A Method and System for EEG-based Assessment of Pelvic Floor Muscle Motion Brain State

By integrating the temporal and frequency characteristics and functional connectivity characteristics of EEG signals, a pelvic floor muscle brain state assessment model was constructed, which solved the problem of lack of central nervous system state perception in existing pelvic floor muscle rehabilitation technologies, realized personalized neurofeedback training, and improved the effect of pelvic floor muscle rehabilitation.

CN123101162APending Publication Date: 2026-11-17BEIJING NAOLI TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611460103.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-18
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing pelvic floor muscle rehabilitation techniques lack perception of the central nervous system state. Electromyographic signals can only reflect muscle contraction activity and cannot assess motor intention. The methods for recognizing motor intention are limited, lack personalized adaptability, and cannot comprehensively assess the level of brain function recovery.

Method used

By collecting multi-channel EEG signals, performing time-frequency feature analysis and source-space functional connectivity analysis, and fusing time-frequency features and functional connectivity features, a pelvic floor muscle brain state assessment model is constructed, a neurorehabilitation score is calculated, and the neurofeedback training strategy is dynamically adjusted based on the assessment results.

Benefits of technology

It enables a comprehensive assessment of the central nervous system state of pelvic floor muscle movement intentions, provides personalized neurofeedback training, and improves the compliance and effectiveness of rehabilitation training.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123101162A_ABST
    Figure CN123101162A_ABST
Patent Text Reader

Abstract

This invention discloses a method and system for assessing brain state during pelvic floor muscle movement based on EEG fusion features. The method includes: acquiring multi-channel EEG signals from subjects performing pelvic floor muscle rehabilitation training tasks and preprocessing the EEG signals; performing time-frequency feature analysis on the preprocessed EEG signals to extract time-frequency features related to pelvic floor muscle movement; performing source-space functional connectivity analysis on the preprocessed EEG signals to extract functional connectivity features between different brain regions; fusing the time-frequency features with the functional connectivity features to construct a pelvic floor muscle brain state assessment model based on the fused multidimensional features; calculating a pelvic floor muscle neurorehabilitation score using the brain state assessment model; and classifying and assessing the subject's brain state based on the pelvic floor muscle neurorehabilitation score, and dynamically adjusting the neurofeedback training strategy based on the assessment results. This method helps promote neuroplasticity, compliance, and effectiveness in the pelvic floor muscle rehabilitation process.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of brain-computer interface technology, specifically to a method and system for assessing brain state during pelvic floor muscle movement based on EEG fusion characteristics. Background Technology

[0002] Brain-computer interface (BCI) is a technology that directly connects the brain to external devices, aiming to enable communication between brain signals and computers or other devices. BCI acquires brain activity signals through methods such as electroencephalography (EEG), preprocesses the brain signals to extract useful features (e.g., signal filtering, feature extraction, and classification), and then converts the processed signals into control commands to drive external devices or software to perform corresponding operations.

[0003] Pelvic floor dysfunction (PFD) is a group of diseases caused by damage to the pelvic floor support structures, neuromuscular function, or tissues. It generally includes stress urinary incontinence, chronic pelvic floor pain, and postpartum pelvic floor dysfunction. With an aging population and increasing demand for postpartum rehabilitation among women, pelvic floor rehabilitation has gradually become an important research direction in rehabilitation medicine and obstetrics and gynecology. Currently, widely used rehabilitation methods in clinical practice include self-training, electrical stimulation therapy, and biofeedback training.

[0004] Kegel exercises are a method of actively training and contracting the pelvic floor muscles. By repeatedly and voluntarily contracting and relaxing the pelvic floor muscles, they can improve muscle strength and endurance, and alleviate pelvic floor dysfunction such as urinary incontinence. This method is the most basic form of rehabilitation training. However, it has the following problems: 1) Patients often have difficulty locating the pelvic floor muscles and tend to contract their thigh muscles more; 2) The training cycle is long, lacks feedback, and has low compliance; 3) It is impossible to assess whether neurological function is involved.

[0005] Electrical stimulation therapy applies electrical pulses to the pelvic floor muscles via vaginal or surface electrodes, causing pelvic floor contraction, strengthening the pelvic floor muscles, and improving nerve conduction function. This method is fast-acting and suitable for patients with pelvic floor dysfunction, but it has the following problems: 1) It cannot recognize the patient's movement intention; 2) It cannot adjust according to the brain state, and the training mode is relatively simple and fixed; 3) It is easy to cause an increase in the response threshold, and the long-term effect is greatly reduced.

