Acupuncture manipulation information intelligent perception and regulation method based on multi-modal brain-computer interface

Abstract

The application provides an acupuncture manipulation information intelligent perception and regulation method based on a multi-modal brain-computer interface, adopts advanced machine learning models such as multi-modal Transformers based on an attention mechanism, performs dynamic feature fusion and deep analysis on heterogeneous time sequence data, can intelligently extract and quantify a core parameter combination representing a'receiving qi' state, and constructs a clear'manipulation parameter-physiological effect-brain response mode' mapping atlas. The system can not only objectively identify and evaluate the'receiving qi' state and its intensity, but also provide dynamic manipulation adjustment suggestions for the operator or fine-tune the future driving auxiliary equipment based on real-time analysis results through a closed-loop regulation logic, so as to realize the paradigm shift from experiential operation to digital, visual and intelligent perception and regulation, and lay a solid technical foundation for the standardized teaching, precise treatment and repeatable scientific research of acupuncture manipulation.

Description

Technical Field

[0001] This invention relates to the field of acupuncture technology, and in particular to a method for intelligent perception and control of acupuncture technique information based on a multimodal brain-computer interface. Background Technology

[0002] Acupuncture, a treasure of traditional Chinese medicine, relies on the core principle of "deqi" (the arrival of qi) for its therapeutic effect, as emphasized in the *Ling Shu* (Spiritual Pivot), Chapter 9, "Nine Needles and Twelve Origins," which states that "the key to needling is the arrival of qi for effectiveness." "Deqi" is a complex and highly subjective sensation, manifesting as a feeling of "heaviness and tightness under the hand" for the practitioner and sensations of "soreness, numbness, distension, and heaviness" for the patient. However, for thousands of years, the transmission of acupuncture techniques and the evaluation of their efficacy have heavily relied on the practitioner's personal experience and the patient's vague descriptions. This has led to three major challenges: First, the lack of objective and unified quantitative standards for the unique techniques of different practitioners makes it impossible to accurately reproduce the techniques and maintain stable therapeutic effects. Second, the physiological and biochemical parameters of the core state of "deqi" have not been fully and synchronously revealed and correlated, resulting in an ambiguity in the scientific mechanism between "technique and effect." Finally, the experience-based transmission model makes it difficult to standardize and promote excellent techniques, limiting the modernization and precision development of acupuncture.

[0003] To overcome the aforementioned challenges, modern research has attempted to introduce engineering technology into the field of acupuncture. Existing technological solutions mainly focus on two relatively independent levels: On the one hand, at the level of technique quantification, researchers have developed acupuncture technique parameter analyzers or "sensing needles" integrating force / torque sensors to record the force, torque, and motion parameters experienced by the needle, attempting to convert the manipulation into electrical signals for objective description (e.g., the work of Yang Huayuan, Ding Guanghong, Li Qinghua, etc.). On the other hand, at the level of effect detection, research focuses on the impact of acupuncture on the local microenvironment of acupoints, such as using needle-type electrodes to dynamically monitor changes in the concentration of biochemical indicators such as calcium ions and pH values ​​at acupoints before and after acupuncture (e.g., the research of Guo Yi, Ren Ning, etc.), or attempting to explore the association between acupuncture and the activation of specific brain regions. In recent years, with the development of artificial intelligence and brain-computer interface technology, some solutions have been proposed to use neural signals such as electroencephalogram (EEG) combined with machine learning models to evaluate the effects of acupuncture treatment or recommend treatment plans (such as the one disclosed in patent application CN202211600862.9, which collects brain-computer interface signals during the treatment process and establishes a deep learning model to select or adjust acupuncture treatment plans).

