level: v TCM,4

[0081] Spleen and stomach function index: v TCM,5

[0082] Liver Qi Circulation Index: v TCM,6

[0083] Meridian blockage index (refined to specific meridians): v TCM,7 ,...,v TCM,j

[0084] Vital Energy Level: v TCM,j+1

[0085] …

[0086] Each v TCM,i Meaning: It is a continuous quantitative value (e.g., 0-100 points) that reflects the patient's status in a certain dimension of traditional Chinese medicine.

[0087] Meaning of parameter symbols:

[0088] D raw The original collection of data.

[0089] E phy E pulse E tongue E subj : Embedding vectors after each modality is encoded.

[0090] Attention(·): An attention mechanism function that includes self-attention and cross-attention.

[0091] W TCM ,b TCM The weight matrix and bias vector of the TCM feature mapping layer.

[0092] K: The number of dimensions in the TCM feature space.

[0093] v TCM,i The i-th dimension value of the TCM feature vector VTCM represents the degree of quantification of a certain TCM concept.

[0094] 3. Core Module of the "Therapy-Symptom-Effect" Dynamic Quantitative Correlation Model (TTSEDQM)

[0095] The aim is to establish a quantitative, dynamic, and interpretable correlation between physical therapy interventions, the evolution of TCM syndromes, and clinical effects. TTSEDQM is a complex deep learning network with built-in reinforcement learning (RL) and causal inference mechanisms.

[0096] Module Objectives:

[0097] Real-time identification of syndrome status before and after physiotherapy (S t ).

[0098] Quantitative assessment of specific physical therapy interventions (A t The direct impact on the syndrome state (ΔS) t (A t )).

[0099] How do changes in symptom state lead to clinical effects (ΔE)? t (S t →S t+1 )).

[0100] It provides a quantitative explanation of the effects of physical therapy interventions on the deep TCM mechanisms of the body.

[0101] Model structure:

[0102] ① Pre-Syndrome Quantization Network (PSQN):

[0103] Input: Traditional Chinese medicine feature vector V before physiotherapy TCM,pre .

[0104] Network type: Transformer-based multi-head attention network.

[0105] Output: Initial syndrome state vector S pre ∈[0,1] L , where L is the preset number of core syndromes (such as Qi deficiency, blood stasis, liver stagnation, spleen dampness, etc.), and each dimension represents the degree of quantification of the syndrome (probability or score).

[0106] Parameter: Number of multi-head attention layers (N) head_psqn Hidden layer dimension (D) hidden_psqn ).

[0107] Training: Supervised learning, characterized by "quantitative judgment of syndrome types" by senior TCM doctors on a large number of cases.

[0108] ② Treatment Intervention Effect Network (TIEN):

[0109] Input: Current physical therapy intervention A t (e.g., acupuncture at Zusanli, moxibustion at Guanyuan, etc., represented by unique heat encoding or embedded vectors), the syndrome state S before physiotherapy. pre .

[0110] Network type: Graph Neural Network (GNN) combined with Multilayer Perceptron (MLP). GNN is used to process complex correlation maps between acupoints and manipulation techniques.

[0111] Function: Quantify current physical therapy interventions A t For each symptom dimension S j The direct, quantitative impact, that is, the generation of a symptom impact vector ΔS(A t )∈R L For example, the effect of acupuncture at Zusanli (ST36) on "spleen deficiency" syndrome is -0.3 (indicating improvement), and the effect on "liver stagnation" is +0.1 (indicating slight aggravation or no effect).

[0112] Parameter: Number of GNN layers (N) layer_gnn ), node embedding dimension (D) node_embed ).

[0113] Training: Combining extensive clinical data with counterfactual analysis. For example, using matched patient (control group) data, estimate the changes in symptoms without intervention, then compare with the intervention group to learn the causal effects of physical therapy.

[0114] Causal inference mechanism: In TIEN, causal discovery algorithms (such as Granger Causality, DoWhy framework) are introduced to clarify the causal relationship between physical therapy intervention and symptom changes, rather than just correlation.

[0115] ③Syndrome Dynamic Evolution Network (SDEN):

[0116] Input: Initial syndrome state S pre The effect vector of physical therapy intervention ΔS(A) t ), and the history of physical therapy intervention sequences over a period of time H A ={A0,...,A t}

[0117] Network type: Recurrent Neural Network (RNN) with attention mechanism or Transformer.

[0118] Function: Predicts real-time symptom status after (or during) physical therapy. post .

[0119] S post =SDEN(S pre ,ΔS(A t ),H A ).

[0120] Parameters: RNN hidden layer dimension (D) hidden_rnn ), time step (T) step ).

[0121] Training: Supervised learning, using patient sequence data and TCM physicians' dynamic syndrome assessment as labels.

[0122] ④ Clinical Effect Quantization Network (CEQN):

[0123] Input: Symptom status after physiotherapy (S) post And the TCM feature vector V after physiotherapy TCM,post .

[0124] Network type: Multilayer Perceptron (MLP).

[0125] Function: Transform changes in symptom state into specific quantitative indicators of clinical effect, such as the degree of pain relief, improvement of fatigue, and improvement of sleep quality.

[0126] Output: Clinical effect vector E clinical ∈[0,1] P , where P is the number of clinical effect indicators.

