Chronic disease acupuncture conditioning multi-modal fusion adaptive decision support system

Abstract

The application discloses a chronic disease acupuncture conditioning multi-modal fusion adaptive decision support system, which is based on traditional Chinese medicine theory and modern algorithm, constructs a whole cycle multi-modal data fusion architecture of "preoperative baseline-intraoperative real-time-postoperative interval-treatment iteration", accurately quantifies acupuncture long-term lag effect and independent intervention effect through time series mutual information, causal forest and other algorithms, builds a treatment level adaptive decision engine to realize functions such as syndrome type drift adaptation and acupoint tolerance correction, combines operation quality control closed loop and multidisciplinary collaborative optimization, solves the core pain points of the prior art, improves the accuracy, safety and accessibility of acupuncture conditioning, provides a standardized and digital solution for non-drug intervention of chronic diseases, and is suitable for whole cycle acupuncture conditioning management of various chronic diseases.

Description

Technical Field

[0001] This invention belongs to the fields of intelligent diagnosis and treatment of traditional Chinese medicine, digital health management of chronic diseases, and artificial intelligence-assisted acupuncture decision-making, specifically a multimodal fusion adaptive decision support system for acupuncture treatment of chronic diseases. Background Technology

[0002] Currently, my country's chronic disease management model focuses on long-term drug therapy and lifestyle interventions. However, it suffers from several pain points, including poor long-term medication adherence, cumulative adverse drug reactions, irreversible target organ damage, and insufficient primary healthcare service capacity. Acupuncture, as a core non-drug therapy in Traditional Chinese Medicine, has clear evidence-based medical support in chronic disease management. It can effectively improve core physiological indicators, alleviate clinical symptoms, reduce the risk of complications, and improve patients' quality of life. However, its clinical application highly depends on the personal experience of acupuncturists, presenting challenges such as difficulty in managing long-term treatment courses, poor replicability of treatment plans, difficulty in controlling efficacy fluctuations, and insufficient accessibility at the primary healthcare level. These factors make it difficult to meet the core clinical needs of long-term, gradual, dynamic, and personalized chronic disease management.

[0003] With the development of digital healthcare and artificial intelligence technologies, digital acupuncture systems have become a research hotspot in the field of intelligent equipment for traditional Chinese medicine, and related technical solutions are gradually being translated into clinical applications. Among them, the core authorized patent in the current field of digital acupuncture, with authorization announcement number CN116313029B and patent number ZL202211653854.0, entitled "A Method, System, and Device for Dynamic Control Optimization of Digital Acupuncture," discloses the following technical solution: It dynamically identifies acupoint locations using machine vision, collects real-time data on patient physiological signs, acupoint physiological data, environmental factors, and acupuncture execution parameters during acupuncture, comprehensively assesses patient physiological safety, emotional stress levels, and immediate effects of acupuncture, and generates a real-time optimization strategy for a single acupuncture treatment, achieving closed-loop dynamic control in acupuncture. This solution addresses the technical issues of insufficient precision in acupuncture procedures and the inability to optimize immediate effects in real time. However, it has fundamental technical limitations: the solution is designed only for the intraoperative phase of a single acupuncture treatment and does not take into account the long-term nature of acupuncture treatment for chronic diseases. It does not address core needs of chronic disease management such as postoperative interval data collection between two acupuncture treatments, quantification of the lag in acupuncture effects, correction of acupoint stimulation tolerance during long treatments, and dynamic drift tracking of patients' TCM syndrome types. Therefore, it cannot achieve adaptive optimization of acupuncture plans at the treatment course level and is completely unsuitable for the clinical scenario of full-cycle management of chronic diseases.

[0004] Overall, current digital acupuncture and chronic disease acupuncture management technologies all suffer from the following long-standing core technical pain points: First, the data link is limited to a single treatment session and has not built a full-cycle time-series data fusion architecture. It completely ignores the core characteristic of the lag effect of acupuncture and cannot quantify the long-term effect of a single acupuncture intervention over a long period, resulting in a lack of core time-series data support for the optimization of acupuncture plans. Second, it is impossible to isolate the interference of multiple factors. The physiological indicators of patients with chronic diseases are affected by multiple factors such as acupuncture, medication, and lifestyle. Existing technology cannot quantify the independent effect contribution of acupuncture, which can easily lead to fundamental deviations in the optimization of the treatment plan. Third, the decision-making logic is static, which cannot adapt to the core characteristics of dynamic drift of patient syndrome and decay of acupoint stimulation tolerance in the long-term management of chronic diseases, and cannot achieve course-level adaptive iterative optimization of acupuncture plans. Fourth, the lack of an effect bias attribution mechanism makes it impossible to distinguish whether poor effect is due to non-standard acupuncture operation or unsuitable protocol, which can easily lead to blind adjustment of the protocol. At the same time, no protocol error tolerance adaptation mechanism has been set up for patients with different operation abilities. Fifth, it has not been integrated into the multidisciplinary comprehensive management system for chronic diseases, and the coverage of scenarios is insufficient, making it impossible to achieve synergistic optimization between acupuncture intervention and routine management of chronic diseases, and making it difficult to maximize the effect of acupuncture in the whole cycle management of chronic diseases. Summary of the Invention

[0005] The purpose of this invention is to provide a multimodal fusion adaptive decision support system for acupuncture treatment of chronic diseases. By constructing a full-cycle multimodal data fusion architecture covering pre-operative, intra-operative, inter-operative, and treatment courses, a treatment course-level adaptive acupuncture decision engine, an operational quality control closed loop, and a multidisciplinary collaborative optimization system, this invention thoroughly solves the core pain points of existing digital acupuncture technologies, such as the inability to adapt to the long-term management needs of chronic diseases, the difficulty in quantifying acupuncture effects, the attenuation of efficacy over long treatment courses, blind adjustments to treatment plans, and insufficient scenario coverage. This invention achieves deep integration of acupuncture treatment with the full-cycle management of chronic diseases, significantly improving the accuracy, safety, accessibility, and clinical efficacy of acupuncture treatment for chronic diseases. It provides a standardized, replicable, and implementable digital solution for non-drug interventions for chronic diseases, possessing extremely high technological innovation value, clinical application value, and industry promotion value.

[0006] The technical solution adopted in this invention is as follows: A multimodal fusion adaptive decision support system for acupuncture treatment of chronic diseases includes a multimodal data acquisition module, a full-cycle time-series data fusion and acupuncture effect quantification module, a treatment course-level adaptive acupuncture decision engine, and an output and interaction module, which are connected in sequence via communication. The multimodal data acquisition module is used to collect multi-source data of the entire acupuncture treatment cycle of the target chronic disease patient, including at least preoperative baseline data, real-time data during acupuncture, postoperative interval data between two acupuncture treatments, time-series follow-up data of continuous treatment courses, chronic disease medication data and lifestyle data. The full-cycle time-series data fusion and acupuncture effect quantification module is used to perform time alignment and feature fusion on full-cycle time-series multimodal data, quantify the long-term lag effect of a single acupuncture intervention through a time-series correlation algorithm, and separate the independent contributions of acupuncture intervention, medication adjustment, and lifestyle changes to changes in chronic disease indicators through a causal inference algorithm, and output the independent effect quantification results of acupuncture intervention. The treatment course-level adaptive acupuncture decision engine is used to generate acupuncture treatment plans that are adapted to the disease stage, TCM syndrome type, and intervention tolerance of the target chronic disease patients based on full-cycle fusion data and acupuncture independent effect quantification results, and to achieve closed-loop iterative optimization of the plan based on the effect data of continuous treatment courses. The output and interaction module is used to output the generated acupuncture treatment plan and receive user operation feedback and clinical effect data to update the system algorithm model.

[0007] Preferably, the multimodal data acquisition module includes a home-based inter-treatment data acquisition unit that connects to wearable devices and patient self-assessment terminals, used to collect core physiological indicators of chronic diseases, symptom self-assessment data, sleep, diet and activity data, and acupoint sensitization time-series data between two acupuncture treatments; the full-cycle time-series data fusion and acupuncture effect quantification module has a built-in time-series mutual information algorithm unit, used to establish a time-series correlation model of "this acupuncture treatment plan - postoperative inter-treatment multimodal data changes - preoperative baseline data for the next acupuncture treatment", and quantify the long-term lag effect coefficient of a single acupuncture intervention.

[0008] Preferably, the treatment course-level adaptive acupuncture decision engine includes an acupoint tolerance adaptive correction unit and a syndrome dynamic drift decision unit; the acupoint tolerance adaptive correction unit is used to calculate the stimulation effect attenuation coefficient of the target acupoint based on the quantitative results of the independent effects of acupuncture in continuous treatment courses, and adaptively adjust the stimulation parameters and acupoint combinations of the acupuncture plan; the syndrome dynamic drift decision unit has a built-in incremental learning TCM syndrome differentiation model, which is used to dynamically update the patient's TCM syndrome differentiation results based on multimodal data in continuous treatment courses, and adaptively adjust the acupoint combinations and tonification / sedation strategies of the acupuncture plan based on the drift of the syndrome differentiation results.

[0009] Preferably, the treatment-level adaptive acupuncture decision engine includes a chronic disease rhythm-acupuncture spatiotemporal fusion unit and a symptom-symptom-urgency adaptive switching unit; the chronic disease rhythm-acupuncture spatiotemporal fusion unit incorporates the Meridian Flow and Linggui Bafa acupuncture time medicine algorithm models to analyze the patient's chronic disease pathological rhythm data, establish a matching mapping relationship between "pathological rhythm peak - acupoint opening and closing time", calculate the optimal acupuncture intervention time window, and dynamically adjust acupoint combinations; the symptom-symptom-urgency adaptive switching unit is preset with multimodal data trigger thresholds to automatically switch the symptom-symptom intervention plan when the patient's indicators exceed the warning level, and automatically switch back to the symptom-symptom treatment plan when the indicators return to a stable period.