[0004] However, existing technical solutions still have significant limitations and unresolved technical problems: First, most existing studies are "single-modal" or "dual-modal," meaning they either focus only on the mechanical parameters of the manipulation or only on the local biochemical signals or central nervous system responses at acupoints, lacking high-precision, time-synchronized multi-modal fusion acquisition of the practitioner's mechanical parameters, the patient's peripheral biological signals at acupoints (temperature, pressure, ion concentration, etc.), and the patient's central nervous system responses (EEG, etc.). This fragmentation makes it impossible to construct a dynamic correlation map between "manipulation-peripheral response-central response," thus failing to truly reveal the entire chain and objective physiological essence of "deqi" (the sensation of qi). Second, even if some solutions mention the use of brain signals, their purpose is mostly limited to macroscopic "efficacy assessment" or "treatment recommendation," without delving into using brain-computer interface signals to provide real-time feedback and intelligent perception of the practitioner's specific manipulation operations (such as the angle and frequency of twisting), failing to form a real-time closed loop of "perception-decision-control," and unable to guide the optimization of the manipulation at the moment of operation. Third, existing data processing models are relatively simple or have a single purpose, lacking dynamic feature fusion and intelligent analysis models specifically designed for the aforementioned multimodal, high-dimensional, and time-series data, in order to extract the core parameter combination that can characterize the "qi" state from complex data and achieve real-time prediction and adaptive control of the effects of techniques.

[0005] Therefore, the urgent technical problem to be solved in this field is: how to construct an integrated system that can synchronously and in real time collect multi-dimensional signals from the entire chain of acupuncture operations, and perform dynamic fusion analysis through advanced artificial intelligence models, so as to achieve intelligent perception, quantitative evaluation of the "deqi" state, and real-time guidance or adaptive control of acupuncture techniques, thereby transforming acupuncture from a "fuzzy art" that relies on personal experience into a "precise science" based on objective data. Summary of the Invention

[0006] The purpose of this invention is to provide a method for intelligent perception and control of acupuncture technique information based on a multimodal brain-computer interface, so as to solve the problems existing in the prior art.

[0007] To achieve the above objectives, the present invention provides the following solution: This invention provides a method for intelligent perception and control of acupuncture technique information based on a multimodal brain-computer interface, comprising the following steps: S1. Synchronously collect multimodal time-series data during the acupuncture operation. The multimodal time-series data includes at least: time-series data of mechanical and kinematic parameters generated when the practitioner performs acupuncture techniques, time-series data of at least one peripheral biological signal at the acupuncture point of the patient, and time-series data of the patient's central nervous system response signal. S2. The multimodal time-series data collected in step S1 is preprocessed and features are extracted to obtain the manipulation feature vector, peripheral biological feature vector and central nervous system response feature vector, respectively. S3. Input the technique feature vector, peripheral biological feature vector and central nervous system response feature vector into the trained multimodal dynamic fusion model. The multimodal dynamic fusion model calculates the weight of each modality feature at each time step based on the attention mechanism and performs weighted fusion to output the fused dynamic feature sequence. S4. Based on the dynamic feature sequence, the patient's acupuncture "deqi" state is identified and quantified in real time using a trained intelligent perception model. S5. Based on the "Deqi" state identified in step S4, generate and output adjustment instructions for the current acupuncture technique.

[0008] Preferably, in step S1: The time-series data of the mechanical and kinematic parameters are acquired through a six-dimensional force / torque sensor and inertial measurement unit integrated into the needle handle, including at least axial force, torque, twist angle, lifting and inserting displacement, and operating frequency; The peripheral biological signal time-series data is collected by a flexible microarray sensing patch placed on the skin surface of the acupoint, and includes at least one of temperature, tissue pressure, calcium ion concentration and pH value. The central nervous system response signal timing data were acquired using high-density electroencephalography (EEG) equipment or functional near-infrared spectroscopy (FIR) imaging equipment, and at least covered brain region signals related to somatosensory and pain processing.

[0009] Preferably, in step S1, all data acquisition devices are controlled by a unified master clock to ensure that the time-series data of mechanical and kinematic parameters, peripheral biological signals, and central nervous system response signals have strictly aligned timestamps.