[0006] Muscle biofeedback training collects electromyographic (EMG) signals from the surface of the pelvic floor muscles, providing real-time feedback on the intensity of muscle contraction or relaxation and offering numerical feedback on EMG intensity to assist patients in completing pelvic floor muscle rehabilitation training. This method can, to some extent, improve patients' understanding of the pelvic floor muscle rehabilitation process, increasing training interest and compliance. However, it has the following problems: 1) EMG signals can only represent muscle contraction and cannot reflect brain state; 2) In some patients with nerve damage, EMG signals may not be available.

[0007] In recent years, brain-computer interface technology has been gradually applied to the field of rehabilitation. It mainly collects the electroencephalogram (EEG) signals of patients during motor imagery, analyzes the motor intentions, and drives external devices to guide patients through the corresponding rehabilitation process. The main areas include active and passive rehabilitation of the upper and lower limbs, hand function rehabilitation, and language decoding. However, brain-computer interface technology has rarely been involved in the field of pelvic floor muscle rehabilitation.

[0008] In the field of brain-computer interfaces (BCIs), some existing technologies actively collect patients' intentions to contract their pelvic floor muscles through BCI. Upon reaching a corresponding EEG response threshold, the pelvic floor magnetic stimulation module is activated to assist the patient in achieving pelvic floor muscle contraction. Other existing technologies collect relevant pelvic floor muscle electromyography (EMG) data to assess contraction, then upload the data to immersive VR devices. Combined with biofeedback systems and virtual reality scenarios, this provides real-time feedback on training effectiveness. Still other existing technologies utilize coordinated EEG and EMG data acquisition, integrating EMG characteristics into EEG characteristics to enhance the application value of EEG signals in pelvic floor muscle rehabilitation. However, these technologies primarily involve the time-frequency domain characteristics of EEG.

[0009] Given that current pelvic floor muscle rehabilitation primarily focuses on muscle contraction with little involvement or feedback from the central nervous system, such as brain status, and rarely involves personalized rehabilitation training, there is a need for a method and system for assessing brain status and providing adaptive neurofeedback in pelvic floor muscle rehabilitation based on the fusion of multidimensional EEG features. Summary of the Invention

[0010] Therefore, embodiments of the present invention provide a method and system for assessing the brain state of pelvic floor muscle movement based on EEG fusion characteristics, in order to solve at least one technical problem existing in the prior art.

[0011] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions: This invention provides a method for assessing the brain state of pelvic floor muscle movement based on electroencephalogram (EEG) fusion features, the method comprising: Multichannel electroencephalogram (EEG) signals were collected from the subjects during the pelvic floor muscle rehabilitation training task, and the EEG signals were preprocessed. Time-frequency feature analysis was performed on the preprocessed EEG signals to extract time-frequency features related to pelvic floor muscle movement; source-space functional connectivity analysis was performed on the preprocessed EEG signals to extract functional connectivity features between different brain regions. The time-frequency features and the functional connectivity features are fused together, and a pelvic floor muscle brain state assessment model is constructed based on the fused multidimensional features. The pelvic floor muscle nerve rehabilitation score is calculated through the brain state assessment model. The pelvic floor muscle nerve rehabilitation score includes at least the task participation index, motor cortex activation index, brain network synergy index, effective maintenance index, and resting inhibition index. The subjects' brain status was graded and assessed based on the pelvic floor muscle neurorehabilitation score, and the neurofeedback training strategy was dynamically adjusted based on the assessment results.

[0012] In some embodiments, multi-channel electroencephalogram (EEG) signals are collected from the subject during the performance of a pelvic floor muscle rehabilitation training task, and the EEG signals are preprocessed, specifically including: The pelvic floor muscle rehabilitation training task includes a left pelvic floor muscle stimulation task, a right pelvic floor muscle stimulation task, an active pelvic floor muscle contraction task, and a resting task. The left and right pelvic floor muscle stimulation tasks are achieved through artificial physical stimulation by vaginal digital examination by a gynecological rehabilitation therapist. The active pelvic floor muscle contraction task is achieved by the subject actively contracting the pelvic floor muscles, and the resting task is achieved by the subject maintaining a relaxed state. EEG signals are acquired synchronously using a multi-channel EEG acquisition device, with a sampling rate not lower than a preset sampling frequency; the preprocessing includes, in sequence: downsampling, bandpass filtering, power frequency notch filtering, segmentation, artifact removal, average reference, and baseline correction.

