Proprioception-based brain-computer interface closed-loop rehabilitation training system and method

Abstract

The application discloses a kind of based on proprioception guide's brain-computer interface closed loop rehabilitation training system and method, it is related to brain-computer interface technical field, to solve the problem that existing fusion stimulation's brain-computer interface system lacks the individualized rehabilitation training method of stage, guiding, and with promoting central nervous remodeling as basic goal in rehabilitation training, system includes stimulation module and the central control module being connected with stimulation module, the stimulation module applies external stimulation to target muscle group;The central control module is respectively connected with intention decoding module and multi-modal feedback module, the intention decoding module judges whether motion intention characteristic meets standard according to built-in algorithm model;The intention decoding module is connected with electroencephalogram acquisition and processing module, and the electroencephalogram acquisition and processing module real-time acquisition and pre-processes electroencephalogram signal;The technical scheme of the application can enhance the authenticity and effectiveness of motor imagination, promote neural function remodeling.

Description

Technical Field

[0001] This invention relates to the field of brain-computer interface technology, specifically to a closed-loop rehabilitation training system and method based on proprioception-guided brain-computer interface. Background Technology

[0002] Motor imagery therapy is a commonly used neurorehabilitation method that guides patients to repeatedly simulate specific movements in their minds, promoting the reorganization of motor cortex function and thus aiding in the recovery of motor function. However, for patients with preserved sensory function but impaired motor output, the effectiveness of motor imagery is often greatly reduced due to the lack of real sensory input, leading to vague imagery, poor concentration, and ineffective training. Current technologies, such as functional electrical stimulation and vibrational stimulation, are used to induce muscle contraction or motor sensation, but these are often directly combined with the patient's active attempts, lacking fine guidance and separation of sensory input. Furthermore, traditional brain-computer interface rehabilitation systems typically directly detect motor intentions and trigger feedback, lacking a systematic training protocol on how to utilize the patient's residual proprioceptive function to "calibrate" and "enhance" the motor imagery process.

[0003] In recent years, several methods incorporating external stimuli have emerged to enhance the characteristics of motor imagery EEG signals. For example, Chinese invention patent application CN121101607A proposes a motor imagery brain-computer interface rehabilitation method incorporating upper limb muscle vibration stimulation. This method synchronously and continuously applies vibration stimulation to both upper limbs while the subject performs a left-hand or right-hand motor imagery task, using the stimulation frequency as prior information to construct a spatial filter for the EEG signal. However, this type of method still has the following limitations: First, it primarily uses external stimulation as a supplementary enhancement, completely overlapping with the patient's active attempts in time, without separating and guiding the neural processes of "sensory input" and "motor intention generation." Second, the core objective of this method is to optimize the immediate brain state decoding performance to control peripheral devices, rather than designing progressive rehabilitation training steps for neural function remodeling. For patients with intact sensory function but impaired motor pathways, they need a training paradigm that can utilize residual sensation to accurately calibrate and gradually strengthen their intrinsic motor command generation ability, rather than simply providing supplementary sensory noise to improve the signal-to-noise ratio. Existing brain-computer interface systems that incorporate fusion stimulation lack a phased, guided rehabilitation training method that aims to promote central nervous system remodeling. Summary of the Invention

[0004] This invention addresses the problem that existing brain-computer interface systems with fusion stimulation lack a phased, guided, and personalized rehabilitation training method that aims to promote central nervous system remodeling. It proposes a closed-loop rehabilitation training system and method based on proprioception guidance, which enhances the realism and effectiveness of motor imagery and promotes neural function remodeling.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a brain-computer interface closed-loop rehabilitation training system based on proprioception guidance, comprising a stimulation module and a central control module connected to the stimulation module, wherein the stimulation module applies external stimulation to the target muscle group; the central control module is respectively connected to an intention decoding module and a multimodal feedback module, wherein the intention decoding module determines whether the motor intention characteristics meet the standard according to a built-in algorithm model; the intention decoding module is connected to an EEG acquisition and processing module, wherein the EEG acquisition and processing module acquires and preprocesses EEG signals in real time.

[0006] The technical solution of this invention mainly includes a stimulation module, an EEG acquisition and processing module, an intention decoding module, a central control module, and a multimodal feedback module. Through a step-by-step guidance protocol, external sensory stimulation is organically combined with active rehabilitation training via brain-computer interface. Brain-computer interface technology is used to analyze the patient's motor intention and attention state, forming a closed loop of "sensory input - central integration - intention recognition - feedback reinforcement". This aims to enhance the realism and effectiveness of motor imagery and promote neural function remodeling.

[0007] The present invention is further configured such that the stimulation module includes a sensory stimulation module and a motor stimulation module, wherein the sensory stimulation module is capable of applying external stimulation to generate a feeling of muscle contraction in active movement; and the motor stimulation module is capable of applying external stimulation to generate actual limb and muscle movement.

[0008] In this technical solution, the stimulation module can be two separate modules, or it can be the same stimulation module that releases different frequencies and intensities to achieve the purpose of generating sensation and movement.