[0010] Preferably, in step S3, the multimodal dynamic fusion model is a Transformer-based model, which dynamically learns the correlation between different modal features through a self-attention layer and calculates the correlation between each modal feature at time step S3. The attention weights, and the fused dynamic features are calculated by the following formula: ; in, , , They are respectively The feature vectors of techniques, peripheral biological features, and central nervous system response at different times.

[0011] Preferably, in step S4, the intelligent perception model is a recurrent neural network or a temporal convolutional network, whose input is the dynamic feature sequence within a time window, and whose output includes: Discrete classification results of the "gaining qi" state; and / or The continuous regression value of the intensity of "deqi".

[0012] Preferably, the method further includes step S6: training a "manipulation parameter-effect" mapping model based on historically collected multimodal time-series data and labeled "deqi" state tags; in step S5, the generation of the adjustment instruction is at least partially based on the recommendation results of the "manipulation parameter-effect" mapping model, and the recommendation results include a combination of acupuncture manipulation parameters to be adjusted in order to achieve or maintain the target "deqi" effect.

[0013] Preferably, the "manipulation parameter-effect" mapping model is a graph neural network model, which constructs a heterogeneous graph by combining manipulation parameter nodes, peripheral biological signal nodes, brain region response nodes and their interrelationships. It predicts the effect characteristics produced under a specific combination of manipulation parameters by learning the embedding representation of the nodes, or recommends manipulation parameter combinations in reverse based on the input expected effect characteristics.

[0014] This invention also provides an intelligent sensing and control system for acupuncture manipulation information based on a multimodal brain-computer interface, comprising: A multimodal synchronous acquisition module is used to perform step S1, and it includes at least an intelligent needle handle sensing unit, an acupoint peripheral signal sensing unit, and a brain-computer interface signal acquisition unit. The data processing and fusion analysis module is used to execute steps S2, S3 and S4, and includes a processor storing executable instructions, which, when executed, implement the preprocessing, feature extraction, dynamic fusion and intelligent perception functions. The intelligent decision-making and control output module is used to execute step S5, generate control commands based on the output of the intelligent sensing model, and output them through a human-computer interaction interface or an actuator.

[0015] Preferably, the intelligent needle handle sensing unit is encapsulated with a six-dimensional force / torque sensor and an inertial measurement unit; the acupoint peripheral signal sensing unit is a flexible patch integrating at least one of a temperature sensor, a pressure sensor, and an electrochemical sensor.

[0016] The present invention also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a method for intelligent perception and control of acupuncture technique information based on a multimodal brain-computer interface.

[0017] The present invention achieves the following beneficial technical effects compared to the prior art: This invention provides an intelligent perception and control method for acupuncture manipulation information based on a multimodal brain-computer interface. By innovatively integrating a multimodal synchronous acquisition module (including an intelligent needle handle sensor for capturing the mechanical parameters of the practitioner's manipulation techniques such as lifting, thrusting, and twisting; a flexible sensing patch for monitoring peripheral biological signals such as temperature, pressure, and ion concentration at the patient's acupoints; and a brain-computer interface device for acquiring central nervous system responses such as high-density EEG / near-infrared spectroscopy), it achieves for the first time the complete capture of the entire chain of acupuncture information—"manipulation-periphery-central"—under strict time synchronization. Based on this, the system employs advanced machine learning models such as the attention-based multimodal Transformer to dynamically fuse and deeply analyze heterogeneous temporal data. It can intelligently extract and quantify the core parameter combinations representing the "deqi" state, constructing a clear mapping atlas of "manipulation parameters-physiological effects-brain response patterns." In particular, this system can not only objectively identify and assess the "deqi" state and its intensity, but also provide practitioners with dynamic technique adjustment suggestions or future drive auxiliary equipment for fine-tuning based on real-time analysis results and closed-loop control logic. This realizes a paradigm shift from experience-based operation to digital, visual, and intelligent perception and control, laying a solid technical foundation for standardized teaching, precision treatment, and reproducible scientific research of acupuncture techniques. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart of the intelligent perception and control method for acupuncture manipulation information based on a multimodal brain-computer interface provided by the present invention; Figure 2 The system architecture diagram of the intelligent perception and control system for acupuncture manipulation information based on multimodal brain-computer interface provided by the present invention is shown. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The purpose of this invention is to provide a method for intelligent perception and control of acupuncture technique information based on a multimodal brain-computer interface, so as to solve the problems existing in the prior art.