[0013] In some embodiments, time-frequency feature analysis is performed on the preprocessed EEG signal to extract time-frequency features related to pelvic floor muscle movement, specifically including: The Morlet wavelet transform was applied to each segment of the single-channel EEG signal to obtain the instantaneous power in the time-frequency domain. Using the signal power of the baseline window before the task as a reference, calculate the event-related spectral perturbation; The power variation features of the alpha band associated with sensorimotor cortex activation and the beta band associated with motor preparation and execution are extracted from the event-related spectral perturbation and used as the time-frequency features.

[0014] In some embodiments, source-space functional connectivity analysis is performed on the preprocessed EEG signals to extract functional connectivity features between different brain regions, specifically including: A forward model is constructed based on the standard template brain and the boundary element model. The minimum norm estimation method is used to generate the inverse solution operator, which maps the scalp EEG signal to the current density of the cortical source space to obtain the source space time series signal. The source space was divided into multiple brain regions (ROIs) using standard brain atlases, and the average time series of each ROI was extracted. The Hilbert transform is applied to the average time series of each ROI to obtain the analytical signal, the instantaneous power is calculated and a logarithmic transform is performed to obtain the power envelope; The power envelope of each ROI pair is orthogonalized to reduce volume conduction effect and zero-delay leakage interference. The orthogonalized Pearson correlation coefficient is calculated and converted into Fisher z value after symmetry processing to obtain the ROI horizontal functional connectivity matrix under each frequency band, which serves as the functional connectivity feature.

[0015] In some embodiments, the time-frequency features are fused with the functional connectivity features, and a pelvic floor muscle brain state assessment model is constructed based on the fused multidimensional features. The pelvic floor muscle neurological rehabilitation score is calculated using the brain state assessment model, including: The time-frequency features are used to describe the activation level of local motor cortex, and the functional connectivity features are used to describe the coordinated changes between different brain regions. The pelvic floor muscle neurorehabilitation score (PFNRS) is calculated using the following formula: ; in, The task engagement index assesses whether a subject has entered an effective task state by comparing changes in brain activity power between the resting state and the task state. The motor cortex activation index is used to assess the response intensity of the motor cortex by evaluating the degree of task-related desynchronization. It is a brain network synergy index that assesses the degree of brain region synergy by comparing the strength of functional connectivity between the task state and the resting state. To effectively maintain the index, the duration of sustained brain state activation was assessed by the ratio of effective desynchronization duration to total task duration. The resting inhibition index assesses a subject's relaxation ability by measuring the degree of activation in the resting state. , , , , These are the weighting coefficients for each index.

[0016] In some embodiments, the method further includes: Each index is normalized to a score of 1-100; Based on the normalized scoring range, the subjects' brain states are divided into multiple levels, which include at least: effective neural control state, formation neural control state, insufficient motor activation state, insufficient brain network coordination state, insufficient sustained control state, insufficient relaxation control state, and low-involvement training state. Each level corresponds to different brain state characteristics and feedback strategies.

[0017] In some embodiments, the subject's brain state is graded and assessed based on the pelvic floor muscle neurorehabilitation score, and the neurofeedback training strategy is dynamically adjusted based on the assessment results, specifically including: When the assessment indicates an effective state of neural control, reinforce the correct neural pattern, increase reward feedback, and increase the training difficulty and retention time. When assessed as a state of insufficient motor activation, enhance motor cortex recruitment, increase motor imagery cues, and improve feedback intensity in the target area; When assessed as a state of insufficient brain network coordination, promote brain network connectivity, increase the number of task repetitions, and provide overall brain state feedback. When assessed as a state of persistent undercontrol, improve stable control ability, prolong training maintenance time, and reduce transient contraction ratio; When the assessment indicates insufficient relaxation control, improve active relaxation ability, switch relaxation training mode, and increase breathing regulation feedback; When the assessment indicates low training participation, increase training participation, re-guide the task, and increase the frequency of prompts.

