Multi-modal medical information intelligent integration and decision support system for acupuncture rehabilitation

Abstract

The invention relates to a multi-modal medical information intelligent integration and decision support system for acupuncture rehabilitation, which belongs to the technical field of medical information, and comprises a data acquisition and synchronization unit used for acquiring physiological data, psychological data and behavior data of athletes; the data acquisition and synchronization unit is also used for performing timestamp alignment processing on the acquired physiological data, psychological data and behavior data to generate a data stream with a global timestamp; the state characterization and evaluation unit is used for extracting multi-modal features based on the data flow generated by the data acquisition and synchronization unit; the state characterization and evaluation unit is further used for processing the multi-modal features through a preset disposable physiological resource dynamic evaluation model so as to estimate a current state vector representing the current physiological state of the athlete. Advanced algorithms such as a convolutional neural network and wavelet transform are used for extracting deep features from original data, noise and redundant information are filtered out, and highly-condensed feature vectors are generated;

Description

Technical Field

[0001] The present invention relates to the field of medical information, and in particular to a multimodal medical information intelligent integration and decision support system for acupuncture rehabilitation. Background Art

[0002] The core goal of postoperative rehabilitation for elite athletes is to safely return to peak competitive form in the shortest possible time. Existing rehabilitation decision support systems often adhere to standardized operating procedures based on evidence-based medicine, a principle known as procedural compliance. These systems emphasize the phasing and standardization of rehabilitation programs, striving for predictable progress through strict implementation. However, the athlete's recovery process is a highly nonlinear dynamic system, whose immediate bioefficiency is influenced in real time by multiple factors, including training, treatment, and psychological stress. Especially under the influence of high-frequency competitive pressure, an athlete's physiological adaptability threshold is significantly lowered. At this point, a destructive coupling between adherence to standard procedures and the pursuit of maximizing immediate bioefficiency can occur. Forcing athletes to perform high-load, standard tasks can lead to a state of latent overtraining, a sharp decline in recovery rate, and even secondary injury, forming a negative feedback loop. This phenomenon reveals a cognitive blind spot in existing technologies, namely the problem of temporal mismatch of physiological resources. Existing systems plan tasks along the time axis, but ignore the energy axis of the athlete's body, forcibly arranging high-consumption tasks during the low period of physiological resources, resulting in ineffective allocation of resources. Therefore, this field needs a technical solution that can actively manage the athlete's physiological energy cycle and dynamically optimize rehabilitation decisions to solve the defect of temporal mismatch of physiological resources.

[0003] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0004] The purpose of the present invention is to provide a multimodal medical information intelligent integration and decision support system for acupuncture rehabilitation to solve the problems raised in the above background technology.

[0005] The technical solution of the present invention is to include: A data acquisition and synchronization unit is used to collect the athlete's physiological data, psychological data, and behavioral data; the data acquisition and synchronization unit is also used to align the timestamps of the collected physiological data, psychological data, and behavioral data to generate a data stream with a global timestamp; a state characterization and evaluation unit for extracting multimodal features based on the data stream generated by the data acquisition and synchronization unit; the state characterization and evaluation unit is further configured to process the multimodal features using a preset available physiological resource dynamic evaluation model to estimate a current state vector representing the athlete's current physiological state; a state evolution prediction unit configured to, based on the sequence of historical state vectors estimated by the state characterization and assessment unit, process using a preset physiological resource evolution prediction engine to solve a sequence of predicted state vectors at future time steps; a decision optimization and instruction generation unit configured to, based on the sequence of predicted state vectors solved by the state evolution prediction unit, generate an optimal intervention sequence; the decision optimization and instruction generation unit is further configured to parse the optimal intervention sequence into a dynamic rehabilitation plan; a closed-loop correction unit configured to calculate a prediction error after execution of the dynamic rehabilitation plan; the closed-loop correction unit is further configured to, in response to the prediction error exceeding a preset error threshold, adaptively correct system parameters of the disposable physiological resource dynamic assessment model and the physiological resource evolution prediction engine.

[0006] Preferably, the physiological data collected by the data collection and synchronization unit includes heart rate variability data, skin conductance response data, tongue image data, and pulse waveform data; the behavior data includes biochemical indicators and subjective recovery perception data.

[0007] Preferably, the process of extracting multi-modal features by the state characterization and assessment unit is as follows: determining statistical features and frequency domain power distribution features of the heart rate variability data, and generating an HRV feature vector; processing the tongue image data using a preset convolutional neural network to extract a quantized feature vector; analyzing the pulse waveform data using wavelet transform to extract a pulse feature vector.