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0023] Example 1: Please see Figure 1 This invention demonstrates the overall process of the intelligent perception and control method for acupuncture manipulation information based on a multimodal brain-computer interface. The core of this invention lies in constructing a real-time closed-loop system of "acquisition-fusion-perception-control," which for the first time performs high-precision synchronous acquisition and intelligent correlation analysis of the practitioner's manipulation techniques, the microscopic biochemical and physical changes at the patient's acupoints, and the neural responses of the brain's central nervous system. This enables the objective quantification, real-time perception, and dynamic control of the core acupuncture effect of "deqi" (the attainment of qi).

[0024] The first step is the synchronous acquisition of multimodal time-series data. This step forms the data foundation for subsequent intelligent analysis, and the key lies in "synchronization" and "multidimensionality." In practice, the system coordinates all acquisition devices through a unified high-precision master clock signal, ensuring that all data streams have strictly aligned timestamps, laying the foundation for millisecond-level event correlation across modalities. The acquisition work is completed by three parallel subsystems. The first is an intelligent needle handle sensing unit used to capture the practitioner's manipulation techniques. This unit integrates a miniaturized six-dimensional force / torque sensor and a nine-axis inertial measurement unit (IMU) within a specially designed needle handle, forming a rigid connection with the needle body. When the practitioner performs operations such as lifting, inserting, twisting, and rotating, the sensor collects in real time the axial force on the needle body, the torque around the needle axis, and the attitude angle and angular velocity data from the nine-axis IMU. Through integration and filtering of the nine-axis IMU data, the precise twisting angle, lifting and inserting displacement, and the frequency of the manipulation technique can be further calculated. Ultimately, the manipulation technique is converted into a time series. Secondly, there is a flexible acupoint sensing patch for monitoring the local response (peripheral response) of acupoints in patients. This patch uses a biocompatible flexible material with a pre-drilled needle hole in the center, allowing it to adhere tightly to the skin surface around the acupuncture point. The patch integrates a miniature temperature sensor, a micro-piezoresistive pressure sensor, and a microelectrode array based on a specific ion-selective membrane. These sensors can synchronously and in real-time acquire tissue temperature, interstitial fluid pressure, and crucial calcium and hydrogen ion concentrations around the acupuncture point, forming temporal data of peripheral biological signals. Thirdly, there is a brain-computer interface signal acquisition unit for capturing the response of the patient's central nervous system. This embodiment preferably uses a high-density wireless electroencephalography (EEG) system, with electrodes arranged according to the international 10-10 system, focusing on covering brain regions related to somatosensory, pain, and visceral sensory processing, such as the primary somatosensory cortex, anterior cingulate cortex, and insula. Simultaneously, a functional near-infrared spectroscopy imaging device can be used to synchronously acquire blood oxygen dynamics signals from the aforementioned brain regions. Thus, the central nervous system response signal is recorded as a multi-channel time series.