[0018] In some embodiments, the training process of the pelvic floor muscle-brain state assessment model includes: A training dataset is constructed by collecting multiple sets of EEG signal samples with known task labels. The task labels of the samples include at least left-side stimulation, right-side stimulation, active exertion, and resting. Preprocessing, time-frequency feature analysis and source space functional connectivity analysis are performed sequentially on the EEG signal samples in the training dataset. The time-frequency features and functional connectivity features corresponding to each sample are extracted and fused into a multi-dimensional feature vector. A multi-task neural network model is constructed, which includes a task state classification head, a brain state index regression head, and a graded evaluation result classification head. The task state classification head is used to output the classification probability of four types of tasks, the brain state index regression head is used to output the regression prediction values ​​of TPI, MAI, BCI, EMI, and RII, and the graded evaluation result classification head is used to output the prediction results of brain state level. Using the multidimensional feature vector of the training dataset as input and the corresponding task label as supervision signal, a multi-task joint loss function is defined. The multi-task joint loss function is composed of a weighted sum of task state classification loss, regression loss of each brain state index, and classification loss of the graded evaluation result. The model parameters are iteratively updated using the backpropagation algorithm until the model converges, resulting in a trained pelvic floor muscle and brain state assessment model, which is then used to perform the pelvic floor muscle and brain state assessment and neural feedback phase based on pre-trained features.

[0019] This invention also provides a pelvic floor muscle motor brain state assessment system based on EEG fusion features, the system comprising: The signal acquisition and preprocessing module is used to acquire multi-channel electroencephalogram (EEG) signals during the subject's pelvic floor muscle rehabilitation training task, and to preprocess the EEG signals. The time-frequency feature extraction module is used to perform time-frequency feature analysis on the preprocessed EEG signal. It uses continuous wavelet transform to calculate the event-related spectral perturbation and extracts the power change features of the alpha and beta bands. The functional connectivity analysis module is used to perform source space functional connectivity analysis on preprocessed EEG signals. Based on a standard template brain, a forward model is constructed and an inverse solution operator is generated. Scalp signals are mapped to the cortical source space. After dividing the brain regions into ROIs, the orthogonalized power envelope correlation method is used to calculate the functional connectivity matrix at the ROI level. The feature fusion and evaluation module is used to fuse the time-frequency features with the functional connectivity features, and calculate the pelvic floor muscle nerve rehabilitation score based on the fused multidimensional features. The pelvic floor muscle nerve rehabilitation score includes at least the task participation index, motor cortex activation index, brain network synergy index, effective maintenance index, and resting inhibition index. The neurofeedback module is used to grade and assess the subject's brain state based on the pelvic floor muscle neurorehabilitation score, and dynamically adjust the neurofeedback training strategy based on the assessment results to achieve personalized pelvic floor muscle rehabilitation training.

[0020] The present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method described above.

[0021] The present invention provides a method for assessing the brain state of pelvic floor muscle movement based on EEG fusion features. This method involves collecting multi-channel EEG signals from subjects during pelvic floor muscle rehabilitation training tasks and preprocessing the EEG signals. Time-frequency feature analysis is performed on the preprocessed EEG signals to extract time-frequency features related to pelvic floor muscle movement. Source-space functional connectivity analysis is then performed on the preprocessed EEG signals to extract functional connectivity features between different brain regions. The time-frequency features and functional connectivity features are fused, and a pelvic floor muscle brain state assessment model is constructed based on the fused multidimensional features. A pelvic floor muscle neurorehabilitation score is calculated using this brain state assessment model. This score includes at least a task participation index, a motor cortex activation index, a brain network synergy index, an effective maintenance index, and a resting inhibition index. The subject's brain state is graded and assessed based on the pelvic floor muscle neurorehabilitation score, and the neurofeedback training strategy is dynamically adjusted based on the assessment results.

[0022] Thus, the method provided by this invention constructs brain state recognition indicators through multi-domain joint construction, identifies pelvic floor muscle movement intentions, constructs brain state evaluation indicators, and combines adaptive neurofeedback to personalize and adjust feedback strategies according to different brain states, effectively promoting neuroplasticity, compliance, and effectiveness in the pelvic floor muscle rehabilitation process. Attached Figure Description

[0023] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0024] The structures, proportions, sizes, etc. illustrated in this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed herein, and are not intended to limit the conditions under which the present invention can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size, without affecting the effects and objectives that the present invention can produce, should still fall within the scope of the technical content disclosed in the present invention.