[0009] The present invention is further configured such that the central control module executes a predetermined step-by-step guidance process, provides sensory stimulation and multimodal feedback, and controls the start-up and shutdown and parameters of the sensory stimulation module; and provides motor stimulation and multimodal feedback to the patient based on the output of the intention decoding module, and controls the start-up and shutdown and parameters of the motor stimulation module.

[0010] The present invention is further configured such that the intention decoding module includes a first computing unit, a second computing unit, and a third computing unit, wherein the first computing unit performs EEG signal decoding in the perception training stage; the second computing unit performs EEG signal decoding and comparison in the sensory guidance stage; and the third computing unit performs EEG signal decoding and comparison in the imagination control stage.

[0011] In this technical solution, the first arithmetic unit is connected to the second arithmetic unit and the third arithmetic unit respectively, and the second arithmetic unit is connected to the third arithmetic unit.

[0012] The present invention is further configured such that the multimodal feedback module includes a visual feedback unit and an auditory feedback unit, wherein the visual feedback unit utilizes the influence of limb movements provided by the computer and presents dynamic movements using virtual reality technology; and the auditory feedback unit provides movement language guidance.

[0013] In this technical solution, the visual feedback unit includes images of limb movements, model examples, and expert guidance provided by a computer, or it can use virtual reality technology to present the dynamic movement process; the auditory feedback unit includes voice guidance for movement provided by the system, prompts for starting and stopping, and feedback on success or failure.

[0014] The present invention is further configured to include a multi-lead EEG cap, an amplifier, and a filter in the EEG acquisition and processing module, and to perform EEG signal acquisition and preprocessing processes when the EEG acquisition and processing module performs different tasks.

[0015] A proprioceptive-guided brain-computer interface closed-loop rehabilitation training method, applicable to the aforementioned proprioceptive-guided brain-computer interface closed-loop rehabilitation training system, comprising: During the sensory training phase, stimulation is applied to the target muscle groups, and an individualized sensory evoked EEG model and individualized threshold are obtained based on EEG signals. During the sensory guidance phase, the first motor intention feature is analyzed; by comparing it with a set threshold or model, it is determined whether the first motor intention feature meets the activation criteria. During the imagination control phase, the characteristics of the second motor intention are analyzed; it is then determined whether the characteristics of the second motor intention meet the activation criteria.

[0016] The training process of this technical solution mainly includes a perception training stage, a sensory guidance stage, and an imagery control stage. The perception training stage enables the construction of an individualized sensory evoked EEG model and corresponding individualized thresholds. The sensory guidance stage enables patients to attempt to perform the sensations experienced in the perception training stage and determine whether the analyzed first motor intention feature meets the activation criteria. The imagery control stage involves motor imagery and motor attempts based on the established muscle awareness, and determines whether the analyzed second motor intention feature meets the activation criteria.

[0017] The present invention is further configured such that the perception training phase includes: Sensory stimulation is applied, and training is initiated after passing the stimulation test. EEG signals are collected while the stimulation is being experienced, and the EEG signals at this stage are saved as individualized baseline data and individualized thresholds are determined.

[0018] In this technical solution, the stimulation test determines whether the stimulation intensity is appropriate, and if it is not appropriate, the stimulation intensity is adjusted.

[0019] The present invention is further configured such that the sensory guidance stage includes: after the sensory stimulation ends, within a set time window, guiding the patient to try to make the induced action or contract the target muscle, analyzing the EEG signal within this time window in real time, determining whether the first motor intention feature has reached the activation standard, and if so, applying motor stimulation.

[0020] The present invention is further configured such that the imagination control stage includes: directly prompting the patient to autonomously perform motor imagination based on the memory of the previous stage without giving prior sensory stimulation; assessing the degree of motor imagination or attempt through EEG signals; analyzing the second motor intention feature; determining whether the second motor intention feature meets the activation criteria; and if so, applying motor stimulation.

[0021] The proprioception-guided brain-computer interface closed-loop rehabilitation training system and method of the present invention can bring the following beneficial effects: 1. Employing a step-by-step training approach: Through a three-step protocol of "pure sensation → sensation + attempt → pure imagination / attempt," this approach aligns with the laws of neural learning, gradually reducing dependence on external stimuli, strengthening central internal drive, and promoting the transformation of motor patterns from exogenous induction to endogenous generation. Utilizing the patient's residual proprioceptive pathways, it provides concrete and accurate "sensory templates" for abstract motor imagery, greatly enhancing the vividness and accuracy of motor imagery and solving key difficulties in traditional motor imagery training. 2. By using brain-computer interfaces to accurately capture neural markers of "active attempts", functional feedback and assistance are only provided when a valid motor intention is detected, thus achieving intention-driven precise closed-loop rehabilitation and avoiding the drawbacks of passive training. 3. It has a wide range of applications and is suitable for any motor dysfunction patient who retains some sensory function (pelvic floor muscles, upper limbs, lower limbs), with good universality and scalability. Attached Figure Description

[0022] Figure 1 This is a flowchart of the sensory training process for the brain-computer interface closed-loop rehabilitation training method based on proprioception guidance according to the present invention.