[0127] E clinical =CEQN(S post V TCM,post ).

[0128] Parameter: Number of MLP layers (N) layer_mlp ), activation function.

[0129] Training: Supervised learning, using patient subjective feelings scales, changes in physiological indicators, and physician assessments as labels.

[0130] ⑤ Core Integration and Feedback Mechanism (Reinforcement Learning for Causal Chain Explanation):

[0131] The core of TTSEDQM is to use a reinforcement learning (RL) framework to optimize and understand the entire causal chain of "physiotherapy-symptom-effect".

[0132] Agent: The decision-maker (system) for physical therapy interventions.

[0133] Environment: The patient's dynamic TCM state (as determined by V) TCM and S characterization).

[0134] State: The current patient's TCM feature vector V TCM And the syndrome state S.

[0135] Action: Specific physical therapy intervention (A), such as needling a certain acupoint, massaging a certain area, adjusting the intensity of the technique, etc.

[0136] Reward:

[0137] Instant rewards (r immediate ): After physical therapy intervention, the TCM feature vector V TCM And the expected improvement in symptom state S. This requires accurate prediction of TIEN and SDEN.

[0138] Delayed reward (r) delayed): Final clinical effect E clinical And improvements to DEI.

[0139] Training objective: To learn a policy π(A|S) that enables the selection of the optimal physical therapy intervention A to maximize long-term rewards.

[0140] Interpretability: The Value Function or Q-value of RL can quantify "the expected contribution of a certain physical therapy intervention to future efficacy in the current symptom state", thus providing a deeper mechanistic explanation.

[0141] Parameters: RL algorithm (such as Deep Q-Network (DQN) or Proximal Policy Optimization (PPO)), learning rate (α) RL ), discount factor (γ) RL ).

[0142] Meaning of parameter symbols:

[0143] V TCM,pre V TCM,post Traditional Chinese medicine feature vectors before and after physiotherapy.

[0144] S pre ,S post : Syndrome state vector before and after physiotherapy, dimension L.

[0145] A t : Physiotherapy intervention at time t.

[0146] ΔS(A t ): Physiotherapy Intervention A t The influence vector on the syndrome.

[0147] H A Historical physiotherapy intervention sequence.

[0148] E clinical Clinical effect vector, dimension P.

[0149] PSQN(·), TIEN(·), SDEN(·), CEQN(·): Functions corresponding to each subnetwork.

[0150] π(A|S): RL policy function.

[0151] r immediate ,r delayed Instant and delayed rewards.

[0152] α RL ,γ RL :RL algorithm parameters.

[0153] 4. The TCM-based Dynamic Efficacy Index (DEI) generation module: DEI is a comprehensive, dynamic efficacy indicator with TCM characteristics.

[0154] Module objective: Based on the output of TTSEDQM, generate a efficacy index of 0-100 points in real time.

[0155] DEI calculation formula:

[0156] DEI t =w1·E clinical,t overall +w2·(∑ j＝1 L max(0,S pre,j -S post,j )·β j )+w3·KL_Divergence(V TCM,t V TCM,target )·γ t

[0157] in:

[0158] E clinical,t overall Clinical effect vector E clinical The comprehensive weighted average (such as the normalized sum of the improvement in patients' pain, fatigue, sleep, etc.).

[0159] ∑ j＝1 L max(0,S pre,j -S post,j )·β j The total quantitative improvement of core symptoms.

[0160] max(0,S pre,j -S post,j ): Indicates the degree of improvement of the j-th syndrome (only improvement is calculated, deterioration is not included in the bonus).

[0161] β j The importance weight of the j-th syndrome in the current patient's condition (for example, for patients with lower back pain, the improvement weight of the "kidney deficiency" syndrome is higher).

[0162] KL_Divergence(V TCM,t V TCM,target Traditional Chinese Medicine Feature Vector V TCM The TCM feature vector V of the target health state TCM,targetThe Kullback-Leibler divergence between the two values. This represents the gap between the current state of health in Traditional Chinese Medicine and the ideal state of health; the smaller the gap, the better the therapeutic effect. The γ value preceding this term... t The DEI is a dynamic negative weight, meaning the smaller the difference, the higher the DEI.

[0163] w1, w2, w3: Weighting coefficients of each part in DEI, which can be adjusted according to specific diseases and clinical needs.

[0164] V TCM,target It is a predefined or learned average TCM feature vector of a "healthy person" or "target treatment state".

[0165] Dynamics: DEI is calculated in real time and changes dynamically as the physiotherapy progresses and TTSEDQM is updated, reflecting the immediate and cumulative effects of the physiotherapy.

[0166] 5. Intelligent Feedback and Intervention Recommendation Module

[0167] Module objective: Based on the mechanistic explanations of DEI and TTSEDQM, provide personalized and interpretable suggestions for optimizing physical therapy plans.

[0168] Core functions:

[0169] Traditional Chinese Medicine Mechanism Explanation Engine:

[0170] Output: Based on the TTSEDQM model, a quantitative explanation of the TCM mechanism is provided. For example: "Currently, the DEI improvement is slow, and analysis shows that the improvement of the 'spleen deficiency and dampness retention' syndrome is not significant (S... post,脾虚 (Higher than expected). The TIEN model shows that the existing acupoint combinations are insufficient to stimulate this syndrome. Specifically, the tongue coating remains thick and greasy (v TCM,舌苔厚腻 (not significantly reduced), the LF / HF ratio in HRV is high (v TCM,自主神经紊乱 (No improvement).