[0025] Figure 1 A flowchart of the pelvic floor muscle exercise brain state assessment method provided by the present invention; Figure 2 This is a flowchart of the pelvic floor muscle movement brain state assessment method in an embodiment of the present invention; Figure 3 The graph shows the time-frequency analysis results under different task conditions in the embodiments of the present invention; Figure 4This is a source space functional connectivity matrix diagram under different task conditions in the embodiments of the present invention; Figure 5 This is a structural block diagram of a computer device provided by the present invention. Detailed Implementation

[0026] The following specific embodiments illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. 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.

[0027] The existing technology has the following main problems: 1. Existing rehabilitation techniques can only train and assess pelvic floor muscle activity, lacking the perception of central nervous system status; electromyographic signals are peripheral physiological signals, which can only reflect muscle contraction activity, not whether the patient has the intention to move, and cannot assess the activation of the motor cortex.

[0028] 2. Existing methods for motor intention recognition are relatively simple and cannot fully reflect the brain functional activity during pelvic floor movement. Common methods use single time-frequency spatial domain methods such as PSD, ERD / ERS, and CSP, which can only display the power changes of the spectrum or the results of spatial filtering. They are difficult to characterize the time-frequency variation of EEG signals and the information interaction between brain regions, and are insufficient to stably characterize the pelvic floor muscle movement process.

[0029] 3. Current motor intention recognition primarily focuses on classification results, such as whether a patient wants to move their left or right hand. After identification, control is achieved through externally connected devices, lacking a comprehensive assessment of brain state, such as the level of brain function recovery and brain activation.

[0030] 4. Existing pelvic floor muscle rehabilitation training methods lack individualized adaptability; there are significant individual differences in brain state indicators and pelvic floor muscle contraction ability among different patients. Currently, relatively standardized parameters are usually used for training, which lacks adjustment of individual training parameters, resulting in poor long-term training effects.

[0031] In summary, existing pelvic floor muscle rehabilitation techniques mainly suffer from problems such as reliance on peripheral electromyographic signals, limited EEG methods for motor intention recognition, lack of comprehensive brain state evaluation mechanisms, and insufficient personalized training. Therefore, it is necessary to propose a brain-computer interface method and system that can integrate multiple EEG features to achieve comprehensive evaluation of brain state and support individualized pelvic floor muscle rehabilitation training.

[0032] To address the aforementioned technical problems, this invention provides a method and system for assessing brain state during pelvic floor muscle movement based on EEG fusion characteristics.

[0033] In one specific implementation, such as Figure 1 As shown, the pelvic floor muscle motor brain state assessment method based on EEG fusion features provided by this invention includes the following steps: S110: Collect multi-channel EEG signals from the subject during the pelvic floor muscle rehabilitation training task, and preprocess the EEG signals; S120: Perform time-frequency feature analysis on the preprocessed EEG signals to extract time-frequency features related to pelvic floor muscle movement; perform source-space functional connectivity analysis on the preprocessed EEG signals to extract functional connectivity features between different brain regions; S130: The time-frequency features are fused with the functional connectivity features, and a pelvic floor muscle brain state assessment model is constructed based on the fused multidimensional features. The pelvic floor muscle nerve rehabilitation score is calculated through the brain state assessment model. The pelvic floor muscle nerve rehabilitation score includes at least the task participation index, motor cortex activation index, brain network synergy index, effective maintenance index, and resting inhibition index. S140: The subject's brain state is graded and assessed based on the pelvic floor muscle neurorehabilitation score, and the neurofeedback training strategy is dynamically adjusted based on the assessment results.

[0034] Thus, the method provided by this invention synchronously collects multi-channel EEG signals during pelvic floor muscle rehabilitation training and integrates time-frequency features with source-space functional connectivity features to construct a brain state assessment model, transforming the existing single peripheral muscle activity monitoring into a comprehensive perception of central nervous system status. By utilizing multi-dimensional quantitative indicators of brain status, combining rehabilitation dimensions such as task participation, motor cortex activation, brain network synergy, sustained maintenance, and resting inhibition, it overcomes the shortcomings of existing technologies that rely solely on EMG signals and cannot reflect motor intention and brain function activation, providing an objective and quantitative central assessment basis for pelvic floor muscle nerve rehabilitation training.