[0023] Figure 2This is a flowchart illustrating the sensory guidance of the brain-computer interface closed-loop rehabilitation training method based on proprioception guidance according to the present invention.

[0024] Figure 3 This is a flowchart of the imagination control method of the brain-computer interface closed-loop rehabilitation training method based on proprioception guidance according to the present invention. Detailed Implementation

[0025] This embodiment proposes a closed-loop rehabilitation training system based on proprioception-guided brain-computer interface, which mainly includes a stimulation module, an EEG acquisition and processing module, an intention decoding module, a central control module, and a multimodal feedback module. The stimulation module is connected to the central control module, which in turn is connected to both the intention decoding module and the multimodal feedback module. The intention decoding module is connected to the EEG acquisition and processing module. The stimulation module can apply corresponding external stimuli to the patient's target muscle groups. The intention decoding module can determine whether the corresponding motor intention characteristics meet preset standards based on its internal algorithm model. The EEG acquisition and processing module can perform real-time acquisition and preprocessing of EEG signals.

[0026] This embodiment organically combines external sensory stimulation with brain-computer interface active rehabilitation training through a step-by-step guided protocol. It uses brain-computer interface technology to analyze the patient's motor intentions and attentional state, forming a closed loop of "sensory input - central integration - intention recognition - feedback reinforcement". The aim is to enhance the realism and effectiveness of motor imagery and promote neural function remodeling.

[0027] The stimulation module mainly includes a sensory stimulation module and a motor stimulation module. The sensory stimulation module can apply external stimuli to generate a feeling of muscle contraction during active movement; the motor stimulation module can apply external stimuli to generate actual limb and muscle movements.

[0028] In this embodiment, the stimulation module can exist in the form of two separate sensory stimulation modules and a motor stimulation module, or it can be the same stimulation module that generates sensation and movement by emitting different frequencies and intensities.

[0029] More specifically, the sensory stimulation module can apply controllable external stimulation to the patient's target muscle groups or corresponding somatosensory cortex to produce a sensation of muscle contraction or joint position similar to active movement; sensory stimulation includes, but is not limited to, one or more of neuromuscular electrical stimulation, functional magnetic stimulation, and mechanical vibration stimulation. The motor stimulation module can apply controllable external stimulation to the patient's target muscle groups or corresponding somatosensory cortex to produce actual limb and muscle movements; motor stimulation includes, but is not limited to, one or more of neuromuscular electrical stimulation, functional magnetic stimulation, rehabilitation robotic arms, and rehabilitation robots.

[0030] In this embodiment, the stimulation module employs a functional electrical stimulation (fEP) module. Electrodes of the fEP module are placed on target muscle groups to generate sensory and motor stimulation. During rehabilitation training for hemiplegic patients to improve hand grasping function, the fEP module is used, with electrodes placed on the wrist and finger extensor muscles of the affected forearm to generate sensory and motor stimulation. During rehabilitation training for patients with pelvic floor dysfunction, a pelvic floor magnetic stimulation module is used. The treatment chair has a built-in magnetic stimulation coil that can focus stimulation on the pelvic floor muscles. While the patient sits on the chair, sensory and motor stimulation is generated through the magnetic coil.

[0031] The central control module executes a predetermined step-by-step guided process, provides sensory stimulation and multimodal feedback, and controls the start and stop of the sensory stimulation module and its parameters; based on the output of the intention decoding module, it provides the patient with motor stimulation and multimodal feedback (such as visual, auditory, and tactile feedback), and controls the start and stop of the motor stimulation module and its parameters.

[0032] In this embodiment, the central control module includes a computer host and a control motherboard, which can run operating systems such as Windows, Linux and Android, and can receive signals from the EEG intention decoding module and send instructions to the stimulation module.

[0033] The intention decoding module includes a first operation unit, a second operation unit, and a third operation unit. The first operation unit performs EEG signal decoding during the perception training phase; the second operation unit performs EEG signal decoding and comparison during the sensory guidance phase; and the third operation unit performs EEG signal decoding and comparison during the imagination control phase.

[0034] In this technical solution, the first arithmetic unit is connected to the second arithmetic unit and the third arithmetic unit respectively, and the second arithmetic unit is connected to the third arithmetic unit.

[0035] More specifically, in the sensory training phase, the first computational unit decodes EEG signals and constructs individualized sensory-evoked EEG features and models. In the sensory guidance phase, the second computational unit analyzes the signal features of actively imagined / attempted contraction of target muscles / limbs from the acquired EEG signals; by comparing with a set threshold or model, it determines whether the motor intention features have reached the preset activation criteria. In the imagination control phase, the third computational unit, based on established muscle awareness and motor attempts, analyzes the signal features of actively imagined or attempted contraction of target muscles / limbs from the acquired EEG signals; by comparing with a set threshold or model, it determines whether the motor intention features have reached the preset activation criteria.

[0036] The multimodal feedback module includes a visual feedback unit and an auditory feedback unit. The visual feedback unit utilizes the influence of limb movements provided by the computer and presents dynamic movements using virtual reality technology; the auditory feedback unit provides movement language guidance.