[0025] The second step is data preprocessing and feature extraction. The raw acquired signals contain noise and vary in dimension and scale, requiring preprocessing to extract key features that effectively represent the information. For manual manipulation signals and peripheral biological signals, preprocessing mainly includes baseline drift removal and bandpass filtering to remove power line interference and non-physiological high-frequency noise. For EEG signals, a more complex preprocessing pipeline is required, including bandpass filtering, EEG and EMG artifact removal based on independent component analysis, and rereference. After preprocessing, the system uses a sliding time window approach to extract features from each modality of data. For manual manipulation data, time-domain statistical features, frequency-domain features, and nonlinear dynamic features are extracted within each window. For peripheral biological signals, the main extraction methods are the rate of change, cumulative change, and instantaneous value relative to the pre-acupuncture baseline. For EEG signals, feature extraction is more comprehensive: in the time domain, the amplitude and latency of event-related potentials can be extracted; in the frequency domain, the relative or absolute power of each frequency band can be calculated; in the spatial domain, a dynamic brain functional connectivity network can be constructed by calculating the phase lock value or Granger causality value between different electrode pairs, thereby extracting the network's topological properties, such as clustering coefficients and feature path lengths. Finally, each time window is transformed into three feature vectors: a manipulation feature vector, a peripheral biological feature vector, and a central nervous system response feature vector.

[0026] The third step is multimodal dynamic feature fusion based on an attention mechanism. This is one of the key innovations of this invention, aiming to solve the problem of how to intelligently weigh the relative importance of different modal information in judging the "qi" state over time. The system uses a multimodal dynamic fusion model based on the Transformer architecture to accomplish this task. The core of this model is a self-attention layer, which receives the concatenated multimodal feature sequence as input. Specifically, at each time t, the model first concatenates the three feature vectors into a comprehensive vector. Then, it maps this vector to a set of initial attention scores through a learnable linear transformation. These scores are then normalized using a Softmax function to obtain the attention weight vectors of each modality at that time. This process can be formally represented as: ; in, The weight matrix is ​​a learnable matrix. For a moment The mechanical characteristic vector of the technique, For a moment The peripheral biological feature vector of acupoints, For a moment The central nervous system response feature vector, For learnable bias terms; The weights are not fixed, but dynamically calculated by the model based on the feature content of all current modalities. For example, in the initial stage of acupuncture, mechanical features may have higher weights; while when the "deqi" sensation occurs, the weights of EEG response features may increase significantly. Finally, the calculated weights are used to sum the original feature vectors to generate a fused dynamic feature vector. ; in, , , They are respectively The feature vectors of techniques, peripheral biological features, and central nervous system responses at different times; , , The weight scalars calculated by Formula 1 correspond to the weights of techniques, peripheral features, and central features, respectively. This dynamic feature vector is a unified representation that can dynamically reflect the strength of the "manipulation-periphery-central" correlation, and serves as the input to the downstream intelligent perception model.

[0027] The fourth step is the intelligent perception and quantification of the "Deqi" state. The goal of this step is to use the fused dynamic feature sequence to determine in real time whether the patient has entered the "Deqi" state and its intensity. The system uses a trained recurrent neural network or temporal convolutional network as the intelligent perception model. This model takes a sequence within a time window as input. The intelligent perception model can effectively capture the temporal dependencies of features during the occurrence and development of "Deqi". The model's output layer is designed with a multi-task learning architecture: one output branch provides a discrete classification result of the "Deqi" state through a Softmax function, such as the probability distribution of "no Deqi", "mild Deqi", and "significant Deqi"; the other output branch provides a continuous regression value of "Deqi" intensity through a linear layer, with higher values ​​indicating stronger "Deqi" sensation. The training of this model relies on a high-quality dataset, which is generated in real time by experienced acupuncturists who simultaneously collect data and annotate it based on the patient's complaints and the practitioner's own "tactile sensations".