[0008] Preferably, the disposable physiological resource dynamic assessment model is a linear Gaussian state space model; the current state vector includes disposable physiological resources, physiological fatigue index, and inflammation level; the state characterization and assessment unit uses a Kalman filter algorithm to estimate the current state vector in combination with the state at the last time and the observation vector composed of the multi-modal features at the current time.

[0009] Preferably, the physiological resource evolution prediction engine is a long short-term memory network based on an attention mechanism; the input of the state evolution prediction unit includes a sequence of historical state vectors and a known future input sequence; the attention mechanism is used to assign weights to events in the historical sequence when predicting future states.

[0010] Preferably, the decision optimization and instruction generation unit generates the optimal intervention sequence by constructing and solving a quadratic objective function; the quadratic objective function aims to: minimize the weighted error between the sequence of predicted state vectors and a preset target state trajectory; minimize control input penalty; minimize deviation penalty from a standard procedure.

[0011] Preferably, the decision optimization and instruction generation unit also processes multiple constraints when solving the optimal intervention sequence; the multiple constraints include an upper limit of the daily total physiological load, and the upper limit of the daily total physiological load does not exceed a preset load threshold.

[0012] Preferably, the process of calculating the prediction error by the closed-loop correction unit is: Determine the actual change in the state vector after the intervention ends; The prediction error is quantified based on the actual change and compared with the estimated restoration benefits stored in the knowledge base of the cost-effectiveness of the intervention.

[0013] Preferably, the process of the closed-loop correction unit performing adaptive correction on the system parameters is: The prediction error is used as a feedback signal to fine-tune the state transfer matrix, input matrix, and observation matrix of the dynamic evaluation model of available physiological resources. The prediction error is used as a feedback signal to fine-tune the network weights of the physiological resource evolution prediction engine.

[0014] The present invention provides a multimodal medical information intelligent integration and decision support system for acupuncture rehabilitation through improvements. Compared with the existing technology, it has the following improvements and advantages: 1. This system achieves a comprehensive, profound, and quantitatively accurate assessment of an athlete's physiological state. Unlike existing technologies that rely on single or discrete indicators for one-sided assessment, this system uses a data acquisition and synchronization unit to fuse modern physiological data, such as heart rate variability and galvanic skin response, which reflect the state of the autonomic nervous system, with tongue images and pulse waveform data based on Traditional Chinese Medicine theory. This integration of multimodal information provides an unprecedentedly comprehensive and robust data foundation for subsequent assessment. The system's state characterization and assessment unit uses advanced algorithms such as convolutional neural networks and wavelet transforms to extract deep features from the raw data, filter out noise and redundant information, and generate highly concise feature vectors. This approach elevates the understanding of an athlete's state from traditional qualitative description to precise quantitative characterization. 2. A fundamental shift from passive response to forward-looking prediction has been achieved. This invention utilizes a linear Gaussian state-space model to abstract and quantify an athlete's internal state into core vectors, such as available physiological resources, physiological fatigue index, and inflammation level, for the first time, and uses a Kalman filter algorithm for optimal estimation. Furthermore, a state evolution prediction unit employs a long-short-term memory network based on an attention mechanism as its prediction engine. This engine not only learns the long-term patterns of state evolution, but its unique attention mechanism also enables intelligent identification and focus on historical events with significant future impact. This enables the system to accurately predict the peaks and troughs of future physiological resources, thereby shifting the basis for formulating rehabilitation strategies from a delayed response to past conditions to a forward-looking approach to future trends. 3. The rehabilitation decision-making is realized from experience-driven to scientific optimization; the decision optimization and instruction generation unit in the application constructs the generation of the rehabilitation plan as a rigorous multi-objective and multi-constraint optimization problem; it balances the cost, safety and compliance with evidence-based medical standard procedures of the intervention measures systematically by solving a quadratic objective function under the core premise of ensuring that the rehabilitation effect is promoted towards the preset target trajectory; at the same time, the introduction of key constraint conditions such as the upper limit of the total daily physiological load provides a solid safety guarantee for the entire rehabilitation process, effectively avoiding the risk of overtraining; the dynamic rehabilitation plan generated by the mechanism is the optimal solution under the condition of meeting all safety boundary conditions, thereby maximizing the efficiency of the rehabilitation process under the premise of ensuring safety; 4. The system is upgraded from a static model to adaptive evolution; the application innovatively contains a closed-loop correction unit; the unit forms a feedback signal by quantifying the error between the actual effect and the predicted effect after the execution of the rehabilitation plan; when the error exceeds the threshold of statistical significance, the system will use the signal to adaptively fine-tune the system parameters of the internal core model, including the state transition matrix, the input matrix and the observation matrix of the dynamic assessment model of the available physiological resources, and the network weights of the physiological resource evolution prediction engine; this self-correction and evolution capability enables the system to continuously learn and adapt to the individual uniqueness of the athlete, and its accuracy of evaluation, prediction and decision-making will continuously improve over time, achieving truly individualized precision rehabilitation. BRIEF DESCRIPTION OF DRAWINGS

[0015] The application will be further explained in conjunction with the accompanying drawings and examples: Figure 1 is the flow chart of the system of the application. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical scheme and advantages of the application more clear and explicit, the application will be further described in detail in conjunction with specific examples.