[0037] Furthermore, the visual feedback unit includes images of limb movements, model examples, and expert guidance provided by a computer, or it can use virtual reality technology to present the dynamic movement process; the auditory feedback unit includes voice guidance for movement provided by the system, prompts for starting and stopping, and feedback on success or failure.

[0038] The EEG acquisition and processing module includes a multi-lead EEG cap, amplifier, and filter, and describes the EEG signal acquisition and preprocessing process when the EEG acquisition and processing module performs different tasks.

[0039] This embodiment also proposes a closed-loop rehabilitation training method based on proprioception-guided brain-computer interface, referencing... Figure 1 , Figure 2 as well as Figure 3 It mainly includes the following three stages.

[0040] During the sensory training phase, stimulation is applied to the target muscle groups, and an individualized sensory evoked EEG model and individualized threshold are obtained based on EEG signals. The sensory training phase includes: applying sensory stimuli, starting training after passing a stimulus test, collecting EEG signals while experiencing the stimuli, saving the EEG signals of this phase as individualized baseline data, and determining individualized thresholds.

[0041] In this technical solution, the stimulation test determines whether the stimulation intensity is appropriate, and if it is not appropriate, the stimulation intensity is adjusted.

[0042] In this embodiment, the central control module triggers the sensory stimulation module to apply one or more standardized stimuli; at the same time, the EEG acquisition and processing module records the patient's EEG signals during this period; the system prompts the patient to perceive the movement or muscle contraction sensation brought about by the stimulation, without making any active movements or attempts, only establishing muscle awareness.

[0043] During the sensory guidance phase, the first motor intention feature is analyzed; by comparing it with the set threshold or model, it is determined whether the first motor intention feature meets the activation criteria.

[0044] The sensory guidance phase includes: after the sensory stimulation ends, within a set time window, guiding the patient to try to perform the induced action or contract the target muscle, analyzing the EEG signals within this time window in real time, determining whether the first motor intention feature has reached the activation standard, and if so, applying motor stimulation.

[0045] In this embodiment, the central control module triggers the sensory stimulation module to apply stimulation. After the stimulation ends, a preset time window is entered. Within this time window, the system prompts the patient to actively try to replicate the sensations experienced during the sensory training phase, i.e., to try to perform the induced movement or contract the target muscle. The EEG acquisition and processing module analyzes the EEG signals within this time window in real time. Once the intention recognition module determines that the parsed first motor intention feature meets the preset standard, the central control module immediately triggers a positive feedback signal (e.g., a "success!" prompt sound, virtual limb animation), and can selectively initiate a functional auxiliary training (e.g., triggering functional electrical stimulation to assist in completing an actual movement, or driving an exoskeleton / robot to assist in joint movement). If the standard is not met, an encouraging prompt is given after the time window ends, and training restarts.

[0046] During the imagination control phase, the characteristics of the second motor intention are analyzed; it is then determined whether the characteristics of the second motor intention meet the activation criteria.

[0047] In this embodiment, the central control module directly prompts the patient to autonomously perform motor imagery based on their prior memory without providing any pre-existing sensory stimulation. The system assesses the degree of motor imagery or attempt through electroencephalogram (EEG) signals and provides corresponding quantitative feedback to strengthen the patient's intrinsic motivation. Once the intention recognition module determines that the parsed second motor intention feature meets the preset standard, the central control module immediately triggers a positive feedback signal (e.g., a "success!" sound or virtual limb animation) and can selectively initiate a functional auxiliary training (e.g., triggering functional electrical stimulation to assist in completing an actual movement, or driving an exoskeleton / robot to assist in joint movement). If the standard is not met, an encouraging prompt is given after the time window ends, and training restarts.

[0048] Example 1 was applied to the rehabilitation training of hand grasping function in hemiplegic patients. It includes a stimulation module, an EEG acquisition and processing module, an intention decoding module, a central control module, and a multimodal feedback module.

[0049] The stimulation module employs functional electrical stimulation (fEP) as a means of sensory induction and functional assistance to implement the step-by-step guided training of this invention. The fEP module places electrodes on the wrist and finger extensor muscles of the patient's affected forearm to generate sensory and motor stimulation. The EEG acquisition and processing module uses a portable multi-lead EEG device to monitor and record EEG signals from the contralateral sensorimotor cortex (e.g., C3 / C4 / Cz channels). The central control module includes a computer host and a control motherboard, capable of running Windows, Linux, and Android operating systems. It can receive signals from the EEG intention decoding module and send commands to the stimulation module. The multimodal feedback module includes a display, projector, speakers (headphones), and VR glasses.

[0050] The training process includes the following steps.

[0051] Step S1, Perceptual Training: The multimodal feedback module provides on-screen and voice prompts, "Please concentrate and feel the sensation of grasping." Subsequently, the electrical stimulation module outputs a 3-second, constant-parameter suprathreshold electrical stimulus (e.g., 30Hz frequency, 200μs pulse width, intensity sufficient to induce finger extension without causing functional movement or discomfort). During this period, the patient experiences a slight extension of the hand and accompanying muscle contraction induced by the electrical stimulation, without making any active attempts. The EEG acquisition and processing module simultaneously records and saves the EEG signals during this phase as individualized baseline data for the sensory evoked period. Through repeated stimulation-sensory training, this helps stroke patients establish muscle awareness of hand grasping movements.