[0010] Preferably, the treatment-level adaptive acupuncture decision engine includes an intelligent acupuncture intervention method adaptation unit and a full-scenario tiered solution adaptation unit. The intelligent acupuncture intervention method adaptation unit has a built-in acupuncture intervention method-syndrome-effect association knowledge base, which is used to automatically recommend the optimal acupuncture intervention method and match the corresponding acupoint combination and stimulation parameters based on the patient's syndrome differentiation results, disease stage, and tolerance. The acupuncture intervention methods include at least filiform needles, electroacupuncture, warm acupuncture, moxibustion, auricular acupressure, and acupoint application. The full-scenario tiered solution adaptation unit has preset three-level scenario adaptation rules of "hospital-community-home", which is used to generate acupuncture solutions for corresponding scenarios and provide corresponding operation specifications and safety warning rules based on the patient's chronic disease risk stratification and data stability.

[0011] Preferably, the system also includes an acupuncture operation standardization monitoring and correction module that is communicatively connected to the multimodal data acquisition module and the output and interaction module. This module is used to acquire acupuncture operation process image data in real time through machine vision equipment, compare it with preset standard operation specifications to generate an operation standardization score, and output visual operation correction prompts in real time.

[0012] Preferably, the system also includes an effect deviation attribution analysis module that is connected to the full-cycle time-series data fusion and acupuncture effect quantification module, the acupuncture operation standardization monitoring and correction module, and the treatment course-level adaptive acupuncture decision engine. This module is used to perform attribution analysis based on the operation standardization score and full-cycle multimodal data when the acupuncture intervention effect fails to reach the preset target: if the operation standardization score is lower than a preset threshold, it is determined to be an effect deviation caused by non-standard operation, and operation correction guidance is output; if the operation standardization score meets the target, it is determined to be a mismatch between the treatment plan and the plan optimization process of the decision engine is triggered.

[0013] Preferably, the system also includes a multidisciplinary collaborative decision-making module that is communicatively connected to the full-cycle time-series data fusion and acupuncture effect quantification module, the treatment-level adaptive acupuncture decision engine, and the output and interaction module. This module is used to establish a collaborative optimization model for acupuncture intervention, medication adjustment, and lifestyle intervention based on multimodal fusion data. It automatically optimizes the acupuncture plan to mitigate safety risks and enhance synergistic effects when adjusting patient medication. Simultaneously, it outputs collaborative suggestions for comprehensive chronic disease management based on the acupuncture effect quantification results. The system also incorporates an incremental reinforcement learning unit for continuously iteratively optimizing the system's full-process algorithm model based on clinical effect data and user feedback.

[0014] A multimodal data fusion and adaptive decision-making method for acupuncture treatment in chronic disease management includes the following steps: S1. Collect multi-source data of the entire acupuncture treatment cycle for target chronic disease patients, including at least preoperative baseline data, real-time data during acupuncture, postoperative interval data between two acupuncture treatments, time-series follow-up data of continuous treatment courses, chronic disease medication data and lifestyle data. S2. Perform time alignment and feature fusion on full-cycle time-series multimodal data, quantify the long-term lag effect of a single acupuncture intervention through a time-series correlation algorithm, and separate the independent contributions of acupuncture intervention, medication adjustment, and lifestyle changes to changes in chronic disease indicators through a causal inference algorithm, and output the quantitative results of the independent effect of acupuncture intervention. S3. Based on the full-cycle fusion data and the quantitative results of the independent effects of acupuncture, generate acupuncture treatment plans that are suitable for the disease stage, TCM syndrome type, and intervention tolerance of the target chronic disease patients, and realize closed-loop iterative optimization of the plans based on the effect data of continuous treatment courses. S4. Output the generated acupuncture treatment plan and receive user feedback and clinical effect data to update the algorithm model.

[0015] Preferably, step S3 further includes: generating optimal intervention timing and acupoint combination schemes based on the patient's chronic disease pathological rhythm matched with the acupuncture time medicine algorithm; adaptively switching between symptomatic and root-cause treatment schemes based on preset thresholds; and stepwise adaptation of acupuncture intervention methods and application scenarios based on the patient's condition. The method also includes a step of real-time monitoring and correction of acupuncture operation standardization, a step of attribution analysis of acupuncture effect deviation, a step of multidisciplinary collaborative decision-making of acupuncture, chronic disease medication and lifestyle intervention, and a step of iterative optimization of the algorithm model based on incremental reinforcement learning.

[0016] The beneficial effects of this invention are: This invention addresses the core deficiency of existing digital acupuncture technologies, which focus only on single-intraoperative optimization and cannot adapt to long-term chronic disease management. Based on traditional Chinese medicine acupuncture theory and a multidisciplinary chronic disease management system, it constructs a multimodal fusion adaptive decision support system for chronic disease acupuncture treatment, achieving a fundamental technological leap from "single-treatment optimization" to "intelligent management of the entire chronic disease cycle." The core breakthrough of this invention lies in building a full-cycle data fusion architecture encompassing "preoperative baseline - intraoperative real-time - postoperative interval - treatment iteration." Through a temporal mutual information algorithm, it achieves for the first time precise quantification of the long-term lag effect of acupuncture. Furthermore, a causal forest algorithm is used to remove confounding factors such as medication and lifestyle, accurately quantifying the independent intervention effect of acupuncture. This solves the long-standing industry pain points of difficulty in quantifying and attributing the effects of acupuncture, providing an unbiased core basis for solution optimization.

[0017] The adaptive acupuncture decision engine constructed in this invention strictly adheres to the core principles of TCM syndrome differentiation and treatment. Through an incremental learning model, it achieves real-time tracking of the dynamic drift of patient syndrome types and adaptive adjustment of treatment plans, truly implementing the TCM clinical requirement of "prescription changing with syndrome." Addressing the common acupoint tolerance problem in long-term interventions, it establishes a dynamic correction mechanism for acupoint sensitization and desensitization, effectively solving the clinical challenge of diminishing efficacy in long-term acupuncture treatments. Simultaneously, it integrates the meridian flow chronomedicine and the pathological rhythms of chronic diseases, achieving precise three-dimensional matching of "time-acupoint-disease." It also features automatic switching of symptom-condition-urgency strategies and a step-by-step treatment plan adaptation function across all scenarios, breaking the hospital-based limitations of acupuncture applications and significantly improving the accessibility of acupuncture treatment at the grassroots level and in home settings.

[0018] This invention pioneers a closed-loop quality control process encompassing "operation monitoring - effect attribution - protocol optimization." Utilizing machine vision, it enables real-time monitoring and correction of acupuncture procedures, accurately distinguishing whether poor intervention effects are primarily due to improper operation or a mismatch between the protocol and the pathogenesis. This completely resolves the common industry problem of blindly adjusting prescriptions in existing technologies, ensuring the stability of acupuncture intervention effects and operational safety across different scenarios. Simultaneously, it constructs a multidisciplinary collaborative decision-making module to achieve synergistic optimization of acupuncture with medication and lifestyle interventions, deeply integrating acupuncture into the comprehensive management system for chronic diseases. This invention combines technological innovation with clinical practicality, providing safe and effective non-pharmacological intervention options for patients with chronic diseases while also promoting the standardization and digitalization of acupuncture therapy, possessing extremely high clinical application and industry promotion value. Attached Figure Description

[0019] Figure 1 This is a diagram showing the overall module architecture of the multimodal fusion adaptive decision support system for chronic disease acupuncture treatment as described in this invention. Figure 2 This is a flowchart illustrating the complete process steps of the method described in this invention. Detailed Implementation

[0020] 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.

[0021] like Figure 1 and 2As shown, the multimodal fusion adaptive decision support system for acupuncture treatment of chronic diseases adopts a four-layer medical-grade hardware architecture: terminal layer, network layer, platform layer, and application layer. The terminal layer hardware includes: an industrial-grade computer at the hospital, a medical tablet terminal at the community, a smartphone at home, a medical-grade wearable physiological monitoring device, an industrial-grade machine vision camera, a multi-parameter physiological monitor, an electroacupuncture parameter monitor, and a high-precision temperature sensor. The network layer uses 5G / Ethernet / WiFi dual-link transmission, and the transmission protocol adopts the HL7FHIR protocol, which conforms to medical data standards, to ensure the security and compatibility of communication between modules. The platform layer uses a dedicated medical cloud server. The application layer includes a web interface for physicians, a mobile APP for patients, and a backend system for management, realizing full-process human-computer interaction and data closure.

[0022] The multimodal data acquisition module follows the full-cycle pathogenesis evolution mechanism of chronic diseases in Traditional Chinese Medicine (TCM) and the full-chain expression law of acupuncture effects: the occurrence and development of chronic diseases is a multi-factor, long-cycle, and dynamic evolution process, with its core pathogenesis exhibiting the core characteristics of "deficiency of the root and excess of the branch, and dynamic transformation"; the effect of acupuncture in treating chronic diseases is not limited to the immediate effect during the procedure, but more importantly, the long-term regulatory effect on organ function achieved through the conduction of Qi and blood in the meridians during the inter-operative period. The expression of this effect is closely related to the patient's physiological state, medication intervention, lifestyle, and acupoint sensitization status. Existing technologies only collect single-point data during the procedure, completely ignoring the inter-operative expression law of acupuncture effects and the dynamic evolution characteristics of chronic disease pathogenesis, and cannot fully capture the full-chain effect of acupuncture treatment. This module, through full-cycle multimodal data acquisition, fully covers the entire effect cycle of acupuncture treatment—"pre-operative-intra-inter-operative-treatment course"—providing a complete data foundation consistent with the TCM pathogenesis laws for subsequent effect quantification and decision-making.