[0171] Quantitative attribution: It can quantify the contribution of current physical therapy interventions to each TCM characteristic dimension and syndrome dimension.

[0172] Personalized Intervention Recommendation Engine (PIRE):

[0173] Input: Current syndrome state S post , DEI, and the mechanistic explanation of TTSEDQM.

[0174] mechanism:

[0175] Combined with the "Traditional Chinese Medicine Acupoint / Intervention Knowledge Graph": This graph contains information on the efficacy, compatibility, contraindications, techniques, and intensity of stimulation of acupoints, as well as the strength of their association with syndromes and symptoms.

[0176] Using the RL strategy of TTSEDQM: Based on the current patient state S and DEI, PIRE uses the RL strategy π(A|S) to recommend the physical therapy intervention A′ that maximizes future rewards (i.e. DEI improvement).

[0177] Consider the "individual effect profile": Based on the patient's historical physiotherapy data, the system learns their individual responses to different acupoints and techniques, and makes fine adjustments. For example, if a patient has high sensitivity to electroacupuncture, a slightly lower electroacupuncture intensity is recommended.

[0178] Output:

[0179] Specific suggestions for adjusting the physiotherapy plan: such as "adding acupuncture to Zusanli, moxibustion to Pishu, adjusting the massage technique to a tonifying method, and increasing the duration by 10 minutes."

[0180] Recommended acupoint combinations: Based on the current symptoms and target DEI, the best acupoint combinations and their stimulation parameters are recommended.

[0181] Risk warning: If a continuous decline in DEI or a deterioration of a certain TCM characteristic is detected, a warning will be issued.

[0182] Meaning of parameter symbols:

[0183] S post,脾虚 : Quantitative value of spleen deficiency syndrome.

[0184] v TCM,舌苔厚腻 Quantitative value of thick, greasy tongue coating in the TCM characteristic space.

[0185] v TCM,自主神经紊乱 Quantitative values ​​of autonomic nervous system disorders in the characteristic space of traditional Chinese medicine.

[0186] A′: Recommended physical therapy intervention.

[0187] 6. User Interaction Module

[0188] Physician / Therapist Workstation: Provides an intuitive dashboard displaying DEI curves, TTSEDQM mechanistic explanations (such as radar charts of symptom changes and bar charts showing the contribution of each physical therapy intervention), personalized recommendations, and early warning information. Supports physicians in reviewing and modifying recommended plans.

[0189] Patient-side APP / display interface: Displays DEI in real time, explaining the current therapeutic effect and TCM principles in easy-to-understand language, enhancing patient participation and compliance.

[0190] Interpretable report generation: Automatically generates detailed physiotherapy reports, including data before, during and after physiotherapy, DEI trends, syndrome evolution, mechanism explanations and follow-up recommendations.

[0191] Knowledge base and data management

[0192] Traditional Chinese Medicine Acupoint / Intervention Knowledge Graph (TCM-Acupoint / Intervention Knowledge Graph):

[0193] Entities: acupoints, meridians, massage techniques, moxibustion sites, physiotherapy tools, syndromes, symptoms, diseases, and effects.

[0194] Relationships include: belonging to (is_a), located at (located_at), treating (treats), compatibility (combines_with), contraindicates (contraindicates), corresponding to (corresponds_to_syndrome), affecting (affects_syndrome_dimension), improving (improves_symptom), enhancing (enhances_effect_with), etc.

[0195] Historical physiotherapy records and individual effect records: Store all historical physiotherapy data of patients for individualized learning and PIRE recommendation optimization in the TTSEDQM model.

[0196] Traditional Chinese Medicine Theory Ontology (TCM Ontology): This theory formally defines the core concepts and relationships of Traditional Chinese Medicine, providing semantic support for MATN and TTSEDQM.

[0197] Additionally, to better illustrate the technical solution of this invention, this invention also provides a method for providing feedback on the effects of traditional Chinese medicine physiotherapy. This method is executed on a computing device including a processor and a memory, and includes the following steps:

[0198] a) Receive physiological parameter data, pulse data, tongue data, and subjective feeling data collected and preprocessed by multimodal sensors;

[0199] b) The data received in step a) is processed by a multi-layer attention mechanism transformation network MATN to generate a high-dimensional, continuous, and quantized TCM feature vector V. TCM ;

[0200] c) The TCM feature vector V generated in step b) TCM The input is processed into a dynamic quantization correlation model, TTSEDQM. This step specifically includes:

[0201] i. Generate an initial symptom state vector S using the initial symptom identification and quantization network PSQN in TTSEDQM. pre ;

[0202] ii. Using the TIEN network, a physical therapy intervention effect quantification network within the TTSEDQM, based on data A representing the current physical therapy intervention... t and the initial symptom state vector S pre Generate a syndrome influence vector ΔS(A) t );

[0203] iii. Using the SDEN network, a symptom dynamic evolution prediction network in TTSEDQM, based on the initial symptom state vector S pre and the influence vector of symptoms ΔS(A) t Generate a real-time syndrome state vector S after physiotherapy. post ;

[0204] iv. Using the Clinical Effect Quantification Network (CEQN) within the TTSEDQM, based on the real-time syndrome state vector S post Traditional Chinese Medicine Feature Vector V after Physiotherapy TCM,post Generate a clinical effect vector E clinical ;

[0205] d) Based on the pre-set model, combined with the clinical effect vector E generated in step c)iv, clinical Initial syndrome state vector S pre With the real-time symptom state vector S post The difference between the values ​​of the TCM feature vector and the KL divergence between the TCM feature vector and a target health state vector are used to calculate a dynamic therapeutic efficacy index (DEI) value.