[0017] Example 1 Please refer to Figure 1 , the application provides a multi-modal medical information intelligent integration and decision support system for acupuncture rehabilitation, which includes: A data acquisition and synchronization unit is used to acquire physiological data, psychological data and behavior data of athletes; the data acquisition and synchronization unit is also used to perform time stamp alignment processing on the acquired physiological data, psychological data and behavior data to generate a data stream with a global time stamp; a state representation and evaluation unit configured to extract multi-modal features based on the data stream generated by the data acquisition and synchronization unit, and to process the multi-modal features by a pre-set disposable physiological resource dynamic evaluation model to estimate a current state vector representing the current physiological state of the athlete; a state evolution prediction unit configured to process the historical state vector sequence estimated by the state representation and evaluation unit by a pre-set physiological resource evolution prediction engine to solve a predicted state vector sequence at future time steps; a decision optimization and instruction generation unit configured to generate an optimal intervention sequence based on the predicted state vector sequence solved by the state evolution prediction unit, and to parse the optimal intervention sequence into a dynamic rehabilitation plan; In addition, the data acquisition and synchronization unit further comprises an outlier detection module. When it is detected that the sensor data exceeds the pre-set reasonable physiological range, for example, the heart rate is 0 or exceeds 250bpm, or the data stream has no change for a long time, the system will trigger an alarm and process the data source by interpolation or rejection to ensure the robustness of the subsequent state evaluation model. For example, the parsing process can be based on a pre-set intervention measure knowledge base which maps each dimension of the control vector to a specific intervention measure, such as acupuncture acupoint, physiotherapy method, duration, intensity, etc. When a certain dimension value of the solved optimal is a specific value, the system queries the knowledge base to generate corresponding specific rehabilitation operation instructions. a closed-loop correction unit configured to calculate the prediction error after the execution of the dynamic rehabilitation plan, and to perform adaptive correction on the system parameters of the disposable physiological resource dynamic evaluation model and the physiological resource evolution prediction engine in response to the prediction error exceeding a pre-set error threshold; The present application provides a multi-modal medical information intelligent integration and decision support system for acupuncture rehabilitation. The system aims to solve the physiological resource timing mismatch problem caused by sticking to standardized operations in the existing rehabilitation process, and to maximize the rehabilitation efficiency through accurate evaluation, dynamic prediction and adaptive decision of the physiological state of the athlete. The system comprises a data acquisition and synchronization unit which is configured to obtain comprehensive data reflecting the state of the athlete from multiple heterogeneous devices. In this embodiment, this unit continuously collects various data of the athlete through wearable sensing devices, special quantitative four-diagnosis devices of traditional Chinese medicine and software interfaces. After the data collection is completed, all data streams are time-aligned by applying the network time protocol and uniformly adding global time stamps to form time series data required for subsequent processing. The system further includes a state characterization and evaluation unit, which aims to extract deep physiological features from the raw data and estimate core physiological states that cannot be directly measured. In this embodiment, this unit performs feature engineering on the time-stamped data stream, and the processing results are input into a preset dynamic evaluation model of available physiological resources. This model can integrate multi-dimensional features to estimate the current state vector representing the athlete's current physiological state in real time. The system further includes a state evolution prediction unit, which aims to deduce the possible changes in the athlete's physiological state over a period of time in the future based on historical state information. In this embodiment, this unit receives the historical state vector sequence output by the state table EPC and the evaluation unit, and processes it using a preset physiological resource evolution prediction engine to calculate a predicted state vector sequence for multiple time steps in the future. The system further includes a decision optimization and instruction generation unit, which aims to calculate the optimal combination of intervention measures based on future predictions. In this embodiment, this unit uses the predicted state vector sequence output by the state evolution prediction unit as the core input, constructs and solves a multi-objective optimization problem to generate the optimal intervention sequence. This unit parses this sequence into a specific, executable dynamic rehabilitation plan. The system further includes a closed-loop correction unit, which is used to reversely correct the system's internal model by comparing the actual results of the plan execution with the expected results, thereby enabling the system to have self-learning and self-adaptive capabilities. In this embodiment, after the dynamic rehabilitation plan cycle is completed, this unit quantifies the prediction error generated. When the error exceeds a preset error threshold, the unit will be activated and use the error as a feedback signal to adaptively correct the internal system parameters of the available physiological resource dynamic assessment model and the physiological resource evolution prediction engine. Through the collaborative work of the above five units, this embodiment constructs a complete technical closed loop from data collection, status assessment, future prediction, optimization decision-making to closed-loop correction; it transforms the rehabilitation process from a traditional, fixed timeline task to a dynamic resource axis management, and intelligently schedules rehabilitation tasks by proactively predicting the peaks and troughs of physiological resources, thereby solving the risk of ineffective allocation of physiological resources and even secondary injuries caused by blindly executing standardized processes, and significantly improving the safety and efficiency of the rehabilitation process.