[0052] Step S2: Sensory Guidance Training The multimodal feedback module provides on-screen and voice prompts: "Please concentrate, feel the sensation of grasping, and try to perform this action yourself." The stimulation module outputs the exact same stimulus as S1; after the stimulation ends, the system prompts the patient to actively imagine or attempt to grasp, setting a 5-10 second time window. Within this time window, the patient must try to autonomously replicate the grasping action they just felt. The EEG signal processing module analyzes the EEG signals within the time window in real time, decoding the user's motor intention; when the decoded first motor intention exceeds a preset individualized threshold, the system immediately triggers two feedback mechanisms: functional auxiliary feedback: triggering functional electrical stimulation (FES) of adjustable intensity, frequency, and duration to assist the patient in completing an actual hand grasping action, reinforcing the "intention-outcome" association. The multimodal feedback module actively confirms the feedback: playing a success prompt sound and displaying an animation of a virtual hand successfully extending on the screen.

[0053] If no valid first motor intention is detected within the time window, the system provides an encouraging prompt of "Please try again" and proceeds to the next cycle to provide sensory stimulation again.

[0054] Step S3: Visualization Control Training: The system prompts, "Now, please imagine your hand extending." No electrical stimulation is given beforehand at this stage. The patient must rely entirely on the "sensory-motor" memory established in the first two steps to autonomously imagine extending their hand for 5-10 seconds. The system decodes the user's motor intention. When the decoded second motor intention exceeds a preset individualized threshold, the system immediately triggers two feedback mechanisms: functional auxiliary feedback: triggering functional electrical stimulation (FES) of adjustable intensity, frequency, and duration to assist the patient in completing an actual hand grasping motion, reinforcing the "intention-outcome" association; and active confirmation feedback through a multimodal feedback module: playing a success prompt tone and displaying an animation of a virtual hand successfully extending on the screen.

[0055] If no valid second motor intention is detected within the time window, the system provides an encouraging prompt of "Please try again" and proceeds to the next cycle, providing sensory stimulation again.

[0056] More specifically, the operation process of this embodiment includes: The system is activated. The therapist puts an EEG cap on the patient and places electrical stimulation electrodes on the wrist and finger extensor muscles of the patient's affected forearm to generate sensory and motor stimulation. The therapist applies electrical stimulation of specific intensity and frequency, allowing the patient to experience the sensation of the limbs and muscles when grasping with their hand. The stimulation intensity is adjusted based on the patient's feedback until the patient can clearly feel the movement of the limbs and muscles when grasping with their hand. Step S1, sensory training, is initiated. A single cycle includes: a resting state before stimulation, 2 seconds of sensory stimulation, and a resting state after stimulation. The patient's EEG signals are recorded during this process. After 3-5 minutes of stimulation, the patient is helped to establish muscle awareness. The collected EEG signals are analyzed to establish a patient EEG model. Step S2, sensory guidance training, is then initiated. Similar to step S1, a suitable sensory stimulation intensity is found for the patient, and training is started. A single cycle includes: a resting state before stimulation, 2 seconds of sensory stimulation, prompting the patient to actively imagine / attempt hand grasping movements after stimulation, the system decoding the user's brainwave motor intention, and when the decoded motor intention exceeds the preset individualized threshold or preset score, the system provides the user with functional electrical stimulation to complete the full grasping action. At the same time, the system actively confirms the feedback through the multimodal feedback module: plays a success prompt tone and displays an animation of the virtual hand successfully extending on the screen.

[0057] If no effective motor intention is detected within the time window, the system provides an encouraging prompt of "Please try again," proceeds to the next cycle, and provides sensory stimulation again until the training ends. Once the patient has established complete muscle awareness and no longer needs sensory prompting, step S3, imagery control training, is initiated. After training begins, the system prompts the patient to actively imagine / attempt hand grasping movements. The system decodes the user's EEG motor intention. When the decoded motor intention exceeds a preset individualized threshold or a preset score, the system provides functional electrical stimulation to the user to complete the grasping action. Simultaneously, the system actively confirms feedback through the multimodal feedback module: playing a success prompt tone and displaying an animation of a virtual hand successfully extended on the screen. If no effective motor intention is detected within the time window, the system provides an encouraging prompt of "Please try again," proceeds to the next cycle, and continues until the training ends.

[0058] Example 1 was applied to the rehabilitation training process of patients with pelvic floor dysfunction, which includes a stimulation module, an EEG acquisition and processing module, an intention decoding module, a central control module, and a multimodal feedback module.

[0059] The stimulation module employs pelvic floor magnetic stimulation as a sensory induction and functional aid, implementing the step-by-step guided training of this invention. The pelvic floor magnetic stimulation module uses a treatment chair with built-in magnetic stimulation coils that can focus stimulation of the pelvic floor muscles. While the patient sits in the chair, sensory and motor stimulation is generated through the magnetic coils. The EEG acquisition and processing module uses a portable multi-lead EEG device to monitor EEG signals from the sensorimotor cortex related to the pelvic floor muscles (such as the Cz, FCz, and CPz channels). The central control module includes a computer host and a control motherboard, capable of running Windows, Linux, and Android operating systems. It can receive signals from the EEG intention decoding module and send commands to the stimulation module. The multimodal feedback module includes a display, projector, speakers (headphones), and VR glasses.