[0206] Furthermore, the cross-modal attention fusion layer of the Multi-Layer Attention Mechanism Transformation Network (MATN) employs self-attention and cross-attention mechanisms to determine the fusion weights between data features from different modalities.

[0207] Furthermore, the Initial Symptom Recognition and Quantization Network (PSQN) is a neural network based on the Transformer architecture.

[0208] Furthermore, the Syndrome Dynamic Evolution Prediction Network (SDEN) is a recurrent neural network (RNN) or Transformer network with an attention mechanism.

[0209] Furthermore, the preset model used by the Dynamic Efficacy Index (DEI) generation module is: DEI t =w1·E clinical,t overall +w2·(∑ j＝1 Lmax(0,S pre,j -S post,j )·β j )+w3·KL_Divergence(V TCM,t V TCM,target )·γ t ;

[0210] Among them, E clinical,t overall S is the weighted average of the clinical effect vectors. pre,j and S post,j These are the quantitative values ​​of the j-th symptom before and after physiotherapy, β j V represents the importance weight of the j-th symptom. TCM,t V represents the current feature vector of Traditional Chinese Medicine. TCM,target Let w1, w2, w3, γ be the preset target health state vector, and KL_Divergence be the Kullback-Leibler divergence between the two. t These are the preset weighting coefficients.

[0211] Furthermore, the following steps are also included:

[0212] a) Receive the policy output by the reinforcement learning component in the Dynamic Quantization Relationship Model (TTSEDQM);

[0213] b) Based on the strategy, query a pre-defined TCM acupoint / intervention knowledge graph;

[0214] c) Based on the query results, generate and output a data signal containing suggestions for optimizing the physiotherapy plan. Specific Implementation

[0216] Example 1: Feedback on the effects of combined physical therapy (massage + moxibustion) on patients with spleen and stomach deficiency and liver qi stagnation.

[0217] Pre-therapy data collection and spatial mapping of TCM features:

[0218] The patient's chief complaint was loss of appetite, abdominal distension, fatigue, low mood, and bitter taste in the mouth.

[0219] Data acquisition: A high-precision pulse sensor identified "a coexistence of wiry and soft pulses," and a tongue image acquisition device identified "a pale and swollen tongue with a white and greasy coating and teeth marks on the edges." Physiological sensors showed a low SDNN value for HRV (autonomic nervous system dysfunction), and a microcirculation monitor showed slightly poor abdominal blood perfusion.

[0220] Traditional Chinese Medicine Feature Space Mapping Module (MATN): Maps the above data into a Traditional Chinese Medicine feature vector V. TCM,pre ,in:

[0221] v TCM,脾胃功能指数40 points (low)

[0222] v TCM,肝气疏泄指数 35 points (lower)

[0223] v TCM,湿邪程度 70 points (higher)

[0224] v TCM,气虚程度 65 points (relatively high)

[0225] v TCM,情绪调畅度 20 points (very low)

[0226] TTSEDQM model (PSQN): Initial syndrome identification is S pre Spleen deficiency with dampness (0.7), liver qi stagnation (0.6), qi deficiency (0.5).

[0227] Develop a physical therapy plan:

[0228] Based on the initial assessment, the TCM doctor formulated a treatment plan: massage of Zhongwan and Zusanli (to strengthen the spleen and stomach), massage of Qimen and Taichong (to soothe the liver and relieve depression), and moxibustion of Guanyuan (to warm and replenish vital energy).

[0229] Real-time feedback and TTSEDQM dynamic assessment during physiotherapy:

[0230] Intervention A1 (massage of Zhongwan and Zusanli acupoints):

[0231] The system monitors the increase in abdominal skin temperature and improvement in microcirculation in real time. The patient reports a warm sensation in the abdomen.

[0232] TTSEDQM model (TIEN): Quantifying the effect of A1 on syndrome ΔS(A1): Spleen deficiency with dampness (-0.15), Qi deficiency (-0.1), Liver Qi stagnation (0).

[0233] TTSEDQM model (SDEN): Predicts the current syndrome state S current .

[0234] DEI generation module: Real-time calculation of DEI t The rise is slow.

[0235] Intervention A2 (massage of Qimen and Taichong points):

[0236] The patient's breathing gradually stabilized, and the LF / HF ratio of HRV decreased slightly.

[0237] TTSEDQM model (TIEN): Quantifying the effect of A2 on syndrome ΔS(A2): Liver Qi stagnation (-0.2), Spleen deficiency and dampness (0.05, slightly aggravated due to excessive stimulation).

[0238] DEI generation module: DEI tIt has improved slightly, but is still not ideal.