[0018] The physiological data collected by the data acquisition and synchronization unit include heart rate variability data, galvanic skin response data, tongue image data, and pulse waveform data; behavioral data include biochemical indicators and subjective recovery perception data; In this embodiment, the collection content of the data collection and synchronization unit is limited; The physiological data collected by this unit, in this embodiment, includes: heart rate variability data and galvanic skin response data collected by wearable devices, both of which are used to reflect the balance and arousal level of the autonomic nervous system; tongue image data obtained by standard light source and high-resolution camera, and pulse waveform data obtained by multi-array pressure sensor, both of which are used to quantify the body state from the perspective of traditional Chinese medicine; The behavioral data collected by this unit, in this embodiment, includes: biochemical indicators such as blood lactate and salivary cortisol entered through the professional equipment interface, and subjective recovery perception data entered through the software application, such as recovery scale scores; This embodiment achieves a deep integration of modern physiological indicators and quantitative information from the four diagnostic methods of traditional Chinese medicine by clarifying specific data modalities; data such as heart rate variability and galvanic skin response provide objective physiological information, while tongue and pulse data provide supplementary information from a macro and holistic perspective; this multimodal, multi-dimensional information input provides a more comprehensive and robust data foundation for subsequent state characterization and evaluation units, thereby significantly improving the accuracy and depth of the assessment of athletes' complex physiological states.

[0019] The process of extracting multimodal features by the state representation and evaluation unit is as follows: Determine the statistical characteristics and frequency domain power distribution characteristics of heart rate variability data and generate HRV feature vectors; A preset convolutional neural network is used to process tongue image data to extract quantitative feature vectors; Analyze pulse waveform data using wavelet transform to extract pulse characteristic vector; In this embodiment, the process of extracting multimodal features by the state representation and evaluation unit is described; The process of generating HRV feature vector is as follows: processing heart rate variability data, calculating its statistical characteristics, such as the standard deviation of adjacent heartbeat intervals, and its frequency domain power distribution characteristics, such as the ratio of high-frequency power to low-frequency power, and combining them into the vector; The process of extracting quantitative feature vectors involves processing tongue image data using a pre-trained convolutional neural network. This deep learning model is pre-trained on a dataset consisting of tens of thousands of tongue images annotated by experienced traditional Chinese medicine practitioners. This neural network automatically learns and identifies visual patterns associated with different physiological states, such as tongue color, tongue coating, and morphology, through multi-layer convolution and pooling operations, and outputs a quantitative feature vector representing the tongue image information. A. Convolutional neural network for extracting tongue image features: For example, the preset convolutional neural network can adopt a transfer learning strategy, that is, fine-tuning the ResNet-50 model pre-trained on the ImageNet large-scale dataset. The tens of thousands of tongue image datasets used for fine-tuning may include tongue color, such as pale white, red, and crimson purple; tongue coating, such as thin white and greasy yellow; and tongue shape, such as fat and tooth marks. The cross entropy loss function is used during model training, and the learning rate is . The Adam optimizer was trained for 200 cycles to obtain a model that can stably extract quantized feature vectors. The extraction process of pulse characteristic vector is as follows: using wavelet transform to analyze pulse waveform data; wavelet transform can effectively characterize the local characteristics of the signal in both time domain and frequency domain. Through this analysis, the dynamic characteristics reflecting the pulse intensity, rate, rhythm and morphology can be extracted and combined into the pulse characteristic vector; This embodiment achieves deep feature extraction of raw multimodal data by adopting specific, advanced signal processing and artificial intelligence technologies. Compared with simply using raw data, the HRV feature vectors, tongue quantification feature vectors, and pulse feature vectors generated by this embodiment can reflect the athlete's internal physiological state in a more essential and concise manner, effectively filtering out data noise and redundant information, and providing high-quality input guarantee for the accuracy of subsequent state assessment models.