[0060] The training process includes the following steps.

[0061] Step S1: Perceptual Training The patient sits in the magnetic stimulation chair, and the system prompts, "Please concentrate and feel the upward contraction of the pelvic floor." The magnetic stimulator outputs a suprathreshold stimulus lasting 2 seconds (e.g., frequency 20Hz, intensity sufficient to clearly allow the patient to feel the pelvic floor muscle contraction and lifting). The patient experiences the proprioceptive sensation of pelvic floor muscle contraction and lifting induced by the stimulus. The EEG acquisition and processing module records the EEG signals of the relevant cortical areas during this stage as the basic model.

[0062] Step S2: Sensory Guidance Training The modal feedback module provides on-screen and voice prompts: "Please concentrate, feel the upward contraction of the pelvic floor, and try to contract your pelvic floor muscles." The stimulation module then outputs the exact same magnetic stimulation as S1. After stimulation, the system prompts the patient to actively imagine or try to contract their pelvic floor muscles, setting a 5-10 second time window. Within this time window, the patient must try to replicate the pelvic floor contraction they just experienced. The EEG acquisition and processing module analyzes the EEG signals within the time window in real time, decoding the user's motor intention. When the decoded first motor intention exceeds a preset individualized threshold, the system immediately triggers two feedback mechanisms: functional auxiliary feedback: triggering a pelvic floor magnetic stimulation with adjustable intensity, frequency, and duration to assist the patient in completing an actual pelvic floor muscle contraction; and positive confirmation feedback through the multimodal feedback module: playing a success prompt tone and displaying an animation of a virtual hand successfully extended on the screen. If no effective first motor intention is detected within the time window, the system provides an encouraging prompt of "Please try again," and proceeds to the next cycle, providing sensory stimulation again.

[0063] Step S3: Visualization Control Training: The system prompts, "Please independently imagine pelvic floor contraction." No sensory stimulation is provided beforehand at this stage. The patient must rely entirely on the "sensory-motor" memory established in the first two steps to autonomously imagine pelvic floor muscle contraction for 5-10 seconds. The system decodes the user's motor intention. When the decoded second motor intention exceeds a preset individualized threshold, the system immediately triggers two feedback mechanisms: functional auxiliary feedback: triggering a pelvic floor magnetic stimulation with adjustable intensity, frequency, and duration to assist the patient in completing an actual pelvic floor muscle contraction; and active confirmation feedback through a multimodal feedback module: playing a success prompt tone and displaying an animation of a virtual hand successfully extending on the screen. If no valid second motor intention is detected within the time window, the system provides an encouraging prompt, "Please try again," and proceeds to the next cycle, providing sensory stimulation again.

[0064] More specifically, the operation process of this embodiment includes: The system is activated, and the patient sits correctly in the chair equipped with the pelvic floor magnetic stimulation module. The therapist puts an EEG cap on the patient. The therapist provides pelvic floor magnetic stimulation of a specific intensity and frequency, allowing the patient to feel the pelvic floor muscles contracting and lifting. The therapist adjusts the stimulation intensity based on the patient's feedback until the patient can clearly feel the pelvic floor muscles contracting and lifting.

[0065] Step S1, sensory training, is initiated. A single cycle includes: a resting state before stimulation, 2 seconds of sensory stimulation, and a resting state after stimulation. The patient's EEG signals are recorded simultaneously during this process. After 3-5 minutes of stimulation, the patient is helped to establish muscle awareness. The collected EEG signals are then analyzed to create a patient EEG model.

[0066] Initiating step S2, sensory-guided training, begins similarly to step S1, finding an appropriate sensory stimulation intensity for the patient and starting the training. A single cycle includes: a resting state before stimulation, 2 seconds of sensory stimulation, followed by prompting the patient to actively imagine / attempt pelvic floor muscle contraction and lifting. The system decodes the user's EEG motor intention. When the decoded motor intention exceeds a preset individualized threshold or a preset score, the system provides pelvic floor magnetic stimulation to complete the entire process of pelvic floor muscle contraction, lifting, holding, and relaxation. Simultaneously, the multimodal feedback module actively confirms feedback: playing a success prompt tone and displaying an animation of pelvic floor muscle contraction on the screen.

[0067] If no valid movement intention is detected within the time window, the system provides an encouraging prompt of "Please try again" and enters the next cycle, providing sensory stimulation again until the training ends.

[0068] Once the patient has established complete muscle awareness and no longer needs sensory prompting, step S3, the imagery control training, is initiated. After the training begins, the system prompts the patient to actively imagine / attempt pelvic floor muscle contraction and lifting movements. The system decodes the user's brainwave motor intentions. When the decoded motor intentions exceed the preset individualized threshold or preset score, the system provides the user with pelvic floor magnetic stimulation to complete the entire process of pelvic floor muscle contraction, lifting, holding, and relaxation. At the same time, the system actively confirms feedback through the multimodal feedback module: plays a success prompt tone and displays an animation of pelvic floor muscle contraction on the screen.