[0239] System intelligent feedback and intervention recommendations:

[0240] Mechanism Explanation Engine: "The current DEI score is 55. After massage, there is no significant improvement in spleen deficiency with dampness (currently 0.55) and liver qi stagnation (currently 0.4) symptoms. The TTSEDQM model shows that massage of Qimen and Taichong points has a lower-than-expected improvement on liver qi stagnation and a slight negative impact on spleen deficiency with dampness. Analysis revealed a white and greasy tongue coating (v...") TCM,湿邪程度 Still at 65 points), emotional well-being (v) TCM,情绪调畅度 Still 30 points).

[0241] PIRE recommends: "Adding moxibustion to the Qimen acupoint to enhance its liver-soothing and mood-lifting effects, and adjusting the massage technique to a gentle tonifying method to avoid excessive abdominal stimulation and thus prevent depletion of spleen qi. Simultaneously, moxibustion on the Zusanli acupoint is recommended to enhance its spleen-strengthening and dampness-removing effects. It is estimated that this adjustment can increase DEI by 15% in subsequent treatments."

[0242] Post-physiotherapy assessment and recommendations:

[0243] After the treatment plan was adjusted, the patient's bitter taste in the mouth lessened, the warm sensation in the abdomen became more pronounced, and the patient's mood improved.

[0244] TTSEDQM model (SDEN & CEQN): Predicts S post Spleen deficiency with dampness (0.3), liver qi stagnation (0.2), qi deficiency (0.4). Clinical effects: improved appetite (0.6), reduced abdominal distension (0.7), improved fatigue (0.5), improved mood (0.7).

[0245] DEI generation module: The final DEI score is 85.

[0246] Mechanism Explanation Engine: "This combined physical therapy, through soothing the liver and regulating qi, strengthening the spleen and removing dampness, and warming and tonifying the vital energy, significantly improved the patient's spleen and stomach function and liver qi circulation. In particular, the addition of moxibustion at Qimen and Zusanli points effectively targeted the dampness and liver stagnation that had not responded well in the early stages. The quantitative values ​​of the patient's spleen deficiency and dampness syndrome and liver qi stagnation syndrome decreased by more than 50%, corresponding to a thinner tongue coating, improved emotional well-being, and a significant increase in DEI."

[0247] PIRE recommends: "Consolidate the existing treatment plan, and next time you can appropriately extend the moxibustion time, and it is also recommended to combine it with spleen-strengthening dietary therapy such as yam porridge."

[0248] To verify the effectiveness of the technical solution of this invention, the following verification report is presented:

[0249] 1. Experimental Objective

[0250] This study aims to verify, through a clinical controlled trial, the superiority of the "Therapeutic Effect Feedback System Based on Dynamic Mapping of Traditional Chinese Medicine Feature Space" (hereinafter referred to as "this System") disclosed in this invention over traditional Chinese medicine physiotherapy evaluation methods. Specific verification objectives are as follows:

[0251] Objective 1 (Objective Quantification and In-Depth Interpretability Verification): Verify that the system can achieve objective quantification of the therapeutic effect and provide in-depth, interpretable feedback on the TCM mechanism, and that its evaluation results are highly consistent with expert consensus.

[0252] Objective 2 (Verification of Real-time Feedback and Personalized Optimization Effects): Verify that the real-time feedback and personalized intervention recommendations provided by this system can effectively guide physical therapists to optimize their treatment plans, thereby improving the final clinical efficacy of physical therapy.

[0253] Objective 3 (Validation of the effectiveness of the comprehensive efficacy index DEI): To verify that the system’s unique “Dynamic Efficacy Index of Traditional Chinese Medicine Physiotherapy (DEI)” can reflect the overall improvement of patients more comprehensively and sensitively than traditional single indicators (such as VAS pain score).

[0254] 2. Experimental Design and Subjects

[0255] 2.1 Experimental Design

[0256] A prospective, randomized, single-blind, controlled clinical trial design was adopted.

[0257] Forward-looking: Pre-determine the plan and collect data in advance.

[0258] Randomization: Subjects are randomly assigned to the experimental group and the control group.

[0259] Single-blind: Subjects are unaware of their assigned group, but the therapist and data evaluator are aware of the group assignments (double-blind is difficult to achieve due to different intervention methods).

[0260] Control group: The experimental group used this system for auxiliary assessment and program optimization; the control group used traditional Chinese medicine physiotherapy assessment methods.

[0261] 2.2 Subject Selection

[0262] Inclusion criteria:

[0263] Age between 30 and 60 years old, gender not limited.

[0264] The patient was diagnosed with "cervical and shoulder syndrome of spleen deficiency and dampness" by two TCM physicians at the associate chief physician level or above. The main symptoms include: soreness, heaviness, and limited movement in the neck and shoulders, accompanied by fatigue, poor appetite, and loose stools.

[0265] Neck pain visual analog scale (VAS) score ≥ 4.

[0266] Participants are encouraged to voluntarily participate in this trial and sign an informed consent form.

[0267] Exclusion criteria:

[0268] Individuals suffering from severe organic diseases such as cardiovascular and cerebrovascular diseases, liver diseases, and kidney diseases.

[0269] Those with serious lesions such as cervical spine fractures, tumors, or tuberculosis.

[0270] Pregnant or breastfeeding women.

[0271] Patients currently receiving other treatments that may affect the results of this trial.