[0020] The dynamic assessment model for available physiological resources is a linear Gaussian state-space model. The current state vector includes available physiological resources, physiological fatigue index, and inflammation level. The state representation and assessment unit uses a Kalman filter algorithm to combine the previous state with the current observation vector composed of multimodal features to estimate the current state vector. In this embodiment, the core model and algorithm used by the state representation and evaluation unit are defined; The dynamic evaluation model of available physiological resources used by this unit is, in this embodiment, a linear Gaussian state space model. The reason for choosing this model is that those skilled in the art should understand that although the physiological system of athletes is essentially nonlinear, within a specific, small working range, the use of a linear Gaussian state space model is an effective and computationally feasible approximation method. This system uses a closed-loop correction unit to adjust the model parameters. The continuous fine-tuning partially compensates for the deviation between the linear model and the real scene, and can optimally estimate the intrinsic, low-dimensional physiological state that cannot be directly measured from the noisy multi-dimensional observation data; the model consists of the state equation and the observation equation; The state equation is:

[0021] in, yes The current state vector at time t is defined as a three-dimensional vector, containing the three core latent states of the available physiological resources, the physiological fatigue index, and the inflammation level; : the time point in the time series; To enable those skilled in the art to understand and implement, the three components of the state vector are now exemplarily described: It should be noted that this three-dimensional state vector is a simplified representation of the athlete's core physiological state, aiming to grasp the main contradictions, and other important factors, such as psychological stress and sleep duration, are taken into the input vector as external influences, thereby indirectly affecting the evolution of the core state vector; Available physiological resources: This can be quantified as a normalized comprehensive score, for example, mainly composed of the high-frequency power in heart rate variability and subjective recovery perception scales such as RPE scores. The higher the score, the more abundant the resources; Physiological fatigue index: This can be quantified as an index related to exhaustion, for example, mainly positively correlated with the ratio of low-frequency to high-frequency power in heart rate variability and blood lactate concentration. The higher the index, the deeper the fatigue; Inflammation level: This can be quantified through proxy indicators, for example, by correlating with the quantified values of tongue color and tongue coating thickness in tongue image features, as well as the fluctuation characteristics of skin galvanic response; is the state vector at the previous time point; is the input vector at time t, with data sources from the behavior data collected by the data acquisition and synchronization unit and known intervention measures, including external influencing factors such as daily training load and psychological stress score; is the state transition matrix, describing the natural evolution of physiological state over time; is the input matrix, used to linearly quantify the direct influence of different interventions on each state change rate. This linear assumption is approximated under the closed-loop correction mechanism of the system by adaptively adjusting the parameters of matrix to simulate the average dose-effect relationship under different states; is the process noise, assumed to be Gaussian white noise; The observation equation is:

[0022] where, is the observation vector composed of multi-modal features extracted by the above method; is the observation matrix, which relates the internal physiological state to the externally measurable multi-modal features; is the measurement noise, also assumed to be Gaussian white noise; the matrix The initial value of is set based on physiological prior knowledge and is continuously optimized in subsequent closed-loop calibration; A. For the state-space model matrices A, B, C: For example, the initialization of these matrices can follow the following principles: State transition matrix : describes the natural evolution of the state. Its diagonal elements are usually close to 1, indicating that the state has a certain degree of persistence. The non-diagonal elements reflect the mutual influence between states. For example, an increase in the physiological fatigue index may lead to a slight increase in the inflammation level at the next moment. The element at the corresponding position in the matrix can be set to a small positive value; Input Matrix : Quantifying external intervention The impact of If the high-intensity training load is included in the matrix In the equation, the coefficient of the input corresponding to the state of disposable physiological resources should be negative, and the coefficient corresponding to the state of physiological fatigue index should be positive; Observation Matrix : Associated internal state With external observation For example, an increase in the inflammation level in the state vector is expected to lead to an increase in the quantified value of the tongue redness in the tongue image observation feature, then the matrix The coefficients of the corresponding positions of these two variables should be positive; To solve the model, the state representation and evaluation unit uses the Kalman filter algorithm in this embodiment; the algorithm is a recursive estimation algorithm that combines the state of the previous moment and the current observation vector composed of multimodal features , estimate the current state vector in real time The optimal value of This embodiment constructs a clear state-space model based on modern control theory to abstract and quantify the athlete's recovery status into core indicators such as available physiological resources. By utilizing the Kalman filter algorithm, it is possible to accurately understand the athlete's internal state from noisy multi-source data, providing a solid, quantitative foundation for subsequent predictions and decision-making, and achieving a leap from qualitative description to dynamic quantitative modeling of athlete status assessment.