[0023] This module is used to collect multi-source data from the entire acupuncture treatment cycle for target chronic disease patients. All collected data is stored in a structured format with millisecond-level timestamps, providing a basic input for subsequent data fusion and decision-making. This module consists of four parallel data acquisition units, and the specific implementation of each unit is as follows: First, the preoperative baseline data collection unit collects the following information: patient demographic information (name, gender, age, height, weight, ethnicity, contact information), chronic disease history information (diagnosis time, complications, previous treatment history, surgical history, allergy history), TCM four diagnostic methods data (tongue appearance, pulse appearance, TCM symptom quantitative score), core physiological and biochemical indicators of chronic diseases, acupoint sensitivity data, and acupuncture contraindication assessment data. Tongue appearance data is collected using a medical tongue imaging device, pulse appearance data using a pressure-type pulse imaging device, TCM symptom scores using the quantitative scoring table accompanying the "Standards for Diagnosis and Efficacy of Internal Medicine Diseases in Traditional Chinese Medicine" published by the China Association of Traditional Chinese Medicine, and acupoint sensitivity data using a pressure threshold tester. After all data is collected, structured coding is automatically completed and synchronized to the system database, performing preliminary screening for missing and outlier values. The baseline data collected in this unit is a complete quantitative expression of the patient's initial pathogenesis state and is the core basis for formulating acupuncture treatment plans, conforming to the TCM clinical principle of "differentiation of syndromes and treatment, establishing the root cause first."

[0024] Second, the intraoperative real-time data acquisition unit collects data including: the entire acupuncture operation process data acquired by machine vision, the patient's real-time physiological data during the operation, acupuncture stimulation parameter data, and treatment environment data. Specifically, acupuncture operation data is acquired in real-time via an industrial machine vision camera at a frame rate of 30fps and a resolution of 1920×1080, including acupoint location coordinates, needle insertion angle, needle insertion depth, needling technique, and needle retention time. Intraoperative physiological data is acquired via a multi-parameter physiological monitor at a sampling frequency of 100Hz, including ECG, heart rate, blood oxygen saturation, non-invasive blood pressure, skin conductance, and skin temperature. Acupuncture stimulation parameters are acquired in real-time via an electroacupuncture parameter monitor and a temperature sensor at a sampling frequency of 1Hz, including the frequency, intensity, pulse width, and waveform of electroacupuncture; the needle temperature and duration of warm acupuncture; and the skin surface temperature, distance, and duration of moxibustion. Environmental data is acquired via temperature and humidity sensors and a noise sensor, including the temperature, humidity, and noise level of the treatment environment. All intraoperative data is synchronized to the system cache in real-time and preprocessed for time alignment. The intraoperative data collected in this unit is a complete record of the acupuncture "technique" execution process. It is the core basis for evaluating the immediate effect of acupuncture and controlling the safety of operation, which is in line with the core logic of the integration of "theory-method-prescription-technique" in clinical acupuncture.

[0025] Third, the home-based data collection unit connects to medical-grade wearable devices via Bluetooth 5.0, including continuous glucose monitors (CGM), medical electronic blood pressure monitors, and medical-grade smartwatches. The collected data includes: continuous monitoring data of core physiological indicators for chronic diseases (e.g., CGM collects blood glucose data every 5 minutes, and the electronic blood pressure monitor collects blood pressure data twice daily at 8 AM and 8 PM), daily TCM symptom self-assessment data, sleep duration and quality data, dietary structure and calorie intake data, physical activity type and duration data, and weekly acupoint self-touch sensitivity score data. After data collection via the patient's mobile app, outliers are first removed using the 3σ principle in the local preprocessing unit, short-term missing values ​​are processed using linear interpolation, and missing values ​​exceeding 24 hours are processed using multiple imputation. Then, the data is synchronized to the cloud server via an encrypted transmission link. The collection period fully covers the entire interval between two acupuncture treatments, with no data gaps. This unit is one of the core designs of this invention that breaks through the prior art. The interval data it collects is the core carrier of the long-term lag effect of acupuncture. It fully captures the continuous effect of meridian qi and blood regulation between two acupuncture sessions, and solves the core defect of the prior art that ignores the expression of the acupuncture interval effect.

[0026] Fourth, the medication and lifestyle data collection unit connects to the patient's electronic medical record (EMR) and rational drug use system through the hospital information system integration platform. It automatically collects data on the types of medications used for chronic diseases, single doses, frequency of use, timing of use, and medication adjustment history. Simultaneously, through a structured input interface on the patient's mobile app, it collects lifestyle data such as diet, exercise, smoking, and alcohol consumption. All data is structured and stored chronologically, forming a precise temporal correspondence with acupuncture intervention data, providing covariate input for subsequent causal effect isolation. The confounding variable data collected in this unit is the core foundation for isolating the independent effects of acupuncture and eliminating multifactorial interference, consistent with the clinical principles of multifactorial comprehensive intervention for chronic diseases.

[0027] The full-cycle time-series data fusion and acupuncture effect quantification module follows the mechanisms of acupuncture effect lag and nonlinear expression, multimodal data information complementarity, and causal inference anti-confounding bias: ① Acupuncture stimulates acupoints to activate meridian qi and blood and regulate the deficiency and excess of the internal organs. Its improvement effect on the core pathogenesis of chronic diseases has a significant time lag and is not expressed linearly and instantly. Traditional correlation analysis cannot capture this nonlinear and lag-dependent effect association; ② Data from different modalities reflect the overall state of patients from different dimensions such as physiology, pathogenesis, and behavior, and there is significant information complementarity. It is necessary to strengthen the core pathogenesis characteristics related to acupuncture effect and suppress irrelevant noise through cross-modal attention mechanisms; ③ The changes in indicators of patients with chronic diseases are affected by multiple factors such as acupuncture, medication, and lifestyle. Traditional correlation analysis cannot distinguish the independent contributions of each factor and is easily affected by confounding bias, leading to distortion of effect quantification. It is necessary to use causal inference models to remove confounding variables and accurately quantify the true intervention effect of acupuncture.

[0028] This module communicates with the multimodal data acquisition module, receives preprocessed full-cycle multimodal data, performs time alignment, feature fusion, and effect quantification, and outputs the independent effect quantification results of acupuncture intervention, providing core input for the decision engine. The specific implementation of this module is as follows: First, the time alignment and multimodal feature fusion of full-cycle time series data are performed. Using timestamps as the sole benchmark, all multimodal data are aligned to a unified time axis. Multi-scale time windows are used to adapt to the time characteristics of different data types: high-frequency intraoperative data use 10-second time windows, daily data during intervals use 1-day time windows, and follow-up data during treatment use 1-week time windows, ensuring that data with different sampling frequencies achieve accurate time series correspondence. Feature fusion is achieved using a multimodal feature fusion network (MMFN), which consists of three layers: The first layer is the feature extraction layer, employing a three-layer one-dimensional convolutional neural network (1D-CNN) to extract deep features from different modalities. The kernel size is 3, the stride is 1, and each layer has 64 kernels. The ReLU activation function is used, outputting a 128-dimensional feature vector for each modality. The second layer is the cross-modal attention layer, using an 8-head multi-head attention mechanism to calculate the correlation weights between features from different modalities, strengthening key features related to acupuncture effects and suppressing irrelevant noise features. The third layer is the fusion output layer, which concatenates the weighted multimodal feature vectors to output a 256-dimensional unified fusion feature vector, completing the deep fusion of multimodal data. This unit solves the problems of time scale mismatch and information weight imbalance between different modalities through multi-scale time alignment and cross-modal attention fusion, fully preserving the temporal features and pathogenesis information of the entire acupuncture effect cycle.

[0029] Second, the temporal mutual information algorithm is used to establish a temporal correlation model of "current acupuncture treatment plan - postoperative interval multimodal data changes - preoperative baseline data for the next acupuncture treatment", quantifying the long-term lag effect coefficient of a single acupuncture intervention. This unit is designed based on the information theory nonlinear correlation quantification mechanism: mutual information can quantify the nonlinear correlation strength between two random variables, without being limited by the linear assumption, perfectly adapting to the nonlinear and lag correlation between acupuncture effects and pathogenesis changes, solving the core defect of traditional linear correlation analysis that cannot capture the lag effect of acupuncture.

[0030] Specifically, the acupuncture intervention protocol parameters (acupoint combinations, stimulation parameters, and intervention methods) are used as intervention variables X, and the rate of change of core chronic disease indicators and the rate of change of TCM symptom scores during the postoperative interval are used as outcome variables Y. The time-series mutual information values ​​of X and Y are calculated using a sliding time window at different lag times. The formula for calculating the time-series mutual information is as follows: ; in, Intervention variables With outcome variables The temporal mutual information value between the two is used to quantify the strength of the nonlinear temporal correlation between them, and the value range is [value range missing]. The larger the value, the more likely it is to represent and The stronger the correlation; Intervention variables represent the set of parameters for this acupuncture intervention, including but not limited to acupoint combinations, electroacupuncture stimulation parameters (frequency, intensity, pulse width, waveform), warm needling / moxibustion temperature and duration, and type of intervention. Outcome variables represent the set of outcome indicators related to the acupuncture effect in the later postoperative period, including but not limited to the rate of change of core physiological indicators of chronic diseases (fasting blood glucose, 2-hour postprandial blood glucose, systolic blood pressure, diastolic blood pressure, etc.), the rate of change of quantitative scores of TCM symptoms, and the rate of change of acupoint sensitization. Intervention variables With outcome variables The joint probability density function describes and The probability distribution of a certain set of values ​​is obtained by calculating the Gaussian kernel density estimation method. Intervention variables The marginal probability density function describes The probability distribution of taking a single set of values; : Ending variable The marginal probability density function describes The probability distribution of taking a single set of values; Natural logarithm function (in mathematical terms) (as base) , : These are the intervention variables and outcome variables The integral differential element represents the integral of a continuous variable. and Integration operations.

[0031] The kernel function bandwidth is automatically optimized using the Silverman criterion. The sliding time window length is set to 1-4 weeks with a step size of 1 day. After traversing and calculating, the lag time corresponding to the mutual information peak is obtained, which is the peak effect lag time of this acupuncture intervention. At the same time, the lag effect coefficient in the 0-1 interval is output to quantify the long-term effect intensity of a single acupuncture intervention in the postoperative interval, solving the industry pain point of the inability to quantify the lag of acupuncture effect.

[0032] Third, causal inference algorithms are used to isolate the independent contributions of acupuncture intervention, medication adjustment, and lifestyle changes to changes in chronic disease indicators. This unit is designed based on a latent outcome model and a dual robust causal inference mechanism: traditional machine learning models can only fit the correlation between variables and cannot solve the spurious associations caused by confounding variables. Causal forests estimate heterogeneous treatment effects by constructing random forests, which can effectively control confounding variables such as medication and lifestyle, accurately isolate the independent causal effects of acupuncture, and avoid the distortion of effect quantification caused by multi-factor coupling.