[0272] Those with broken skin or a history of allergies are not suitable for wearing sensors or undergoing physical therapy.

[0273] Sample size: A total of 60 eligible subjects were recruited and randomly divided into an experimental group (30 people) and a control group (30 people) using a random number table.

[0274] 3. Trial grouping and intervention measures

[0275] Both groups of subjects received standard TCM physiotherapy for 2 weeks, 3 times a week, for a total of 6 sessions.

[0276] Basic physical therapy plan (same for both groups):

[0277] Massage: Apply rolling, pressing, kneading, and grasping techniques to the neck and shoulders (Fengchi, Jianjing, Tianzong, cervical Jiaji, etc.) for 15 minutes each time.

[0278] Moxibustion: Moxibustion on Zusanli, Zhongwan, and Pishu acupoints for 15 minutes each time.

[0279] Evaluation and plan adjustment methods (two different groups):

[0280] Control group (traditional method):

[0281] Assessment: The physical therapist mainly relies on the subjective experience of the four diagnostic methods of "inspection, auscultation, inquiry, and palpation" (including observing the patient's expression, asking about feelings, taking the pulse, and observing the tongue) and the patient's verbal feedback to assess the effectiveness of each physical therapy session.

[0282] Treatment plan adjustments: Based on their clinical experience, physical therapists decide after each treatment whether the techniques or duration of the next session need adjustment. However, the basis for these adjustments lacks quantifiable standards.

[0283] Experimental group (assisted by this system):

[0284] Assessment: Comprehensive data collection (multimodal physiological parameters, three-dimensional pulse diagnosis, multispectral tongue diagnosis, and structured subjective feelings) was conducted using this system before, during, and after each physiotherapy session. The system automatically generated a Traditional Chinese Medicine feature vector (V).TCM The system includes syndrome quantification (S), dynamic efficacy index (DEI), and provides a mechanism explanation report.

[0285] Treatment Plan Adjustment: During the treatment, therapists can view the real-time feedback provided by this system. If the system indicates "DEI growth is slow, and the 'dampness' symptoms are not improving well," and recommends "adding moxibustion to Fenglong acupoint to enhance the dampness-removing effect," the therapist can adopt this suggestion based on their own judgment and adjust the treatment plan in real time. Any adjustments made by the therapist to the treatment plan are recorded by the system.

[0286] 4. Efficacy evaluation indicators

[0287] 4.1 Main therapeutic indicators

[0288] Neck and Shoulder Dysfunction Index (NDI): Assessments were conducted before treatment (day 0) and after treatment (day 15). The NDI is an internationally recognized scale for assessing neck dysfunction; the lower the score, the better the function.

[0289] Visual analog scale (VAS) for pain: Assessment was performed before and after each treatment. 0 points indicated no pain, and 10 points indicated severe pain.

[0290] 4.2 Secondary efficacy indicators

[0291] The Spleen Deficiency and Dampness Syndrome Scoring Scale, developed by an expert team, includes 10 items such as fatigue, poor appetite, heaviness in the neck and shoulders, and thick, greasy tongue coating. Each item is scored from 0 to 3, for a total of 30 points. Assessments are conducted before and after treatment.

[0292] System evaluation metrics (experimental group only):

[0293] Traditional Chinese Medicine Feature Vector (V) TCM Key dimensions in the index, such as "degree of dampness" and "spleen and stomach function index".

[0294] Syndrome quantification value (S), such as the syndrome score for "spleen deficiency and dampness".

[0295] Dynamic efficacy index (DEI).

[0296] Consistency of expert blind review: Data before and after physiotherapy of 10 patients in the experimental group (excluding the results of the systematic evaluation) were randomly selected and three senior TCM doctors who did not participate in this trial conducted a "back-to-back" blind review (evaluating the efficacy level: significant effect, effective, ineffective). The results were then compared with the efficacy level output by this system (Kappa test).

[0297] 5. Test Implementation Steps

[0298] Baseline assessment (Day 0): Demographic information registration, NDI questionnaire, VAS score, and spleen deficiency and dampness syndrome score scale assessment were performed on all 60 subjects, and all system data before physiotherapy were collected.

[0299] Randomization: Subjects are randomly assigned to either the experimental group or the control group.

[0300] Intervention implementation (days 1 to 14): Both groups received 6 sessions of physical therapy as planned.

[0301] Control group: Therapists performed and assessed patients using traditional methods.

[0302] Experimental group: This system was used for monitoring and assistance throughout the entire process. The system recorded all real-time feedback, recommendations, and the therapists' adoption of these recommendations.

[0303] Data collection: VAS scores were recorded before and after each treatment. The experimental group received a full set of systemic data for each treatment.

[0304] Endpoint assessment (day 15): All subjects were assessed again using the NDI questionnaire, VAS score, and Spleen Deficiency and Dampness Syndrome Score Scale.

[0305] Blind review by experts: After the trial, the data of the 10 patients in the experimental group were compiled for blind review by experts.

[0306] Data preparation and statistical analysis: SPSS 25.0 software was used for data analysis. t-tests were used for continuous data, and chi-square tests were used for categorical data. 2 The test or Kappa test can be used. A p-value < 0.05 is considered statistically significant.