[0023] Example 2 The physiological resource evolution prediction engine is an attention mechanism based long short-term memory network; the input of the state evolution prediction unit includes a historical state vector sequence and a known future input sequence; the attention mechanism is used to assign weights to events in the historical sequence when predicting future states; In this embodiment, the core prediction engine used by the state evolution prediction unit is described; The physiological resource evolution prediction engine used by the unit is an attention mechanism based long short-term memory network in this embodiment; the reason for choosing this model is that the long short-term memory network can effectively capture the long-term dependence relationship in the time series data through its internal gating mechanism, and the attention mechanism further enhances the performance of the model; the attention mechanism refers to a mechanism that simulates human cognitive attention, which can automatically give higher computational weights to the most influential events in the historical sequence when predicting future states; B. For the physiological resource evolution prediction engine: For example, the attention mechanism based long short-term memory network can be specifically designed as a network structure containing two layers of stacked LSTM, each layer containing 256 hidden units; in each time step prediction, the attention mechanism will calculate the weight according to the value size of the physiological fatigue index in the historical sequence, so as to give higher attention to those extremely fatigued historical state points; the network uses mean square error as the loss function for training; The input of the prediction engine includes two parts: the historical state vector sequence output by the state representation and evaluation unit, and the known future input sequence input externally, such as the planned training impulse or rehabilitation therapy arrangement in the future for several days; the output of the engine is a predicted state vector sequence of multiple time steps in the future; The prediction engine used in this embodiment not only learns the evolution law of the state like traditional time series models, but also realizes intelligent focusing on key historical events through the attention mechanism; this makes the prediction result more sensitive and accurate to the individual experience of the athlete, and can foresee the future physiological resource trough that may be caused by a certain specific historical event, thereby providing high-quality prediction information for the system to make forward-looking and protective decision optimization, significantly improving the forward-looking and accuracy of the decision.

[0024] Embodiment 3 The decision optimization and instruction generation unit generates the optimal intervention sequence by constructing and solving a quadratic objective function; the quadratic objective function aims to: Minimize the weighted error between the predicted state vector sequence and the preset target state trajectory; Minimize the control input penalty; Minimize the deviation penalty from the standard process; The decision optimization and instruction generation unit also handles multiple constraints when solving the optimal intervention sequence; the multiple constraints include an upper limit on the total daily physiological load, which must not exceed a preset load threshold; In this embodiment, the optimization method and constraints used by the decision optimization and instruction generation unit to generate the optimal intervention sequence are described in detail; This unit generates the optimal intervention sequence by constructing and solving a quadratic objective function. This objective function aims to find the optimal control strategy within a finite future time domain within the framework of model predictive control. The mathematical form is as follows:

[0025] in, :Indicates that the goal of this formula is to minimize the objective function ; is the objective function to be minimized; : index of the time step; is the predicted future state vector output by the state evolution prediction unit; It is a preset target state trajectory that is predefined by the rehabilitation specialist based on clinical guidelines and the athlete's rehabilitation phase goals; is the sequence of interventions to be solved; It is an intervention sequence corresponding to the standard rehabilitation process obtained from an external knowledge base or standardized rehabilitation program; and are the lengths of the prediction and control horizons, respectively; Both are semi-positive definite symmetric weighted matrices, whose values ​​are preset tuning parameters to balance different optimization objectives; : represents the quadratic norm of the vector, or the weighted quadratic norm, used to quantify the error or cost; Crucially, the elements of these weight matrices are not dimensionless pure numbers, but have specific units to ensure that the objective function All items in have the same, dimensionless cost units, specifically: matrix The element unit is , used to convert the square of the state error Converted into dimensionless cost; matrix and The element unit is , used to square the control input Converted into dimensionless cost; In this way, the objective function The three components of the cost of state error, control input cost, and process deviation cost can be unified in terms of dimension, so as to perform meaningful addition and minimization solutions; B. Optimize the weighted matrices Q, R, S for decision making: These weighting matrices are set based on the focus of the rehabilitation strategy, such as: matrix : Determines the degree of penalty for deviation from the target trajectory. If the primary goal at the current stage is to quickly eliminate inflammation, then The weight corresponding to the error in the inflammation level should be set relatively large; matrix : Determines the degree of penalty for the cost of the intervention. If acupuncture or physical therapy is expensive, The weight corresponding to the intervention should be set high to encourage the system to find lower-cost alternatives; matrix : Determines the degree of punishment for deviation from the standard process. In the early stages of rehabilitation, a higher weights to make the plan closer to the standard plan; in the later stages of rehabilitation, the weights can be appropriately reduced. Weight, giving the system more room for personalized optimization; The quadratic objective function specifically aims to achieve three mutually balanced goals: Minimize the weighted error between the predicted state vector sequence and the preset target state trajectory, which is determined by the first term Implementation: This is the core driver to ensure that the rehabilitation program moves towards the ideal state; Minimize the control input penalty, given by the second term This measure aims to avoid the use of interventions that are too costly or too intensive, reflecting economic and safety considerations; Minimize the penalty for deviation from the standard process, given by the second term Implementation; This ensures that the dynamic plan generated by the system will not deviate indefinitely from the standard process verified by evidence-based medicine, ensuring the reliability of decision-making; When solving the above optimal intervention sequence, the unit also handles multiple constraints; in this embodiment, the key constraint is the upper limit of the total daily physiological load, that is, ensuring that the sum of the physiological costs generated by all intervention measures and training tasks in a day does not exceed the preset load threshold This threshold is a safety upper limit set based on statistical analysis of a large number of athletes' physiological data or authoritative sports physiology guidelines. Its function is to prevent overtraining or secondary injuries caused by excessive load in a single day. The embodiment realizes the scientization and optimization of the decision by constructing the rehabilitation decision as a rigorous, constrained model predictive control optimization problem; instead of being based on isolated rules or experience, the multiple goals of rehabilitation effect, intervention cost, process compliance and physiological safety are systematically balanced in a predictive framework; this method can generate a dynamic intervention sequence that is optimal in comprehensive benefit under the condition of meeting all safety constraints, thereby maximizing the rehabilitation process under the premise of ensuring safety.