[0035] In step S110 above, multi-channel electroencephalogram (EEG) signals are collected during the subject's pelvic floor muscle rehabilitation training task, and the EEG signals are preprocessed, specifically including: The pelvic floor muscle rehabilitation training task includes a left pelvic floor muscle stimulation task, a right pelvic floor muscle stimulation task, an active pelvic floor muscle contraction task, and a resting task. The left and right pelvic floor muscle stimulation tasks are achieved through artificial physical stimulation by vaginal digital examination by a gynecological rehabilitation therapist. The active pelvic floor muscle contraction task is achieved by the subject actively contracting the pelvic floor muscles, and the resting task is achieved by the subject maintaining a relaxed state. EEG signals are acquired synchronously using a multi-channel EEG acquisition device, with a sampling rate not lower than a preset sampling frequency; the preprocessing includes, in sequence: downsampling, bandpass filtering, power frequency notch filtering, segmentation, artifact removal, average reference, and baseline correction.

[0036] Thus, the method provided by this invention achieves comprehensive acquisition of EEG responses under different pelvic floor muscle movement states through four standardized rehabilitation training tasks: bilateral stimulation, active exertion, and resting. It utilizes a combination of artificial physical stimulation and active contraction via vaginal digital examination by gynecological rehabilitation therapists, accommodating both passive induction and active intention rehabilitation scenarios. By sequentially executing a standardized preprocessing pipeline of downsampling, bandpass filtering, power frequency notch filtering, segmentation, artifact removal, averaging reference, and baseline correction, it reduces non-neurophysiological noise such as power frequency interference, baseline drift, and electrode artifacts, providing a high-quality, high signal-to-noise ratio EEG data foundation for subsequent time-frequency analysis and functional connectivity calculations.

[0037] In step S120 above, time-frequency feature analysis is performed on the preprocessed EEG signal to extract time-frequency features related to pelvic floor muscle movement, specifically including: The Morlet wavelet transform was applied to each segment of the single-channel EEG signal to obtain the instantaneous power in the time-frequency domain. Using the signal power of the baseline window before the task as a reference, calculate the event-related spectral perturbation; The power variation features of the alpha band associated with sensorimotor cortex activation and the beta band associated with motor preparation and execution are extracted from the event-related spectral perturbation and used as the time-frequency features.

[0038] Thus, the method provided by this invention achieves a joint high-resolution characterization of EEG signals in both time and frequency dimensions by employing Morlet wavelet transform for time-frequency analysis. This overcomes the limitation of traditional power spectrum analysis, which can only reflect frequency domain energy changes and cannot locate time-varying features. By calculating event-related spectral perturbations with the pre-task baseline window signal power as a reference, it can quantify the dynamic power changes of each frequency band relative to the resting baseline during task execution. By focusing on the power change characteristics of the alpha and beta bands, it directly corresponds to the event-related desynchronization and event-related synchronization phenomena in the sensorimotor cortex, providing time-frequency features with clear neurophysiological significance for assessing local cortical activation related to pelvic floor muscle movement.

[0039] In step S120 above, source-space functional connectivity analysis is performed on the preprocessed EEG signal to extract functional connectivity features between different brain regions, specifically including: A forward model is constructed based on the standard template brain and the boundary element model. The minimum norm estimation method is used to generate the inverse solution operator, which maps the scalp EEG signal to the current density of the cortical source space to obtain the source space time series signal. The source space was divided into multiple brain regions (ROIs) using standard brain atlases, and the average time series of each ROI was extracted. The Hilbert transform is applied to the average time series of each ROI to obtain the analytical signal, the instantaneous power is calculated and a logarithmic transform is performed to obtain the power envelope; The power envelope of each ROI pair is orthogonalized to reduce volume conduction effect and zero-delay leakage interference. The orthogonalized Pearson correlation coefficient is calculated and converted into Fisher z value after symmetry processing to obtain the ROI horizontal functional connectivity matrix under each frequency band, which serves as the functional connectivity feature.

[0040] Thus, the method provided by this invention constructs a forward model based on a standard template brain and a boundary element model, and uses a minimum norm estimation inverse operator to estimate and map scalp EEG signals to the cortical source space, realizing spatial localization from scalp recording to intracranial neural source activity; it uses standard brain atlases to divide brain regions of interest (ROIs) and extract average time series, realizing the structured organization of source space signals; by orthogonalizing the power envelope, it reduces the brain capacitive conduction effect and zero-delay leakage interference, improving the biological reliability of functional connectivity estimation; finally, it constructs functional connectivity matrices at the ROI level for each frequency band based on Fisher's z-value, providing relatively accurate network features for quantifying the intensity of information interaction between different brain regions, making up for the inability of single time-frequency analysis to characterize brain region coordination.