[0028] The fifth step is intelligent decision-making and closed-loop control. This step achieves a closed loop from perception to action, representing the ultimate manifestation of system intelligence. Control decisions are based on two levels. The first level is direct reactive control: the system monitors the intensity of the "deqi" (a sensation of energy attainment) output by the intelligent perception model in real time. When the intensity value falls below the preset target threshold or shows a downward trend during treatment, the system immediately triggers the control mechanism. The second level is predictive optimization control: a "manipulation parameter-effect" mapping model is pre-set within the system or invoked online. This model is preferably constructed using a graph neural network, which uses different manipulation parameters, peripheral response indicators, and brain region activation patterns as nodes in the graph, learning the complex relationships between nodes through historical data. When control is needed, the system can input the current state into this model, and the model can infer the most likely adjustment suggestion for the acupuncture manipulation parameters to achieve that state, such as "increasing the twisting frequency by 15%" or "changing to a small-amplitude lifting and thrusting technique." The final control instructions are presented to the practitioner in real time through a user-friendly human-computer interaction interface, such as visual cues on a tablet or gentle vibration feedback generated by the micro-tactile motor of the intelligent needle handle, guiding the practitioner to adjust the manipulation technique. In future, more automated implementations, this instruction can directly drive a parallel robotic arm to perform sub-millimeter-level precise motion adjustments on the needle handle, thereby achieving automatic optimization of manipulation parameters.

[0029] Example 2: Please see Figure 2 The system architecture for implementing the aforementioned methods is demonstrated. This system primarily comprises: a multimodal synchronous acquisition module, integrating the aforementioned smart needle handle, flexible acupoint patch, and high-density EEG cap; a data processing and fusion analysis module, hosted by a high-performance embedded processor or server, running the aforementioned preprocessing, feature extraction, fusion model, and perception model algorithms; and an intelligent decision-making and control output module, responsible for generating and transmitting control commands. All modules interact with each other via a network or internal bus, collaboratively completing the entire process from signal acquisition to intelligent control.

[0030] In summary, this invention, through the specific embodiments described above, constructs a complete technical solution combining hardware and software. It goes beyond simply collecting multiple signals; rather, through an innovative synchronous acquisition scheme, dynamic weighted fusion algorithm, and closed-loop control logic, it achieves for the first time a deep digital analysis and real-time interactive guidance of the "deqi" process during acupuncture. This provides a revolutionary tool for the standardization of acupuncture therapy, the visualization of teaching, and the objective evaluation of therapeutic effects, enabling the wisdom and experience of traditional medicine to be interpreted, inherited, and optimized using the language of modern technology.

[0031] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0032] It should be noted that the components mentioned in the above embodiments are all general standard parts or components known to those skilled in the art. Their structures and principles can be learned by those skilled in the art through technical manuals or conventional experimental methods.

[0033] This invention has illustrated its principles and implementation methods using specific examples. The descriptions of these embodiments are merely illustrative of the method and its core ideas; furthermore, those skilled in the art will recognize that modifications may be made to the specific implementation methods and application scope based on the principles of this invention. Therefore, the content of this specification should not be construed as limiting the invention.

Claims

1. A method for intelligent perception and control of acupuncture technique information based on a multimodal brain-computer interface, characterized in that, Includes the following steps: S1. Synchronously collect multimodal time-series data during the acupuncture operation. The multimodal time-series data includes at least: time-series data of mechanical and kinematic parameters generated when the practitioner performs acupuncture techniques, time-series data of at least one peripheral biological signal at the acupuncture point of the patient, and time-series data of the patient's central nervous system response signal. S2. The multimodal time-series data collected in step S1 is preprocessed and features are extracted to obtain the manipulation feature vector, peripheral biological feature vector and central nervous system response feature vector, respectively. S3. Input the technique feature vector, peripheral biological feature vector and central nervous system response feature vector into the trained multimodal dynamic fusion model. The multimodal dynamic fusion model calculates the weight of each modality feature at each time step based on the attention mechanism and performs weighted fusion to output the fused dynamic feature sequence. S4. Based on the dynamic feature sequence, the patient's acupuncture "deqi" state is identified and quantified in real time using a trained intelligent perception model. S5. Based on the "Deqi" state identified in step S4, generate and output adjustment instructions for the current acupuncture technique.