[0069] If no valid movement intention is detected within the time window, the system provides an encouraging prompt of "Please try again" and enters the next cycle, providing sensory stimulation again until the training ends.

[0070] In this embodiment, in step S1, the patient's pre-stimulation resting state, sensory stimulation state, and post-stimulation state are analyzed, and the patient's EEG model is calculated. The model is then stored in the intent decoding module.

[0071] For the EEG data collected over a period of time, including pre-stimulation resting state, sensory stimulation state, and post-stimulation state, the data underwent preprocessing, including notch filtering at 50Hz power frequency and bandpass filtering from 0.5 to 40Hz. Data in frequency bands close to the sensory stimulation frequency were removed from the EEG data obtained during the sensory stimulation phase, resulting in preprocessed data. Calculate the characteristic data of each EEG frequency band. EEG frequency band feature data is the result of calculating the mean, variance, standard deviation, coefficient of variation, power spectral density, asymmetry coefficient, Pearson correlation coefficient, and cosine similarity of preprocessed EEG data.

[0072] Specifically, EEG model Represented as .

[0073] in, The power spectrum threshold for electroencephalogram (EEG). The average value of the pre-stimulus resting state characteristics is equal to the characteristic coefficient a1 multiplied by the average value of the pre-stimulus resting state characteristics, minus the characteristic coefficient a2 multiplied by the average value of the pre-stimulus resting state characteristics, and then minus the characteristic coefficient a3 multiplied by the average value of the post-stimulus resting state characteristics. The average value of the pre-stimulus resting state characteristics is the sum of the m pre-stimulus resting state characteristics. The sum of the sums divided by the data length m gives the average of the stimulus state features. The sum of the sums divided by the data length m gives the average of m post-stimulus resting state features. The sum is then divided by the data length m.

[0074] in, The threshold for EEG feature similarity. It consists of two parts: the first part is the weighted result of Pearson related exercise features, and the second part is the weighted result of cosine similarity features.

[0075] The weighted result of the Pearson correlation features is the coefficient k1 multiplied by the first calculation term, which is the Pearson correlation coefficient feature coefficient b1 multiplied by the pre-stimulus resting state features. Subtract the characteristic coefficient b2 of the Pearson correlation coefficient and multiply by the state characteristics in the stimulus. Subtract the characteristic coefficient b3 of the Pearson correlation coefficient and multiply by the post-stimulus state characteristic. .

[0076] The weighted result of the cosine similarity feature is the coefficient k2 multiplied by the second calculation term, which is the cosine similarity feature coefficient c1 multiplied by... Subtract the cosine similarity feature coefficient c2 and multiply by Subtract the cosine similarity feature coefficient c3 and multiply by , , , The cosine similarity is calculated by matching the pre-stimulus resting state features, the mid-stimulus state features, and the post-stimulus state features with the standard feature template.

[0077] The sum of coefficients k1 and k2 is 1. Coefficients k1 and k2 are the Pearson correlation coefficient feature and the cosine similarity feature, respectively.

[0078] in, The threshold for the coefficient of variation of electroencephalogram (EEG). Characteristic coefficients Multiplied by the characteristics of the resting state before stimulation Subtract the characteristic coefficient Multiply by stimulus state characteristics Subtract the characteristic coefficients Multiplied by the resting state characteristics after stimulation .

[0079] In step S2, the first motor intention feature can be specifically represented as an imagined state score. The preset auxiliary score C is calculated by adding the score conversion coefficient h1 multiplied by the first coefficient, then adding the score conversion coefficient h2 multiplied by the second coefficient, and finally adding the score conversion coefficient h3 multiplied by the third coefficient. The first coefficient is the real-time EEG power spectrum threshold. Subtract the difficulty coefficient t1 and multiply by the EEG power spectrum threshold. The second coefficient is the real-time similarity of EEG features. Subtract the difficulty coefficient t2 and multiply by the EEG feature similarity threshold. The third coefficient is the real-time EEG characteristic variation value. Subtract the difficulty coefficient t3 and multiply by the EEG variability threshold. The difficulty levels t1, t2, and t3 can be adjusted to change the training difficulty based on the actual situation.

[0080] Among them, real-time EEG feature similarity value The calculation method and the threshold of EEG power spectrum The calculation method is the same; it only requires adjusting the EEG power spectrum threshold. In the calculation method, the average value of the resting state characteristics after stimulation is replaced with the real-time state characteristics. The average value is sufficient.

[0081] Real-time EEG feature similarity The calculation method and the similarity threshold of EEG features The calculation method is the same; it only requires setting the EEG feature similarity threshold. Calculation method Replace with That's all. The Pearson correlation coefficient is calculated by matching real-time state features after sensory stimulation with standard feature templates.

[0082] Real-time EEG variability The calculation method and the threshold of the coefficient of variation of electroencephalogram (EEG) The calculation method is the same; it only requires setting the EEG variability threshold. In the calculation method Replace with real-time resting state features after stimulation That's all.