[0041] In step S130 above, the time-frequency features and the functional connectivity features are fused, and a pelvic floor muscle brain state assessment model is constructed based on the fused multidimensional features. The pelvic floor muscle nerve rehabilitation score is calculated through the brain state assessment model, including: The time-frequency features are used to describe the activation level of local motor cortex, and the functional connectivity features are used to describe the coordinated changes between different brain regions. The pelvic floor muscle neurorehabilitation score (PFNRS) is calculated using the following formula: ; in, The task engagement index assesses whether a subject has entered an effective task state by comparing changes in brain activity power between the resting state and the task state. The motor cortex activation index is used to assess the response intensity of the motor cortex by evaluating the degree of task-related desynchronization. It is a brain network synergy index that assesses the degree of brain region synergy by comparing the strength of functional connectivity between the task state and the resting state. To effectively maintain the index, the duration of sustained brain state activation was assessed by the ratio of effective desynchronization duration to total task duration. The resting inhibition index assesses a subject's relaxation ability by measuring the degree of activation in the resting state. , , , , These are the weighting coefficients for each index.

[0042] In this way, by multidimensionally fusing the time-frequency features describing local motor cortex activation with the functional connectivity features describing brain region synergistic changes, a quantitative representation of the brain state of pelvic floor muscle movement is achieved from local to global, and from time-frequency to network. The constructed pelvic floor muscle neurorehabilitation score PFNRS has five dimensions: comprehensive task participation index, motor cortex activation index, brain network synergy index, effective maintenance index, and resting inhibition index. It covers the complete neurophysiological process in rehabilitation training from task input, cortical response, network synergy, sustained control to relaxation inhibition. This overcomes the one-sidedness of existing technologies that only focus on single classification results or local features, and provides a quantitative assessment tool for the effect of pelvic floor muscle rehabilitation.

[0043] Furthermore, the method also includes: Each index is normalized to a score of 1-100; Based on the normalized scoring range, the subjects' brain states are divided into multiple levels, which include at least: effective neural control state, formation neural control state, insufficient motor activation state, insufficient brain network coordination state, insufficient sustained control state, insufficient relaxation control state, and low-involvement training state. Each level corresponds to different brain state characteristics and feedback strategies.

[0044] In this way, by normalizing each brain state index to a standardized score of 1-100, comparability and intuitiveness between indicators of different dimensions are achieved. Based on the normalized score range, brain states are divided into seven levels: effective neural control, formation of neural control, insufficient motor activation, insufficient brain network coordination, insufficient sustained control, insufficient relaxation control, and low-involvement training, thus achieving a fine classification of the subjects' rehabilitation status. Each level corresponds to clear brain state characteristics and differentiated feedback strategies, providing a clear and operable classification basis for clinical rehabilitation decisions.

[0045] In step S140 above, the subject's brain state is graded and assessed based on the pelvic floor muscle neurorehabilitation score, and the neurofeedback training strategy is dynamically adjusted based on the assessment results, specifically including: When the assessment indicates an effective state of neural control, reinforce the correct neural pattern, increase reward feedback, and increase the training difficulty and retention time. When assessed as a state of insufficient motor activation, enhance motor cortex recruitment, increase motor imagery cues, and improve feedback intensity in the target area; When assessed as a state of insufficient brain network coordination, promote brain network connectivity, increase the number of task repetitions, and provide overall brain state feedback. When assessed as a state of persistent undercontrol, improve stable control ability, prolong training maintenance time, and reduce transient contraction ratio; When the assessment indicates insufficient relaxation control, improve active relaxation ability, switch relaxation training mode, and increase breathing regulation feedback; When the assessment indicates low training participation, increase training participation, re-guide the task, and increase the frequency of prompts.

[0046] In this way, by dynamically matching differentiated neurofeedback training strategies based on the above-mentioned multi-level brain state assessment results, interventions were achieved for different deficiencies such as insufficient motor cortex activation, insufficient brain network coordination, insufficient sustained control, insufficient relaxation control, and low-participation training. When motor cortex activation was insufficient, motor imagery cues were enhanced; when brain network coordination was insufficient, the number of task repetitions was increased; when sustained control was insufficient, the duration of maintenance training was extended; when relaxation control was insufficient, the relaxation training mode was switched and respiratory regulation feedback was added; and when participation was low, task guidance was re-provided. This achieved personalized rehabilitation through assessment, feedback, and adjustment. It overcame the problems of fixed training parameters and lack of individualized adaptability in existing technologies, and significantly improved the pertinence and effectiveness of pelvic floor muscle rehabilitation training.