[0033] The specific implementation involves using whether or not the patient accepted the acupuncture intervention as a binary treatment variable T (T=1 for accepting the intervention, T=0 for not accepting it). Medication adjustments, diet, exercise, sleep, age, disease duration, and baseline levels are used as confounding covariates. The rate of change in core chronic disease indicators before and after the intervention is used as the outcome variable Y. A causal forest model is constructed with the following parameters: 1000 decision trees, maximum tree depth of 6, minimum number of splits of 10, and a subsample ratio of 0.5. Cross-validation is used to optimize the model's hyperparameters. After model fitting, the mean treatment effect (ATE) for each sample is calculated. This quantifies the independent causal effect of acupuncture intervention on changes in chronic disease indicators after excluding all confounding covariates. Simultaneously, the independent contributions of medication adjustments and lifestyle changes to indicator changes are quantified, accurately isolating multi-factor coupling interference and providing a realistic effect basis for subsequent program optimization.

[0034] Fourth, the incremental reinforcement learning unit is used to continuously iterate and optimize the entire system's algorithm model based on clinical effect data and user feedback. This unit is designed based on Markov decision processes and incremental learning mechanisms: the pathogenesis of patients with chronic diseases is continuously and dynamically changing, and the entire algorithm model needs to be continuously optimized to adapt to the individual characteristics of patients. Incremental reinforcement learning can continuously update the model based on new clinical data without retraining the entire model, avoiding catastrophic forgetting and ensuring the real-time adaptability of the model.

[0035] Specifically, the system employs a Deep Q-Network (DQN) reinforcement learning model. The core reward function is the quantified independent effect of acupuncture intervention. The state space is the patient's full-cycle multimodal fusion features, and the action space is the adjustment strategy of the acupuncture plan. The core parameters are set as follows: experience replay pool size 10,000, batch size 32, learning rate 0.001, discount factor 0.9, target network update frequency 100 steps. An ε-greedy strategy is used to balance exploration and exploitation, with an initial ε value of 1.0, a decay rate of 0.995, and a minimum value of 0.01. After each course of acupuncture intervention, the system automatically stores the current state, action, reward, and next state into the experience replay pool. Batch data is extracted for incremental training of the model, eliminating the need to retrain the entire model. This enables continuous self-optimization of the system algorithm, adapting to the personalized characteristics of different patients.

[0036] The treatment-level adaptive acupuncture decision engine follows the core principles of TCM syndrome differentiation and treatment and the clinical rules of acupuncture conditioning for chronic diseases. Its core design logic is "prescription changes with syndrome, method shifts with disease, addressing both root cause and symptoms, and synergistic effects across time and space," covering the full-dimensional clinical mechanisms of TCM acupuncture conditioning for chronic diseases: ① The dynamic change mechanism of acupoint "sensitization-desensitization": Long-term repeated stimulation of the same acupoint leads to a decrease in acupoint sensitization and a weakening of meridian conduction effects, i.e., acupoint tolerance, which is the core pathogenesis reason for the decline in the efficacy of long-term acupuncture for chronic diseases; ② The dynamic drift mechanism of syndrome types: The core pathogenesis of chronic diseases dynamically evolves with the course of the disease, interventions, and lifestyle changes, with "syndrome changing with disease and prescription changing with syndrome." "This is the core requirement of syndrome differentiation and treatment; ③ The mechanism of "correspondence between man and nature" in acupuncture time medicine: the circulation of qi and blood in the human body's meridians has a circadian rhythm, and the opening and closing state of acupoints changes dynamically with time. Intervention at the opening and closing time of acupoints corresponding to the peak of the pathological rhythm can maximize the stimulation of meridian regulation effect; ④ The mechanism of "treating the symptoms in acute cases and treating the root cause in chronic cases": intervention strategies need to be dynamically switched according to the symptoms and root cause of the patient's disease mechanism, taking into account both safety and effectiveness; ⑤ The mechanism of hierarchical diagnosis and treatment and full-scenario adaptation: patients with different risk levels have different requirements for the professionalism and safety of intervention, and it is necessary to adapt to the operation capabilities and safety requirements of the three-level scenarios of hospital-community-home.

[0037] This engine communicates with the full-cycle time-series data fusion and acupuncture effect quantification module, receiving fused feature vectors and acupuncture independent effect quantification results, generating personalized acupuncture treatment plans tailored to patients, and implementing closed-loop iterative optimization. This engine comprises six parallel decision-making units, the specific implementation methods of which are as follows: First, the acupoint tolerance adaptive correction unit is used to calculate the stimulation effect attenuation coefficient of the target acupoint based on the quantification results of the independent effects of acupuncture over consecutive treatment courses, and adaptively adjust the stimulation parameters and acupoint combinations of the acupuncture plan. This unit is designed based on the dynamic evolution mechanism of acupoint sensitization-desensitization, solving the problem of rigidity in existing technical solutions and their inability to cope with the efficacy attenuation caused by long-term acupoint tolerance.

[0038] Specifically, the implementation involves calculating the effect attenuation coefficient of each main acupoint based on the effect data from consecutive treatment courses. The calculation formula is: Attenuation coefficient = Marginal effect value of the target acupoint in the nth intervention / Marginal effect value of the acupoint in the 1st intervention. The preset effect attenuation threshold is 0.6. When the attenuation coefficient of a certain main acupoint is lower than the threshold for two consecutive times, it is determined that the acupoint has developed stimulation tolerance, that is, the acupoint has entered a dull state, and the meridian conduction effect has significantly decreased. The system automatically executes a two-level correction strategy: The first-level strategy is to adjust the stimulation parameters, increasing the electroacupuncture intensity corresponding to the acupoint by 10%-2%. The initial stimulation frequency was adjusted to twice or half of the original frequency, and the needle retention time was extended by 10 minutes. The stimulation parameters were adjusted to reactivate the acupoint sensitization state. The secondary strategy was to adjust the acupoint combination, replacing the tolerant acupoints with backup acupoints of the same meridian and indication. The backup acupoint database was constructed based on the 4th edition of the "Acupuncture and Moxibustion" textbook in the new century and the "Clinical Guidelines for Acupuncture and Moxibustion" of the China Association of Traditional Chinese Medicine. Each main acupoint corresponds to 3-5 backup acupoints with sufficient evidence-based support. For example, the backup acupoints for Zusanli are Yanglingquan, Fenglong, Sanyinjiao, and Shangjuxu. The intensity of the meridian regulation effect was restored by changing the acupoints.

[0039] Second, the syndrome dynamic drift decision unit incorporates an incremental learning TCM syndrome differentiation model. This model dynamically updates the patient's TCM syndrome differentiation results based on multimodal data from continuous treatment courses, and adaptively adjusts the acupuncture point combinations and tonification / sedation strategies based on the drift of the syndrome differentiation results. This unit is designed to address the dynamic evolution of chronic disease pathogenesis and the core principles of syndrome differentiation and treatment, overcoming the core deficiency of existing static solutions that cannot adapt to the dynamic changes in patients' syndrome patterns.

[0040] Specifically, the model is implemented using an incremental support vector machine (ISVM). The initial model is trained based on the diagnostic and therapeutic standards for chronic diseases in the "Standards for Diagnosis and Treatment of Diseases and Syndromes in Traditional Chinese Medicine" and "Internal Medicine of Traditional Chinese Medicine." The training data comes from the publicly available TCM-MIMIC dataset for chronic diseases, covering all TCM syndrome differentiation types of common chronic diseases such as hypertension and type 2 diabetes. The initial model achieves a diagnostic accuracy of ≥92% on the test set. The incremental update cycle for the model is one treatment course (2 weeks). After each collection of new diagnostic data, symptom data, and chronic disease indicator data from patients, only the newly added data is used to incrementally update the model, without retraining the entire dataset. This ensures the real-time adaptability of the model and avoids the loss of historical pathogenesis information caused by full retraining. The preset threshold for the drift of the syndrome differentiation results is 20%. When the confidence level of the syndrome differentiation results output by the model changes by more than 20%, it is determined that the patient has experienced dynamic syndrome drift, that is, the core pathogenesis has undergone essential changes. The system automatically adjusts the main acupoints, acupoint combinations, and tonification and sedation techniques of the acupuncture plan based on the new syndrome differentiation results. For example, when the patient changes from the Yin deficiency and dryness-heat type to the Qi and Yin deficiency type, the main acupoints are changed from Quchi and Neiting, which clear heat and moisten dryness, to Zusanli, Sanyinjiao, and Taixi, which tonify Qi and nourish Yin. The tonification and sedation techniques are changed from sedation to even tonification and sedation, which fully follows the TCM clinical principle of "the prescription changes with the syndrome".

[0041] Third, the Chronic Disease Rhythm-Acupuncture Spatiotemporal Fusion Unit incorporates standardized algorithm models from the Meridian Flow Method, the Nazi Method, and the Linggui Eight Methods. This unit analyzes patients' chronic disease pathological rhythm data, establishes a matching mapping relationship between "pathological rhythm peak - acupoint opening and closing time," calculates the optimal acupuncture intervention time window, and dynamically adjusts acupoint combinations. Designed specifically for the core mechanism of "harmony between man and nature" in acupuncture time medicine and the characteristics of chronic disease pathological rhythms, this unit achieves precise three-dimensional matching of "time-acupoint-disease," maximizing the therapeutic effect of acupuncture.