[0307] 6. Experimental Results and Data Tables

[0308] 6.1 Validation of Objective 1: Objective Quantification and In-depth Interpretability

[0309] Table 1: Consistency test between the experimental group's systematic evaluation results and the expert blind review (n=10)

[0310]

[0311]

[0312] Results analysis:

[0313] The system's assessment of efficacy levels is highly consistent with the assessment results reached by three senior experts (90%).

[0314] The Kappa value was 0.85 (Kappa > 0.75 is generally considered to indicate excellent consistency), which is statistically significant.

[0315] Conclusion: This demonstrates that the evaluation results of this system are highly objective and accurate, effectively simulating and even quantifying the diagnostic thinking of experts, thus verifying objective 1.

[0316] Table 2: Examples of T003 Mechanism Explanation in the Experimental Group Patients

[0317]

[0318] Results Analysis: Table 2 visually illustrates how the system uses quantitative indicators (V) TCM The system identifies problems and provides interpretable and actionable optimization suggestions based on Traditional Chinese Medicine (TCM) theory. This demonstrates the system's deep interpretability and real-time guidance capabilities.

[0319] 6.2 Validation of Objective 2: Improvement of Clinical Efficacy through Personalized Optimization

[0320] Table 3: Comparison of key efficacy indicators before and after treatment in the two groups (mean ± standard deviation)

[0321]

[0322] Results analysis:

[0323] Before treatment, there were no statistically significant differences between the two groups in NDI, VAS and syndrome scores (P>0.05), and the baselines were comparable.

[0324] After treatment, all indicators in both groups improved (P<0.01).

[0325] Key findings: The experimental group showed significantly greater improvement in NDI scores, VAS scores, and spleen deficiency with dampness syndrome scores compared to the control group (all intergroup differences P < 0.01). The experimental group's NDI score decreased by 16.4 points, while the control group only decreased by 9.3 points.

[0326] Conclusion: This study demonstrates that, with the guidance of real-time feedback and personalized optimization suggestions from this system, physical therapists can adjust their treatment plans more precisely, thereby achieving superior clinical efficacy and validating objective 2.

[0327] 6.3 Target 3 Validation: Validity of DEI

[0328] Table 4: Comparison of DEI and VAS scores during a single physiotherapy session (T007) in the experimental group.

[0329]

[0330] Results analysis:

[0331] In this case, the changes in the VAS score were relatively slow and limited. During the moxibustion phase, although the patient's pain did not change immediately, the system detected a significant improvement in their internal physiological state (TCM characteristics related to spleen and stomach function), which was sensitively captured by the DEI index.

[0332] DEI not only reflects the single dimension of pain, but also integrates symptom improvement (improved spleen and stomach function), physiological indicator improvement (skin temperature, microcirculation) and subjective feelings (improved mental state), which can more comprehensively and dynamically assess the deep and overall effects of physical therapy.

[0333] Conclusion: This study demonstrates that DEI is a more comprehensive, sensitive, and effective indicator of overall efficacy in TCM compared to traditional single indicators, thus validating objective 3.

[0334] 7. Experimental Conclusions

[0335] The results of this clinical controlled trial strongly demonstrate that the "physiotherapy effect feedback system based on dynamic mapping of traditional Chinese medicine feature space" disclosed in this invention has significant and excellent technical effects:

[0336] This system can achieve objective and quantitative evaluation of the effects of traditional Chinese medicine physiotherapy. Its evaluation results are highly consistent with expert consensus and can provide in-depth and interpretable feedback on the mechanisms of traditional Chinese medicine.

[0337] The real-time feedback and personalized intervention recommendations provided by this system can effectively guide physical therapists to optimize treatment plans, thereby significantly improving clinical efficacy and outperforming traditional assessment methods.

[0338] The DEI index, a unique feature of this system, is a more comprehensive and sensitive tool for evaluating therapeutic efficacy than traditional single indicators, and can better reflect the overall regulatory effect of traditional Chinese medicine physiotherapy.

[0339] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A traditional Chinese medicine physiotherapy effect feedback system, characterized in that, Comprise the following components: a) a data processing device comprising at least one processor and a memory, the memory storing computer program instructions, the processor executing the computer program instructions to implement the functions of the following modules: b) a traditional Chinese medicine feature space mapping module, the module comprising a multi-layer attention mechanism conversion network MATN; the MATN is configured to: receive preprocessed physiological parameter data, pulse condition data, tongue condition data and subjective feeling data, and process the received data through a modal encoder, a cross-modal attention fusion layer and a traditional Chinese medicine feature mapping layer in the MATN to generate a high-dimensional, continuous and quantitative traditional Chinese medicine feature vector V TCM ; c) a dynamic quantification and association model TTSEDQM, which is composed of a series of interconnected neural networks stored in the memory, the neural networks comprising: i. an initial syndrome identification and quantification network PSQN, an input end of which is connected to an output end of the TCM feature space mapping module, used for receiving the TCM feature vector V TCM,pre and outputting an initial syndrome state vector S pre ; ii. a therapeutic intervention effect quantification network TIEN, comprising a graph neural network GNN and a causal inference component, the input of which is configured to receive data A representative of a current therapeutic intervention t and an initial syndrome state vector S pre and output a syndrome influence vector ΔS(A t ) quantifying the influence of the therapeutic intervention on the syndrome. iii.a syndrome dynamic evolution prediction network SDEN, an input end of which is configured to receive an initial syndrome state vector S pre , a syndrome influence vector ΔS(A t ) and historical physiotherapy intervention data, and output a real-time syndrome state vector S post after physiotherapy; iv. a clinical effect quantification network CEQN, whose input is configured to receive the real-time syndrome state vector S post and the TCM feature vector V after physiotherapy TCM,post and output a clinical effect vector E quantifying the clinical effect clinical ; d) a dynamic efficacy index DEI generation module configured to calculate and generate a dynamic efficacy index DEI value according to a preset model, in combination of a clinical effect vector E clinical , a difference between an initial syndrome state vector S pre and a real-time syndrome state vector S post , and a KL divergence between the TCM feature vector and a target health state vector.