[0026] Embodiment 4 The process of the closed-loop correction unit calculating the prediction error is: determining the actual change amount of the state vector after the intervention ends; quantifying the prediction error by comparing the actual change amount with the estimated recovery benefit of the intervention measure stored in the intervention measure cost-benefit knowledge base; The process of the closed-loop correction unit adaptively correcting the system parameters is: using the prediction error as a feedback signal to fine-tune the state transition matrix, input matrix and observation matrix of the dynamic assessment model of the allocable physiological resources; using the prediction error as a feedback signal to fine-tune the network weights of the physiological resource evolution prediction engine; In the embodiment, the working mechanism of the closed-loop correction unit is described, including the calculation method of the prediction error and the adaptive correction process of the system parameters; The process of the closed-loop correction unit calculating the prediction error starts with obtaining the actual multi-modal data after the intervention ends through the data acquisition and synchronization unit after the execution of the periodic dynamic rehabilitation plan, and estimating the actual state vector after the intervention ends by the state representation and assessment unit; determining the actual change amount of the state vector after the intervention ends; comparing the actual change amount with the estimated recovery benefit of the intervention measure stored in the internal intervention measure cost-benefit knowledge base of the system, and quantifying the difference between the two as the prediction error of this time; the intervention measure cost-benefit knowledge base is a structured database that stores prior knowledge such as estimated resource cost and recovery benefit associated with various intervention measures; The construction method of the intervention measure cost-benefit knowledge base can adopt one or a combination of the following methods: Evidence-based medical data entry: systematically extract quantitative data about the impact of specific interventions such as acupuncture at specific acupoints and massage with specific techniques on physiological indicators such as heart rate variability and blood lactate from published clinical research and medical guidelines as initial entries of the knowledge base.

[0027] Quantification of expert experience: Design a structured questionnaire and invite experienced Chinese medicine practitioners or rehabilitation therapists to evaluate the resource costs of different interventions, such as time, consumables, and estimated recovery benefits, such as the degree of improvement in fatigue index. This semi-quantitative expert knowledge can be converted into data in the knowledge base; Online learning and filling: The knowledge base is designed as a dynamic database. During the operation of the system, it will continuously record each intervention The actual observed state change When enough data is accumulated, the cost-benefit values ​​of various intervention measures in the knowledge base can be automatically filled and updated through regression analysis and other methods, thereby achieving self-improvement of the knowledge base; The process of adaptive correction of system parameters by the closed-loop correction unit is as follows: When the calculated prediction error exceeds a preset error threshold The correction process is triggered when the error is greater than 0.05. The threshold can be set based on the statistical distribution of historical forecast data, for example, the upper bound of the 95% confidence interval, to ensure that the correction mechanism is triggered only when statistically significant deviations occur. The unit uses this forecast error as a feedback signal and employs algorithms such as recursive least squares or gradient descent to fine-tune the core model parameters within the system. The unit uses the prediction error as a feedback signal to dynamically evaluate the state transition matrix of the available physiological resources model. , input matrix and the observation matrix Fine-tuning the condition assessment model to more accurately reflect the physiological evolution of a specific athlete and their response to interventions; This unit uses prediction errors as feedback signals to fine-tune the network weights of the physiological resource evolution prediction engine. This enables the attention-based long short-term memory network prediction model to more accurately capture the temporal characteristics of the athlete's state changes. This embodiment establishes a complete closed-loop feedback and correction mechanism, giving the entire system the ability to self-evolve and adapt individually. The system is no longer a static model, but is able to dynamically optimize its internal state assessment model and future prediction model by continuously learning the errors generated by its own decision-making. This adaptive capability enables the system to increasingly fit the uniqueness of individual athletes over time, thereby continuously improving the accuracy of its assessment, prediction, and decision-making, and achieving truly personalized and precise rehabilitation.