[0042] Specifically, the Meridian Flow Algorithm calculates the corresponding opening and closing acupoints, five shu points, and eight confluent points for each of the twelve two-hour periods of the day based on the patient's birth date, consultation time, and the Heavenly Stems and Earthly Branches calendar, thus fully restoring the standardized logic of TCM acupuncture time medicine. The pathological rhythm analysis uses the STL time series decomposition algorithm to decompose the time series data of the patient's continuous monitoring of chronic disease core indicators, separating the trend term, seasonal term, and residual term, and extracting the peak periods of the patient's pathological rhythm, such as the morning peak period of blood pressure for hypertensive patients, the postprandial blood glucose peak period for diabetic patients, and the nocturnal exacerbation period for COPD patients. The system establishes a matching mapping table of "pathological rhythm peak - acupoint opening and closing time". It calculates the optimal acupuncture intervention time window through mutual information algorithm, and prioritizes the time window corresponding to the pathological rhythm peak and the time window in which the acupoints opening and closing match the core pathogenesis of chronic disease as the recommended intervention time. At the same time, it dynamically adjusts the acupoint combination based on the optimal time window, and prioritizes the acupoints that open and close at that time as the main acupoints, and combines them with acupoints that target the core pathogenesis of chronic disease to achieve three-dimensional synergistic optimization of "time-acupoint-pathology" and stimulate the maximum regulatory effect of meridian qi and blood at the best time point.

[0043] Fourth, the adaptive switching unit for addressing both symptom and underlying causes is pre-set with multimodal data trigger thresholds. This allows for automatic switching to a symptomatic intervention plan when patient indicators exceed warning levels, and automatic switching back to a root-cause treatment plan when indicators return to a stable state. This unit is designed based on the core principle of Traditional Chinese Medicine (TCM) of "treating the symptoms in acute cases and addressing the root cause in chronic cases," balancing symptom control during acute exacerbations with pathogenesis management during stable periods, while mitigating the safety risks of inappropriate interventions during acute exacerbations.

[0044] Specifically, the trigger thresholds are based on national clinical guidelines such as the "Chinese Guidelines for the Prevention and Treatment of Hypertension (2023 Edition)" and the "Chinese Guidelines for the Prevention and Treatment of Type 2 Diabetes (2024 Edition)". Differentiated acute exacerbation warning thresholds are set for different types of chronic diseases. Taking type 2 diabetes as an example, the preset thresholds are: fasting blood glucose ≥16.7 mmol / L, or random blood glucose ≥19.4 mmol / L, or positive urine ketones, or an acute symptom score exceeding the warning line. Meeting any one of these conditions constitutes an acute exacerbation. The pathogenesis is primarily based on both symptom and exacerbation factors. The system automatically switches to a symptomatic intervention acupuncture program, with acupoint combinations primarily aimed at rapidly lowering blood sugar and improving acute symptoms. It employs strong stimulation parameters and a twisting-reducing method to quickly control acute conditions and mitigate risks. When the patient's fasting blood glucose is stable between 4.4-7.0 mmol / L, glycated hemoglobin <7.0%, and there are no acute symptoms, the patient is considered to be in a stable phase, with the underlying pathology primarily being deficiency. The system then automatically switches back to a root-cause treatment acupuncture program, with acupoint combinations primarily aimed at regulating the internal organs and improving the core pathogenesis of diabetes. It uses gentle stimulation parameters with a balanced tonifying and reducing effect to achieve long-term fundamental management of the chronic disease. All thresholds can be customized by clinicians based on the individual patient's condition, adapting to specific clinical scenarios.

[0045] Fifth, the intelligent acupuncture intervention unit, with a built-in knowledge base linking acupuncture intervention methods, syndrome types, and effects, automatically recommends the optimal acupuncture intervention method and matches corresponding acupoint combinations and stimulation parameters based on the patient's syndrome differentiation results, disease stage, and tolerance. This unit is designed according to the TCM principle of "treatment based on syndrome," as different acupuncture intervention methods have different effect characteristics and indications, requiring precise matching with the patient's syndrome type and disease stage to maximize the effect.

[0046] Specifically, the knowledge base is constructed based on the 4th edition of the "Acupuncture and Moxibustion" textbook in the new century, the clinical guidelines for acupuncture and moxibustion of the China Association of Traditional Chinese Medicine, and evidence-based medicine for acupuncture and moxibustion from the Cochrane Library. It covers the indications, contraindications, effect characteristics, and syndrome matching rules of six core acupuncture intervention methods: filiform needle, electroacupuncture, warm needle moxibustion, moxibustion, auricular acupressure, and acupoint application. For example, warm needle moxibustion and moxibustion are preferred for deficiency-cold syndrome to warm yang and dispel cold; electroacupuncture is preferred for blood stasis syndrome to dredge the meridians; auricular acupressure and acupoint application are preferred for daily conditioning during the stable period of chronic diseases to achieve long-term and gentle conditioning; and filiform needle and electroacupuncture are preferred for acute symptom period to achieve rapid effect. The intervention method recommendation adopts a multi-class XGBoost model. The model input includes the patient's syndrome differentiation results, disease stage, intervention tolerance, and application scenario. The output is a ranking of 6 intervention methods. The model parameters are set as follows: 100 trees, maximum depth 5, learning rate 0.1, and 5-fold cross-validation is used to optimize hyperparameters. The model's recommendation accuracy on the test set is ≥90%. The system automatically matches the corresponding standardized acupoint combination and stimulation parameters based on the recommended optimal intervention method.

[0047] Sixth, the full-scenario tiered solution adaptation unit has pre-set solution adaptation rules for three-level scenarios: "hospital-community-home." This is used to generate corresponding acupuncture solutions based on patient chronic disease risk stratification and data stability, along with corresponding operational guidelines and safety warning rules. This unit is designed based on the mechanism of the national chronic disease hierarchical diagnosis and treatment system, adapting to different scenarios' medical resources, operational capabilities, and safety requirements to achieve full-cycle and broad coverage of acupuncture treatment.

[0048] Specifically, based on the chronic disease risk stratification standards of the "National Basic Public Health Service Standards (Third Edition)," patients are divided into four levels: extremely high risk, high risk, intermediate risk, and low risk. Corresponding to three levels of scenario adaptation rules: For extremely high-risk / acute exacerbation patients, a professional hospital acupuncture plan is generated, including complex acupoint combinations, professional needling techniques, and strict intraoperative safety monitoring rules. This plan is only to be executed by licensed acupuncturists to ensure intervention safety in extreme risk scenarios. For intermediate-high risk / stable patients, a community-assisted acupuncture plan is generated, simplifying acupoint combinations and providing standardized operation guidance, safety operation specifications, and contraindication reminders. This plan is suitable for general practitioners and TCM physicians in community health service centers, adapting to the service capabilities of primary healthcare. For low-risk / recovery patients, a home-based self-intervention plan is generated, prioritizing safe, easy-to-operate, and error-tolerant intervention methods such as moxibustion, auricular acupressure, and acupoint application. This plan includes a visual acupoint location map, step-by-step operation videos, safety warning rules, and emergency response plans, suitable for patients to operate at home, achieving long-term home management of chronic diseases.

[0049] The acupuncture operation standardization monitoring and correction module and the effect deviation attribution analysis module follow the integrated mechanism of "theory-method-formula-technique" in acupuncture clinical practice: the complete expression of acupuncture effects depends on two core aspects: "whether the plan is appropriate (theory, method, and formula)" and "whether the operation is standardized (technique)". Incorrect operation can lead to inaccurate acupoint location, improper tonification and sedation techniques, and insufficient stimulation, failing to effectively stimulate the meridians and qi and blood, resulting in insufficient expression of the effect. Existing technologies attribute all poor effects to unsuitable plans, completely ignoring the core influence of the operation aspect, violating the integrated logic of acupuncture clinical practice, and easily leading to blind adjustments to plans and a continuous decline in efficacy. This module, through operation standardization monitoring and effect deviation attribution, achieves precise separation of "operation deviation correction" and "plan optimization", ensuring quality control throughout the entire acupuncture treatment process.

[0050] The acupuncture operation standardization monitoring and correction module communicates with the multimodal data acquisition module, output and interaction module, and the effect bias attribution analysis module communicates with the full-cycle time series data fusion and acupuncture effect quantification module, the acupuncture operation standardization monitoring and correction module, and the treatment course-level adaptive acupuncture decision engine. The two modules form a complete closed loop of "operation monitoring-correction-attribution-plan optimization".

[0051] The acupuncture operation standardization monitoring and correction module is used to acquire acupuncture operation process image data in real time through machine vision equipment, compare it with preset standard operation specifications to generate operation standardization scores, and output visual operation correction prompts in real time. Specifically, it employs a YOLOv8-based object detection model. The pre-trained model is trained on a publicly available acupuncture operation dataset and 10,000 clinically collected acupuncture operation annotated images. The annotations include standard acupoint locations, needle insertion angles, needle insertion depths, key points of needling techniques, and standard action specifications. The model achieves a keypoint detection accuracy of ≥95% and a recall rate of ≥93% on the test set. For hospital / community scenarios, the model is deployed on an industrial computer, inputting real-time video streams captured by an industrial camera. For home scenarios, a lightweight YOLOv8-nano model is used, deployed on a patient-side mobile app, adapted to mobile computing power, ensuring a real-time detection frame rate of no less than 20fps. The model detects key points of acupuncture procedures in real time and compares them with the preset national standard "Acupuncture Technique Operation Specifications: Filiform Needles" (GB / T21709.1-2008). It calculates an operation standardization score from 0 to 100 points, with the following scoring rules: acupoint location deviation exceeding 1cm deducts 20 points; needle insertion angle deviation exceeding 15 degrees deducts 10 points; needle insertion depth deviation exceeding 20% ​​deducts 10 points; non-standard needling techniques deduct 15 points; and needle retention time deviation exceeding 10 minutes deducts 5 points. The system outputs real-time visual correction prompts, marking the correct acupoint location, needle insertion angle, and depth on the screen. It also provides voice correction guidance, such as "Needle insertion angle is too large; please adjust to vertical insertion," achieving real-time standardized control of acupuncture operations and ensuring the full expression of acupuncture effects from the operational stage.