[0083] In step S3, there is no sensory stimulus process. The stimulus stage characteristics from step S2 can be substituted into the calculation model, or the sensory stimulus stage characteristics can be removed from the calculation formula.

[0084] In this embodiment, the preset auxiliary score C can be set according to actual needs; generally, EEG signals include... frequency band frequency band frequency band frequency band and For frequency bands, you can either select one frequency band for calculation or select multiple frequency bands from multiple leads for calculation as needed.

[0085] In this embodiment, the sensory stimulation module may provide suprathreshold sensory stimulation during the sensory training phase and then stop; it may also provide suprathreshold sensory stimulation during both the sensory training and sensory guidance phases, but the stimulation intensity may be adjusted as needed; or it may provide suprathreshold sensory stimulation during the sensory training phase and subthreshold stimulation during the sensory guidance phase.

Claims

1. A brain-computer interface closed-loop rehabilitation training system based on proprioception guidance, characterized in that, The device includes a stimulation module and a central control module connected to the stimulation module. The stimulation module applies external stimulation to the target muscle group. The central control module is connected to an intention decoding module and a multimodal feedback module. The intention decoding module determines whether the motor intention features meet the standard based on a built-in algorithm model. The intention decoding module is connected to an EEG acquisition and processing module, which acquires and preprocesses EEG signals in real time.

2. The brain-computer interface closed-loop rehabilitation training system based on proprioception guidance according to claim 1, characterized in that, The stimulation module includes a sensory stimulation module and a motor stimulation module. The sensory stimulation module can apply external stimuli to generate a feeling of muscle contraction during active movement; the motor stimulation module can apply external stimuli to generate actual limb and muscle movements.

3. The brain-computer interface closed-loop rehabilitation training system based on proprioception guidance according to claim 2, characterized in that, The central control module executes a predetermined step-by-step guided process, provides sensory stimulation and multimodal feedback, and controls the start and stop of the sensory stimulation module and its parameters; based on the output of the intention decoding module, it provides motor stimulation and multimodal feedback to the patient, and controls the start and stop of the motor stimulation module and its parameters.

4. The proprioceptive-guided brain-computer interface closed-loop rehabilitation training system according to claim 1 or 2, characterized in that, The intention decoding module includes a first computing unit, a second computing unit, and a third computing unit. The first computing unit performs EEG signal decoding during the perception training phase; the second computing unit performs EEG signal decoding and comparison during the sensory guidance phase; and the third computing unit performs EEG signal decoding and comparison during the imagination control phase.

5. The proprioceptive-guided brain-computer interface closed-loop rehabilitation training system according to claim 1 or 2, characterized in that, The multimodal feedback module includes a visual feedback unit and an auditory feedback unit. The visual feedback unit utilizes the influence of limb movements provided by the computer and presents dynamic movements using virtual reality technology. The auditory feedback unit provides movement language guidance.

6. The proprioceptive-guided brain-computer interface closed-loop rehabilitation training system according to claim 1, 2, or 3, characterized in that, The EEG acquisition and processing module includes a multi-lead EEG cap, amplifier, and filter, and describes the EEG signal acquisition and preprocessing process when the EEG acquisition and processing module performs different tasks.

7. A proprioceptive-guided brain-computer interface closed-loop rehabilitation training method, applicable to the proprioceptive-guided brain-computer interface closed-loop rehabilitation training system as described in any one of claims 1-6, characterized in that, include: During the sensory training phase, stimulation is applied to the target muscle groups, and individualized sensory evoked EEG models and individualized thresholds are obtained based on EEG signals. During the sensory guidance phase, the first motor intention feature is analyzed; by comparing it with a set threshold or model, it is determined whether the first motor intention feature meets the activation criteria. During the imagination control stage, the characteristics of the second motor intention are analyzed; Determine whether the second motor intention feature meets the activation criteria.

8. The brain-computer interface closed-loop rehabilitation training method based on proprioception guidance according to claim 7, characterized in that, The perception training phase includes: Sensory stimulation is applied, and training is initiated after passing the stimulation test. EEG signals are collected while the stimulation is being experienced, and the EEG signals at this stage are saved as individualized baseline data and individualized thresholds are determined.

9. The brain-computer interface closed-loop rehabilitation training method based on proprioception guidance according to claim 7 or 8, characterized in that, The sensory guidance phase includes: after the sensory stimulation ends, within a set time window, guiding the patient to try to perform the induced action or contract the target muscle, analyzing the EEG signals within this time window in real time, determining whether the first motor intention feature has reached the activation standard, and if so, applying motor stimulation.

10. The brain-computer interface closed-loop rehabilitation training method based on proprioception guidance according to claim 7 or 8, characterized in that, The imagination control phase includes: without prior sensory stimulation, directly prompting the patient to autonomously perform motor imagination based on the memory of the previous stage; assessing the degree of motor imagination or attempt through EEG signals; analyzing the second motor intention characteristics; determining whether the second motor intention characteristics meet the activation criteria; and if so, applying motor stimulation.

Images