[0028] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A multimodal medical information intelligent integration and decision support system for acupuncture rehabilitation, characterized by: include: Data acquisition and synchronization unit, used to collect athletes' physiological data, psychological data and behavioral data; The data acquisition and synchronization unit is also used to perform time stamp alignment processing on the collected physiological data, psychological data and behavioral data to generate a data stream with a global time stamp; a state characterization and evaluation unit for extracting multimodal features based on the data stream generated by the data acquisition and synchronization unit; The state representation and evaluation unit is further configured to process the multimodal features through a preset available physiological resource dynamic evaluation model to estimate a current state vector representing the athlete's current physiological state; A state evolution prediction unit is used to process the historical state vector sequence estimated by the state representation and evaluation unit using a preset physiological resource evolution prediction engine to solve the predicted state vector sequence of future time steps; a decision optimization and instruction generation unit, configured to generate an optimal intervention sequence based on the predicted state vector sequence calculated by the state evolution prediction unit; the decision optimization and instruction generation unit is further configured to parse the optimal intervention sequence into a dynamic rehabilitation plan; A closed-loop correction unit is used to calculate the prediction error after the dynamic rehabilitation plan is executed; the closed-loop correction unit is also used to adaptively correct the system parameters of the available physiological resource dynamic assessment model and the physiological resource evolution prediction engine in response to the prediction error exceeding a preset error threshold.

2. The multimodal medical information intelligent integration and decision support system for acupuncture rehabilitation according to claim 1 is characterized in that: The physiological data collected by the data acquisition and synchronization unit include heart rate variability data, galvanic skin response data, tongue image data and pulse waveform data; the behavioral data include biochemical indicators and subjective recovery perception data.

3. The multimodal medical information intelligent integration and decision support system for acupuncture rehabilitation according to claim 1 is characterized in that: The process of extracting multimodal features by the state representation and evaluation unit is as follows: Determine the statistical characteristics and frequency domain power distribution characteristics of heart rate variability data and generate HRV feature vectors; A preset convolutional neural network is used to process tongue image data to extract quantitative feature vectors; The pulse waveform data is analyzed using wavelet transform to extract the pulse characteristic vector.

4. The multimodal medical information intelligent integration and decision support system for acupuncture rehabilitation according to claim 1 is characterized in that: The dynamic assessment model of available physiological resources is a linear Gaussian state space model; the current state vector includes available physiological resources, physiological fatigue index, and inflammation level; the state representation and assessment unit uses a Kalman filter algorithm to combine the state at the previous moment with the current observation vector composed of multimodal features to estimate the current state vector.

5. The multimodal medical information intelligent integration and decision support system for acupuncture rehabilitation according to claim 1 is characterized in that: The physiological resource evolution prediction engine is a long short-term memory network based on the attention mechanism; the input of the state evolution prediction unit includes the historical state vector sequence and the known future input sequence; the attention mechanism is used to assign weights to events in the historical sequence when predicting future states.

6. The multimodal medical information intelligent integration and decision support system for acupuncture rehabilitation according to claim 1 is characterized in that: The decision optimization and instruction generation unit generates the optimal intervention sequence by constructing and solving a quadratic objective function; the quadratic objective function aims to: Minimize the weighted error between the predicted state vector sequence and the preset target state trajectory; Minimize control input penalties; Minimize penalties for deviations from standard procedures.

7. The multimodal medical information intelligent integration and decision support system for acupuncture rehabilitation according to claim 6 is characterized in that: The decision optimization and instruction generation unit also processes multiple constraints when solving the optimal intervention sequence; the multiple constraints include an upper limit on the total daily physiological load, which does not exceed a preset load threshold.

8. The multimodal medical information intelligent integration and decision support system for acupuncture rehabilitation according to claim 1 is characterized in that: The process of closed-loop correction unit calculating prediction error is: Determine the actual change in the state vector after the intervention ends; The prediction error is quantified based on the actual change and compared with the estimated restoration benefits stored in the knowledge base of the cost-effectiveness of the intervention.

9. The multimodal medical information intelligent integration and decision support system for acupuncture rehabilitation according to claim 1 is characterized in that: The process of adaptive correction of system parameters by the closed-loop correction unit is as follows: The prediction error is used as a feedback signal to fine-tune the state transfer matrix, input matrix, and observation matrix of the dynamic evaluation model of available physiological resources. The prediction error is used as a feedback signal to fine-tune the network weights of the physiological resource evolution prediction engine.

Images