[0052] The effect bias attribution analysis module is used to perform attribution analysis based on operational standardization scores and full-cycle multimodal data when the acupuncture intervention effect fails to meet the preset target. Specifically, it sets a threshold for achieving the desired acupuncture intervention effect, based on clinical guidelines and industry consensus: an improvement rate of ≥10% for core chronic disease indicators and a decrease of ≥20% in TCM symptom scores over two consecutive treatment courses. When the patient's intervention effect fails to meet this threshold, the attribution analysis process is automatically triggered. The system first retrieves the average standardization score of all acupuncture operations within the treatment course, with a preset score threshold of 70. If the average score is below 70, it is determined that the effect is biased due to non-standard operation, that is, the core reason for the poor effect is the lack of the "technique" aspect, rather than the inappropriate plan. The system does not adjust the core acupuncture plan, but outputs targeted operation correction guidance through the output and interaction modules, including operation error analysis, standardized operation teaching videos, and targeted practice suggestions. If the average score is ≥70, it is determined that the acupuncture plan itself is unsuitable, that is, the core reason for the poor effect is that the "theory, method, and prescription" aspects do not match the patient's pathogenesis. The system automatically triggers the full-process plan optimization process of the treatment course-level adaptive acupuncture decision engine to regenerate a personalized acupuncture plan that is adapted to the patient's current state. This module incorporates a quantitative model of "operational deviation-effect decay," which fits a linear relationship between the operational standardization score and the effect decay rate based on 5,000 historical clinical data. The formula is: Effect decay rate = 1 - 0.008 × Operational standardization score. For patients whose operational standardization score is consistently below 60, the system automatically triggers a course-level adaptive acupuncture decision engine to adjust to an acupuncture intervention method with a higher fault tolerance and lower operational threshold, thereby reducing the difficulty of operation and ensuring the stability of the intervention effect.

[0053] The multidisciplinary collaborative decision-making module and the output and interaction module follow the mechanisms of modern multidisciplinary comprehensive management of chronic diseases and the synergistic effect mechanism of acupuncture intervention: Chronic diseases are complex diseases caused by the combined effects of multiple factors such as genetics, environment, behavior, and metabolism. Single acupuncture intervention cannot achieve comprehensive management of chronic diseases. There are significant synergistic or antagonistic effects between acupuncture and drug and lifestyle interventions. A reasonable synergistic plan can maximize the intervention effect while avoiding the safety risks of combined interventions (such as the risk of hypoglycemia caused by the synergy of hypoglycemic drugs and acupuncture). This module deeply integrates acupuncture conditioning into the multidisciplinary comprehensive management system of chronic diseases, realizes the synergistic optimization of multiple intervention methods, and conforms to the core principles of modern chronic disease management.

[0054] The multidisciplinary collaborative decision-making module communicates with the full-cycle time-series data fusion and acupuncture effect quantification module, the treatment course-level adaptive acupuncture decision engine, and the output and interaction module. The output and interaction module serves as the system's human-computer interaction port, enabling solution output and feedback collection.

[0055] The multidisciplinary collaborative decision-making module is used to establish a collaborative optimization model for acupuncture intervention, medication adjustment, and lifestyle intervention based on multimodal fusion data. When a patient's medication is adjusted, the system automatically optimizes the acupuncture plan to mitigate safety risks and enhance synergistic effects. Simultaneously, based on the quantitative results of acupuncture effects, it outputs collaborative suggestions for comprehensive chronic disease management. Specifically, the collaborative optimization model uses a non-dominated ranking genetic algorithm (NSGA-II) with an elitist strategy for multi-objective optimization. The optimization objectives are three-dimensional: maximizing the improvement rate of core chronic disease indicators, minimizing intervention safety risks, and minimizing the patient's implementation burden. When a patient's medication plan is adjusted, such as increasing or decreasing the dosage of hypoglycemic or antihypertensive drugs, or changing the type of drug, the system automatically adjusts the acupoint combinations and stimulation parameters of the acupuncture plan. For example, when increasing the dosage of hypoglycemic drugs, the stimulation intensity of the main hypoglycemic acupoints is reduced to avoid the risk of hypoglycemia, while optimizing the acupoint combinations to enhance the synergistic hypoglycemic effect with the medication. When a patient's diet, exercise, and sleep patterns change significantly, the system automatically adjusts the timing, frequency, and parameters of acupuncture intervention. Meanwhile, based on the quantitative results of acupuncture effects, the system outputs comprehensive management and collaborative suggestions that conform to the national chronic disease management guidelines, including suggestions for medication adjustment, dietary structure optimization, exercise prescriptions, and sleep improvement. All suggestions are marked with evidence-based levels and clearly stated as "for clinicians' reference only and not a substitute for clinical decision-making by licensed physicians," which meets the compliance requirements of medical software.

[0056] The output and interaction module is used to output the generated acupuncture treatment plan and receive user feedback and clinical effect data to update the system algorithm model. Specifically, it includes three interfaces: a physician-side web interface deployed on hospital / community computers, with core functions including a patient full-cycle data visualization dashboard, an acupuncture plan recommendation and online editing interface, a real-time acupuncture operation monitoring interface, acupuncture effect assessment report, and a multidisciplinary collaborative suggestion viewing interface, supporting clinicians to adjust, confirm, archive, and print the system-recommended plans; a patient-side mobile app deployed on the patient's smartphone, with core functions including acupuncture plan viewing, standardized operation guidance videos, a home data entry portal, a symptom self-assessment interface, safety warning prompts, and a doctor-patient communication channel; and a management-side backend system deployed on a cloud server, with core functions including clinical data statistical analysis, user permission management, knowledge base update and maintenance, system parameter configuration, and security audit log viewing. All interfaces comply with medical software human-computer interaction standards, with simple operation processes adapted to clinical work scenarios.

[0057] The execution flow of the multimodal data fusion and adaptive decision-making method for acupuncture therapy in chronic disease management corresponds one-to-one with the above system modules, fully following the aforementioned TCM mechanisms, algorithmic mechanisms, and clinical evidence-based mechanisms. The complete implementation steps are as follows: Step S1: Multimodal full-cycle data acquisition. Through the multimodal data acquisition module, multi-source data of the entire acupuncture treatment cycle of the target chronic disease patient is collected, including at least preoperative baseline data, real-time data during acupuncture, postoperative interval data between two acupuncture treatments, time-series follow-up data of continuous treatment courses, chronic disease medication data and lifestyle data. Data preprocessing, structured storage and encrypted transmission are completed to fully cover the entire cycle of acupuncture effect expression. Step S2 involves full-cycle data fusion and precise effect quantification. Time alignment and multi-scale feature fusion are performed on full-cycle time-series multimodal data. The long-term lag effect of a single acupuncture intervention is quantified by the time-series mutual information algorithm. The independent contribution of acupuncture intervention, medication adjustment, and lifestyle changes to changes in chronic disease indicators is extracted by the causal forest algorithm. The independent effect quantification results of acupuncture intervention are output, solving the problems of unquantifiable lag of acupuncture effect and effect distortion caused by multiple factors. Step S3, Treatment-level Adaptive Acupuncture Decision Generation: Based on the full-cycle fusion features and the quantitative results of independent acupuncture effects, the treatment-level adaptive acupuncture decision engine generates acupuncture treatment plans that are suitable for the disease stage, TCM syndrome type, and intervention tolerance of the target chronic disease patient. This includes generating the optimal intervention timing and acupoint combination plan by matching the patient's chronic disease pathological rhythm based on the acupuncture time medicine algorithm, completing the adaptive switching between symptomatic treatment and root cause treatment plans based on preset thresholds, and completing the stepwise adaptation of acupuncture intervention methods and application scenarios based on the patient's condition. At the same time, the closed-loop iterative optimization of the plan is realized based on the effect data of continuous treatment courses, which fully follows the core principles of TCM syndrome differentiation and treatment. Step S4, acupuncture operation monitoring and effect attribution optimization: Through the acupuncture operation standardization monitoring and correction module, the acupuncture operation process is monitored in real time, and correction guidance and standardization score are output; when the intervention effect does not reach the preset target, the effect deviation attribution analysis module completes the accurate attribution, and outputs operation correction guidance or triggers the scheme optimization process to achieve precise quality control of "separation of technique and method". Step S5, Multidisciplinary Collaborative Decision-Making and Model Iteration: Through the multidisciplinary collaborative decision-making module, the collaborative optimization of acupuncture intervention, medication adjustment, and lifestyle intervention is achieved, and collaborative suggestions for comprehensive chronic disease management are output. Through the output and interaction module, clinical operation feedback and effect data are received, and the system's full-process algorithm model is continuously iterated and optimized based on incremental reinforcement learning to adapt to the dynamic pathogenesis changes of patients. Step S6, Solution Output and Archiving: The generated acupuncture treatment plan is output to the physician and patient terminals through the output and interaction module, completing the compliant archiving and storage of clinical data throughout the entire process.

[0058] Example 1: This example uses type 2 diabetes patients as the application object to fully verify the whole process execution logic of this system in the management of metabolic chronic diseases, covering the functions and mechanisms of all core modules of the system.

[0059] Patient Wang, male, 56 years old, diagnosed with type 2 diabetes for 4 years. At the time of this visit, his fasting blood glucose was 8.5 mmol / L, his 2-hour postprandial blood glucose was 12.3 mmol / L, and his glycated hemoglobin was 7.9%. He was taking metformin twice daily at the usual dose and had no serious complications such as diabetic nephropathy or retinopathy. According to the four diagnostic methods of traditional Chinese medicine, his tongue was red with little saliva and his pulse was weak and thready. The syndrome was diagnosed as deficiency of both qi and yin with blood stasis. The core pathogenesis was deficiency of qi and yin in the spleen and stomach and obstruction of blood vessels by blood stasis. His chronic disease risk stratification was high risk, which is consistent with the clinical application scope of this system.

[0060] Initial baseline data collection and initial treatment plan generation: The system completes the collection of comprehensive baseline data for patients through a multimodal data acquisition module, including demographic information, diabetes history, tongue and pulse data, blood glucose and biochemical indicators, whole-body acupoint sensitivity test results, medication records, and daily diet and exercise habits data. Data preprocessing and structured storage are then performed to fully quantify the patient's initial pathogenesis. A full-cycle time-series data fusion and acupuncture effect quantification module completes the initial multimodal feature fusion. A treatment-level adaptive acupuncture decision engine generates an initial personalized acupuncture plan based on the patient's core pathogenesis of Qi and Yin deficiency with blood stasis: intervention method. The treatment involved electroacupuncture combined with warm needling. The main acupoints were Zusanli (ST36), Sanyinjiao (SP6), Pishu (BL20), Shenshu (BL23), and Yishu (BL21), with Taixi (KI3), Xuehai (SP10), and Diji (SP8) as secondary acupoints. The tonification and sedation techniques were balanced, taking into account both tonifying Qi and nourishing Yin while promoting blood circulation and unblocking collaterals. Based on the patient's 2-hour postprandial blood glucose peak rhythm and the Meridian Flow Method algorithm, the optimal intervention time was recommended to be Chen Shi (7:00-9:00 AM, when the Stomach Meridian is at its peak, corresponding to the core pathogenesis of spleen and stomach dysfunction in diabetes). The intervention frequency was twice a week, with 2 weeks constituting one course of treatment. At the same time, based on the patient's high-risk stratification, an auxiliary intervention plan suitable for community health service centers was generated, along with standardized operating procedures and safety warning rules.