2. The traditional Chinese medicine treatment effect feedback system according to claim 1, characterized in that, Further comprising: a multi-modal data acquisition device, specifically comprising: a three-dimensional multi-point piezoelectric array pulse sensor for acquiring a three-dimensional pulse waveform array, and a multi-spectral intelligent pulse acquisition device for acquiring multi-spectral images and constructing a three-dimensional tongue model.

3. The traditional Chinese medicine treatment effect feedback system according to claim 2, characterized in that: Further comprising an intelligent feedback and intervention recommendation module; Specifically comprising a recommendation engine; the recommendation engine is configured to: a) receive the policy output by the reinforcement learning component in the dynamic quantification and association model TTSEDQM; b) based on the policy, query a TCM acupoint / intervention knowledge graph stored in the memory; c) based on the query result, generate a data signal containing a physiotherapy scheme optimization suggestion.

4. The traditional Chinese medicine treatment effect feedback system according to claim 3, characterized in that: The TCM acupoint / intervention knowledge graph stored in the memory is a data structure containing entities such as acupoints, meridians, syndromes, and functions, as well as their associated relationships.

5. A traditional Chinese medicine physiotherapy effect feedback method, characterized in that, The method is executed on a computing device comprising a processor and a memory, comprising the following steps: a) receiving physiological parameter data, pulse data, tongue data, and subjective experience data collected and preprocessed by multi-modal sensors; b) processing the data received in step a) through a multi-layer attention mechanism transformation network (MATN) to generate a high-dimensional, continuous, quantized traditional Chinese medicine feature vector V TCM ; c) generating a traditional Chinese medicine feature vector V TCM is input into a dynamic quantitative correlation model TTSEDQM for processing, which specifically includes: i. An initial syndrome state vector S is generated by an initial syndrome identification and quantification network PSQN in the TTSE DM pre ; ii. Through a physiotherapy intervention effect quantification network TIEN in TTSEDQM, according to data A representing the current physiotherapy intervention t and the initial syndrome state vector S pre , generate a syndrome influence vector ΔS(A t ). iii. Through one of the syndrome dynamic evolution prediction networks SDEN in TTSEDQM, according to the initial syndrome state vector S pre and the syndrome influence vector ΔS(A t ), generate a real-time syndrome state vector S post after physiotherapy; iv. Using the Clinical Effect Quantification Network (CEQN) within the TTSEDQM, based on the real-time syndrome state vector S post Traditional Chinese Medicine Feature Vector V after Physiotherapy TCM,post Generate a clinical effect vector E clinical ; d) According to the preset model, the clinical effect vector E generated in step c) iv is combined clinical , the initial syndrome state vector S pre and the difference between the real-time syndrome state vector S post and the KL divergence of the traditional Chinese medicine feature vector and a target health state vector, a dynamic therapeutic effect index DEI value is calculated.

6. The TCM therapy effect feedback method of claim 5, wherein: The cross-modal attention fusion layer of the multi-layer attention mechanism conversion network MATN adopts a self-attention mechanism and a cross-attention mechanism to determine the fusion weights between different modal data features.

7. The Chinese medicine treatment effect feedback method according to claim 6, characterized in that: The initial syndrome identification and quantification network PSQN is a neural network based on the Transformer architecture.

8. The Chinese medicine treatment effect feedback method according to claim 7, characterized in that: The syndrome dynamic evolution prediction network SDEN is a recurrent neural network RNN or Transformer network with attention mechanism.

9. The traditional Chinese medicine treatment effect feedback system according to claim 8, characterized in that, The preset model for the dynamic efficacy index DEI generation module is: DEI t = w1 · E clinical,t overall + w2 · (∑ j＝1 L max(0, S pre,j -S post,j ) · β j ) + w3 · KL_Divergence(V TCM,t , V TCM,target ) · γ t ; wherein, E clinical,t overall is the weighted average of the clinical effect vector, S pre,j and S post,j are the quantification values of the jth syndrome before and after physiotherapy, β j is the importance weight of the jth syndrome, V TCM,t is the current TCM feature vector, V TCM,target is the preset target health state vector, KL_Divergence is the Kullback-Leibler divergence between the two, w1, w2, w3, γ t are preset weight coefficients.

10. The Chinese medicine treatment effect feedback method according to claim 9, characterized in that, Further comprising the following steps: a) receiving the policy output by the reinforcement learning component in the dynamic quantification and association model TTSEDQM; b) based on the policy, query a pre-set TCM acupoint / intervention knowledge graph; c) according to the query result, generate and output a data signal containing a physiotherapy scheme optimization suggestion.

Images