2. The method for intelligent perception and control of acupuncture technique information based on a multimodal brain-computer interface according to claim 1, characterized in that, In step S1: The time-series data of the mechanical and kinematic parameters are acquired through a six-dimensional force / torque sensor and inertial measurement unit integrated into the needle handle, including at least axial force, torque, twist angle, lifting and inserting displacement, and operating frequency; The peripheral biological signal time-series data is collected by a flexible microarray sensing patch placed on the skin surface of the acupoint, and includes at least one of temperature, tissue pressure, calcium ion concentration and pH value. The central nervous system response signal timing data were acquired using high-density electroencephalography (EEG) equipment or functional near-infrared spectroscopy (FIR) imaging equipment, and at least covered brain region signals related to somatosensory and pain processing.

3. The method for intelligent perception and control of acupuncture technique information based on a multimodal brain-computer interface according to claim 1, characterized in that, In step S1, all data acquisition devices are controlled by a unified master clock to ensure that the time-series data of mechanical and kinematic parameters, peripheral biological signals, and central nervous system response signals have strictly aligned timestamps.

4. The method for intelligent perception and control of acupuncture technique information based on a multimodal brain-computer interface according to claim 1, characterized in that, In step S3, the multimodal dynamic fusion model is a Transformer-based model, which dynamically learns the correlation between different modal features through a self-attention layer and calculates the correlation between each modal feature at time step S3. The attention weights, and the fused dynamic features are calculated by the following formula: ； in, , , They are respectively The feature vectors of techniques, peripheral biological features, and central nervous system response at different times.

5. The method for intelligent perception and control of acupuncture technique information based on a multimodal brain-computer interface according to claim 1, characterized in that, In step S4, the intelligent perception model is a recurrent neural network or a temporal convolutional network, whose input is the dynamic feature sequence within a time window, and whose output includes: Discrete classification results of the "Gaining Qi" state; and / or The continuous regression value of the intensity of "deqi".

6. The method for intelligent perception and control of acupuncture technique information based on a multimodal brain-computer interface according to claim 5, characterized in that, It also includes step S6: training a "manipulation parameter-effect" mapping model based on historically collected multimodal time-series data and labeled "deqi" state labels; in step S5, the generation of the adjustment instruction is at least partially based on the recommendation results of the "manipulation parameter-effect" mapping model, and the recommendation results include a combination of acupuncture manipulation parameters to be adjusted in order to achieve or maintain the target "deqi" effect.

7. The method for intelligent perception and control of acupuncture technique information based on a multimodal brain-computer interface according to claim 6, characterized in that, The "manipulation parameter-effect" mapping model is a graph neural network model, which constructs a heterogeneous graph by combining manipulation parameter nodes, peripheral biological signal nodes, brain region response nodes and their interrelationships. It predicts the effect characteristics produced under a specific combination of manipulation parameters by learning the embedding representation of the nodes, or recommends manipulation parameter combinations in reverse based on the input expected effect characteristics.

8. A system for intelligent sensing and control of acupuncture technique information based on a multimodal brain-computer interface for implementing the intelligent sensing and control method of acupuncture technique information based on a multimodal brain-computer interface as described in any one of claims 1-7, characterized in that, include: A multimodal synchronous acquisition module is used to perform step S1, and it includes at least an intelligent needle handle sensing unit, an acupoint peripheral signal sensing unit, and a brain-computer interface signal acquisition unit. The data processing and fusion analysis module is used to execute steps S2, S3 and S4, and includes a processor storing executable instructions, which, when executed, implement the preprocessing, feature extraction, dynamic fusion and intelligent perception functions. The intelligent decision-making and control output module is used to execute step S5, generate control commands based on the output of the intelligent sensing model, and output them through a human-computer interaction interface or an actuator.

9. The intelligent perception and control system for acupuncture manipulation information based on a multimodal brain-computer interface according to claim 8, characterized in that, The intelligent needle handle sensing unit encapsulates a six-dimensional force / torque sensor and an inertial measurement unit; the acupoint peripheral signal sensing unit is a flexible patch integrating at least one of a temperature sensor, a pressure sensor, and an electrochemical sensor.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the intelligent perception and control method for acupuncture manipulation information based on a multimodal brain-computer interface as described in any one of claims 1 to 7.

Images