[0061] Intraoperative operation control and real-time monitoring: During the treatment, the acupuncture operation standardization monitoring and correction module uses an industrial machine vision camera to collect the physician's operation in real time, compares it with the national standard "Acupuncture Technique Operation Specifications" to complete key point detection, and outputs an operation standardization score of 93 points throughout the process, with no obvious operation deviations. Only one minor adjustment prompt is given for the needle insertion depth, ensuring the standardization of the operation. The intraoperative real-time data acquisition unit simultaneously collects physiological data such as the patient's heart rate, blood oxygen, and skin temperature. There are no abnormal fluctuations throughout the process, and the system does not trigger any safety warnings.

[0062] Inter-treatment data collection and effect quantification: During the interval between two treatments, the system connects to the patient's continuous blood glucose monitor and medical-grade smart wearable device through the home-based inter-treatment data collection unit to continuously collect daily blood glucose data, sleep, diet, exercise data, and symptom self-assessment data. Simultaneously, it connects to the hospital's electronic medical record system to confirm that the patient's medication regimen has not been adjusted. The full-cycle time-series data fusion and acupuncture effect quantification module calculates the lag effect coefficient of this acupuncture intervention as 0.85 and the peak effect lag time as 3 days through the time-series mutual information algorithm, accurately capturing the long-term lag effect of acupuncture. After removing multi-factor interference through the causal forest algorithm, it is found that the independent contribution of acupuncture intervention to the patient's fasting blood glucose reduction is 72%, and the contribution of diet and medication adjustment is 28%, accurately quantifying the real conditioning effect of acupuncture.

[0063] Effect Attribution and Adaptive Optimization of the Treatment Plan: After two consecutive treatment courses, the patient's fasting blood glucose stabilized at 7.0-7.5 mmol / L, and the TCM symptom score decreased by 18%, which did not reach the preset effect achievement threshold. The system automatically triggered the effect deviation attribution analysis process. The average standardization score of all operations within the treatment course was retrieved and found to be 90 points, which is higher than the preset threshold of 70 points. The influence of non-standard operation was excluded, and it was determined that the acupuncture plan itself did not match the patient's dynamic pathogenesis, triggering the full-process optimization of the treatment course-level adaptive acupuncture decision engine. System analysis revealed that the effect attenuation coefficient of the main acupoint Zusanli was 0.55, indicating the emergence of stimulation tolerance and the acupoint entering a dull state. Simultaneously, the patient's TCM syndrome differentiation model confidence level changed by 22%, and the syndrome type changed from Qi and Yin deficiency with blood stasis to simple Qi and Yin deficiency. The superficial excess syndrome of blood stasis obstructing the collaterals has been alleviated, and the core pathogenesis has shifted. The system automatically optimized the treatment plan: the tolerant acupoint Zusanli was replaced with the backup acupoint Yanglingquan on the same meridian, the main acupoints were adjusted to Sanyinjiao, Pishu, Shenshu, Yishu, and Yanglingquan, the blood-activating acupoint Xuehai was removed, the electroacupuncture frequency was adjusted from 2Hz to 10Hz, and based on the patient's updated blood glucose rhythm, the optimal intervention time was adjusted to Si Shi (9:00-11:00, when the spleen meridian is dominant, matching the core pathogenesis of Qi and Yin deficiency).

[0064] Efficacy verification and long-term management: After one course of optimized treatment, the patient's fasting blood glucose stabilized at 6.5-7.0 mmol / L, glycated hemoglobin decreased to 7.0%, and the TCM symptom score decreased by 32%, achieving the preset effect target. The system automatically switched to a home-based self-intervention plan based on the patient's low-risk and stable state, recommending auricular acupressure + home-based gentle moxibustion, accompanied by visual acupoint selection videos, operation instructions, and safety warning rules. The multidisciplinary collaborative decision-making module simultaneously outputs appropriate chronic disease management collaborative suggestions, including a low-salt and low-fat diet prescription and a suggestion of 150 minutes of moderate-intensity aerobic exercise per week. It also provides auxiliary optimization references for the patient's medication plan. All suggestions are marked with evidence-based evidence levels and are for clinicians' reference only.

[0065] Example 2: This example uses patients with primary hypertension as the application object to fully verify the whole process execution logic of this system in the management of cardiovascular chronic diseases. It focuses on verifying the core functions and mechanisms of the system, such as pathological rhythm-acupuncture spatiotemporal fusion, adaptive switching between symptoms and urgency, and medication synergistic optimization, forming a differentiated coverage of chronic disease scenarios compared with Example 1.

[0066] Patient Li, female, 62 years old, was diagnosed with stage 3 (very high risk) essential hypertension 5 years ago. At the time of this visit, her office blood pressure was 158 / 96 mmHg, the average 24-hour ambulatory blood pressure was 148 / 92 mmHg, and the highest systolic blood pressure during the morning peak was 172 mmHg. She was taking a regular dose of calcium channel blocker once a day. She had no history of serious cardiovascular or cerebrovascular events such as stroke or myocardial infarction. According to the four diagnostic methods of traditional Chinese medicine, she had dizziness, weakness in the lower back and knees, red tongue with yellow coating, and wiry, thready and rapid pulse. The syndrome was diagnosed as liver yang hyperactivity combined with liver and kidney yin deficiency. The core pathogenesis was liver and kidney yin deficiency and liver yang hyperactivity, which is consistent with the clinical application scope of this system.

[0067] Initial baseline data collection and initial treatment plan generation: The system completes the collection of the patient's full-dimensional baseline data through a multimodal data acquisition module, including hypertension history, target organ damage assessment results, tongue and pulse data, dynamic blood pressure monitoring data, acupoint sensitization test results, antihypertensive medication records, daily routine and emotional state data, and completes data preprocessing and structured storage; the full-cycle time-series data fusion and acupuncture effect quantification module completes the initial multimodal feature fusion, and the treatment-level adaptive acupuncture decision engine generates an initial personalized acupuncture plan based on the patient's core pathogenesis of liver yang hyperactivity and liver and kidney yin deficiency: the intervention method is filiform needle + electroacupuncture, and the main acupoints are Baihui and Fengchi. The acupoints selected were Taichong (LR3), Taixi (KI3), and Sanyinjiao (SP6), with Ganshu (BL18) and Shenshu (BL23) as supplementary points. The tonifying and reducing techniques were a combination of even tonification and reducing combined with the rotating reducing technique at Taichong (LR3), which aimed to both calm the liver and subdue yang while nourishing the liver and kidneys. The morning blood pressure peak period was extracted from the patient's blood pressure using the STL time series decomposition algorithm and identified as Mao Shi (5:00-9:00 AM). Combined with the Meridian Flow and Na Jia method algorithm, the optimal intervention time was recommended to be Mao Shi (5:00-7:00 AM, when the Large Intestine Meridian is dominant, corresponding to the earlier morning blood pressure peak). The intervention frequency was 3 times per week, with 2 weeks constituting one course of treatment. At the same time, based on the patient's high-risk stratification, a professional intervention plan exclusive to the hospital's acupuncture department was generated, along with strict intraoperative blood pressure monitoring rules and safety emergency plans.

[0068] Intraoperative operation control and specimen urgency switching: During the first treatment, the patient's preoperative resting blood pressure was 156 / 94 mmHg. Ten minutes after treatment, the blood pressure rose to 165 / 102 mmHg, accompanied by worsening dizziness. The system captured the abnormal indicators through the intraoperative real-time data acquisition unit. The specimen urgency adaptive switching unit automatically triggered an acute exacerbation warning. Following the principle of "treating the symptoms in emergencies", the system immediately switched to the symptomatic intervention plan: the original balanced tonification and sedation operation was suspended, and strong stimulation of Renying and Quchi acupoints was added for sedation. Blood pressure changes were monitored in real time. After 15 minutes, the patient's blood pressure dropped back to 150 / 90 mmHg, and the dizziness was relieved. The system completed the full process recording and archiving of this emergency event.

[0069] Inter-treatment data collection and effect quantification: During the interval between two treatments, the system connects to the patient's medical electronic blood pressure monitor and medical-grade smart wearable device through the home-based inter-treatment data collection unit to continuously collect blood pressure measurement data at fixed times in the morning and evening, daily dizziness symptom self-assessment data, and sleep and mood data, while simultaneously confirming that the patient's antihypertensive medication regimen remains stable; The full-cycle time-series data fusion and acupuncture effect quantification module calculates the lag effect coefficient of this acupuncture intervention as 0.82 and the peak effect lag time as 2 days through the time-series mutual information algorithm; After removing multi-factor interference through the causal forest algorithm, it is found that the independent contribution of acupuncture intervention to the reduction of the patient's 24-hour average systolic blood pressure is 68%, while the contribution of antihypertensive drugs and lifestyle adjustments is 32%.

[0070] Effect attribution and adaptive optimization of the treatment plan: After two consecutive treatment courses, the patient's blood pressure in the clinic stabilized at around 145 / 90 mmHg, and the dizziness symptom score decreased by 17%, which did not reach the preset effect target threshold. The system automatically triggered the effect deviation attribution analysis process. The average standardization score of all operations in this treatment course was retrieved and found to be 91 points, which was higher than the preset threshold. The influence of non-standard operation was excluded, and it was determined that the treatment plan was not suitable enough, triggering the full-process optimization of the treatment plan. System analysis revealed that the effect attenuation coefficient of the main acupoint Fengchi was 0.58, indicating the emergence of stimulation tolerance and the acupoint entering a desensitized state. Simultaneously, the patient's TCM diagnostic model confidence level changed by 23%, shifting the syndrome from Liver Yang Rising with Liver and Kidney Yin Deficiency to Liver and Kidney Yin Deficiency. The symptoms of Liver Yang Rising had been alleviated, but the core pathogenesis had shifted. The system automatically optimized the treatment plan: replacing the tolerant acupoint Fengchi with the backup acupoint Tianzhu; adjusting the main acupoints to Baihui, Taichong, Taixi, Sanyinjiao, and Tianzhu; removing the liver-calming and yang-subduing acupoint Ganshu; adjusting the electroacupuncture frequency from 50Hz to 2Hz; and, based on the patient's updated blood pressure rhythm, adjusting the optimal intervention time to Chenshi (7:00-9:00 AM) to match the patient's current blood pressure fluctuation pattern and core pathogenesis.

[0071] Efficacy Verification and Multidisciplinary Collaborative Management: After one course of optimized treatment, the patient's office blood pressure stabilized below 130 / 80 mmHg, the 24-hour ambulatory blood pressure average was 128 / 78 mmHg, the morning blood pressure surge was significantly alleviated, and the dizziness symptom score decreased by 35%, achieving the preset effect target. The multidisciplinary collaborative decision-making module provided auxiliary reference suggestions for reducing the dosage of antihypertensive drugs based on the patient's stable blood pressure, and automatically optimized the acupuncture plan to reduce the intensity of acupoint stimulation and avoid the risk of hypotension caused by the synergistic effect of drugs and acupuncture. The system will then switch the patient to a tiered management plan combining community and home care, with weekly community acupuncture intervention, combined with home auricular acupressure self-intervention, and supporting home blood pressure monitoring and early warning rules to achieve long-term stable management of hypertension.

[0072] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multimodal fusion adaptive decision support system for acupuncture treatment of chronic diseases, characterized in that, It includes a multimodal data acquisition module with sequential communication connections, a full-cycle time-series data fusion and acupuncture effect quantification module, a treatment course-level adaptive acupuncture decision engine, and an output and interaction module; The multimodal data acquisition module is used to collect multi-source data of the entire acupuncture treatment cycle of the target chronic disease patient, including at least preoperative baseline data, real-time data during acupuncture, postoperative interval data between two acupuncture treatments, time-series follow-up data of continuous treatment courses, chronic disease medication data and lifestyle data. The full-cycle time-series data fusion and acupuncture effect quantification module is used to perform time alignment and feature fusion on full-cycle time-series multimodal data, quantify the long-term lag effect of a single acupuncture intervention through a time-series correlation algorithm, and separate the independent contributions of acupuncture intervention, medication adjustment, and lifestyle changes to changes in chronic disease indicators through a causal inference algorithm, and output the independent effect quantification results of acupuncture intervention. The treatment course-level adaptive acupuncture decision engine is used to generate acupuncture treatment plans that are adapted to the disease stage, TCM syndrome type, and intervention tolerance of the target chronic disease patients based on full-cycle fusion data and acupuncture independent effect quantification results, and to achieve closed-loop iterative optimization of the plan based on the effect data of continuous treatment courses. The output and interaction module is used to output the generated acupuncture treatment plan and receive user operation feedback and clinical effect data to update the system algorithm model.

2. The system according to claim 1, characterized in that, The multimodal data acquisition module includes a home-based inter-treatment data acquisition unit that connects to wearable devices and patient self-assessment terminals. This unit is used to collect core physiological indicators of chronic diseases, symptom self-assessment data, sleep, diet, activity data, and acupoint sensitization time-series data between two acupuncture treatments. The full-cycle time-series data fusion and acupuncture effect quantification module has a built-in time-series mutual information algorithm unit. This unit is used to establish a time-series correlation model of "this acupuncture treatment plan - postoperative inter-treatment multimodal data changes - preoperative baseline data for the next acupuncture treatment" to quantify the long-term lag effect coefficient of a single acupuncture intervention.

3. The system according to claim 1, characterized in that, The treatment-level adaptive acupuncture decision engine includes an acupoint tolerance adaptive correction unit and a syndrome dynamic drift decision unit. The acupoint tolerance adaptive correction unit is used to calculate the stimulation effect attenuation coefficient of the target acupoint based on the quantitative results of the independent effects of acupuncture in continuous treatment courses, and adaptively adjust the stimulation parameters and acupoint combinations of the acupuncture plan. The syndrome dynamic drift decision unit has a built-in incremental learning TCM syndrome differentiation model, which is used to dynamically update the patient's TCM syndrome differentiation results based on multimodal data in continuous treatment courses, and adaptively adjust the acupoint combinations and tonification / sedation strategies of the acupuncture plan based on the drift of the syndrome differentiation results.

4. The system according to claim 1, characterized in that, The treatment-level adaptive acupuncture decision engine includes a chronic disease rhythm-acupuncture spatiotemporal fusion unit and a symptom-symptom-urgency adaptive switching unit. The chronic disease rhythm-acupuncture spatiotemporal fusion unit incorporates the Meridian Flow and Linggui Bafa acupuncture time medicine algorithm models to analyze patients' chronic disease pathological rhythm data, establish a matching mapping relationship between "pathological rhythm peak - acupoint opening and closing time", calculate the optimal acupuncture intervention time window, and dynamically adjust acupoint combinations. The symptom-symptom-urgency adaptive switching unit is preset with multimodal data trigger thresholds to automatically switch the symptom-symptom intervention plan when the patient's indicators exceed the warning level, and automatically switch back to the symptom-symptom treatment plan when the indicators return to a stable period.

5. The system according to claim 1, characterized in that, The treatment-level adaptive acupuncture decision engine includes an intelligent acupuncture intervention method adaptation unit and a full-scenario tiered solution adaptation unit. The intelligent acupuncture intervention method adaptation unit has a built-in acupuncture intervention method-syndrome-effect association knowledge base, which is used to automatically recommend the optimal acupuncture intervention method and match the corresponding acupoint combination and stimulation parameters based on the patient's syndrome differentiation results, disease stage, and tolerance. The acupuncture intervention methods include at least filiform needles, electroacupuncture, warm acupuncture, moxibustion, auricular acupressure, and acupoint application. The full-scenario tiered solution adaptation unit has preset three-level scenario adaptation rules of "hospital-community-home", which is used to generate acupuncture plans for corresponding scenarios and provide corresponding operation specifications and safety warning rules based on the patient's chronic disease risk stratification and data stability.

6. The system according to claim 1, characterized in that, It also includes an acupuncture operation standardization monitoring and correction module that is communicatively connected to the multimodal data acquisition module and the output and interaction module. The module is used to acquire acupuncture operation process image data in real time through machine vision equipment, compare it with preset standard operation specifications to generate an operation standardization score, and output visual operation correction prompts in real time.

7. The system according to claim 6, characterized in that, It also includes an effect deviation attribution analysis module that is connected to the full-cycle time-series data fusion and acupuncture effect quantification module, an acupuncture operation standardization monitoring and correction module, and a treatment course-level adaptive acupuncture decision engine. The module is used to complete attribution analysis based on operation standardization score and full-cycle multimodal data when the acupuncture intervention effect does not reach the preset target: if the operation standardization score is lower than the preset threshold, it is determined to be an effect deviation caused by non-standard operation, and operation correction guidance is output; if the operation standardization score meets the standard, it is determined to be a mismatch between the plan and the plan optimization process of the decision engine is triggered.

8. The system according to claim 1, characterized in that, It also includes a multidisciplinary collaborative decision-making module that is communicatively connected to the full-cycle time-series data fusion and acupuncture effect quantification module, the treatment course-level adaptive acupuncture decision engine, and the output and interaction module; the module is used to establish a collaborative optimization model of acupuncture intervention, medication adjustment, and lifestyle intervention based on multimodal fusion data, automatically optimize the acupuncture plan to avoid safety risks and enhance the synergistic effect when the patient's medication is adjusted, and output collaborative suggestions for comprehensive chronic disease management based on the acupuncture effect quantification results; The system also has a built-in incremental reinforcement learning unit, which is used to continuously iterate and optimize the system's full-process algorithm model based on clinical effect data and user feedback.

9. A multimodal data fusion and adaptive decision-making method for acupuncture treatment in chronic disease management, characterized in that, Includes the following steps: S1. Collect multi-source data of the entire acupuncture treatment cycle for target chronic disease patients, including at least preoperative baseline data, real-time data during acupuncture, postoperative interval data between two acupuncture treatments, time-series follow-up data of continuous treatment courses, chronic disease medication data and lifestyle data. S2. Perform time alignment and feature fusion on full-cycle time-series multimodal data, quantify the long-term lag effect of a single acupuncture intervention through a time-series correlation algorithm, and separate the independent contributions of acupuncture intervention, medication adjustment, and lifestyle changes to changes in chronic disease indicators through a causal inference algorithm, and output the quantitative results of the independent effect of acupuncture intervention. S3. Based on the full-cycle fusion data and the quantitative results of the independent effects of acupuncture, generate acupuncture treatment plans that are suitable for the disease stage, TCM syndrome type, and intervention tolerance of the target chronic disease patients, and realize closed-loop iterative optimization of the plans based on the effect data of continuous treatment courses. S4. Output the generated acupuncture treatment plan and receive user feedback and clinical effect data to update the algorithm model.

10. The method according to claim 9, characterized in that, Step S3 also includes: generating optimal intervention timing and acupoint combination schemes based on the patient's chronic disease pathological rhythm matched by the acupuncture time medicine algorithm; adaptively switching between symptomatic and root-cause treatment schemes based on preset thresholds; and stepwise adaptation of acupuncture intervention methods and application scenarios based on the patient's condition. The method also includes a step of real-time monitoring and correction of acupuncture operation standardization, a step of attribution analysis of acupuncture effect deviation, a step of multidisciplinary collaborative decision-making of acupuncture, chronic disease medication and lifestyle intervention, and a step of iterative optimization of the algorithm model based on incremental reinforcement learning.

Images