[0047] In some embodiments, the training process of the pelvic floor muscle-brain state assessment model includes: A training dataset is constructed by collecting multiple sets of EEG signal samples with known task labels. The task labels of the samples include at least left-side stimulation, right-side stimulation, active exertion, and resting. Preprocessing, time-frequency feature analysis and source space functional connectivity analysis are performed sequentially on the EEG signal samples in the training dataset. The time-frequency features and functional connectivity features corresponding to each sample are extracted and fused into a multi-dimensional feature vector. A multi-task neural network model is constructed, which includes a task state classification head, a brain state index regression head, and a graded evaluation result classification head. The task state classification head is used to output the classification probability of four types of tasks, the brain state index regression head is used to output the regression prediction values ​​of TPI, MAI, BCI, EMI, and RII, and the graded evaluation result classification head is used to output the prediction results of brain state level. Using the multidimensional feature vector of the training dataset as input and the corresponding task label as supervision signal, a multi-task joint loss function is defined. The multi-task joint loss function is composed of a weighted sum of task state classification loss, regression loss of each brain state index, and classification loss of the graded evaluation result. The model parameters are iteratively updated using the backpropagation algorithm until the model converges, resulting in a trained pelvic floor muscle and brain state assessment model, which is then used to perform the pelvic floor muscle and brain state assessment and neural feedback phase based on pre-trained features.

[0048] Thus, during training, a three-head multi-task neural network model was constructed, comprising a task state classification head, a brain state index regression head, and a classification head for graded evaluation results. This enabled the joint optimization of the three objectives: task recognition, index regression, and state grading. By utilizing a multi-task joint loss function to weightedly fuse classification loss, regression loss, and grading loss, the model can predict the continuous values ​​and discrete levels of each brain state index while learning task differentiation, thereby improving the sharing and generalization ability of feature representations. Through an end-to-end training process, the model can directly output comprehensive evaluation results from the original EEG feature vectors.

[0049] In the above specific embodiments, the pelvic floor muscle exercise brain state assessment method based on EEG fusion features provided by the present invention collects multi-channel EEG signals during the subject's pelvic floor muscle rehabilitation training task, and preprocesses the EEG signals; performs time-frequency feature analysis on the preprocessed EEG signals to extract time-frequency features related to pelvic floor muscle exercise; performs source-space functional connectivity analysis on the preprocessed EEG signals to extract functional connectivity features between different brain regions; fuses the time-frequency features with the functional connectivity features, constructs a pelvic floor muscle brain state assessment model based on the fused multidimensional features, calculates a pelvic floor muscle neurorehabilitation score through the brain state assessment model, the pelvic floor muscle neurorehabilitation score includes at least a task participation index, a motor cortex activation index, a brain network synergy index, an effective maintenance index, and a resting inhibition index; grades and assesses the subject's brain state according to the pelvic floor muscle neurorehabilitation score, and dynamically adjusts the neurofeedback training strategy based on the assessment results.

[0050] Thus, the method provided by this invention constructs brain state recognition indicators through multi-domain joint construction, identifies pelvic floor muscle movement intentions, constructs brain state evaluation indicators, and combines adaptive neurofeedback to personalize and adjust feedback strategies according to different brain states, effectively promoting neuroplasticity, compliance, and effectiveness in the pelvic floor muscle rehabilitation process.

[0051] In addition to the methods described above, this invention also provides a pelvic floor muscle movement brain state assessment system based on EEG fusion features, the system comprising: The signal acquisition and preprocessing module is used to acquire multi-channel electroencephalogram (EEG) signals during the subject's pelvic floor muscle rehabilitation training task, and to preprocess the EEG signals. The time-frequency feature extraction module is used to perform time-frequency feature analysis on the preprocessed EEG signal. It uses continuous wavelet transform to calculate the event-related spectral perturbation and extracts the power change features of the alpha and beta bands. The functional connectivity analysis module is used to perform source space functional connectivity analysis on preprocessed EEG signals. Based on a standard template brain, a forward model is constructed and an inverse solution operator is generated. Scalp signals are mapped to the cortical source space. After dividing the brain regions into ROIs, the orthogonalized power envelope correlation method is used to calculate the functional connectivity matrix at the ROI level. The feature fusion and evaluation module is used to fuse the time-frequency features with the functional connectivity features, and calculate the pelvic floor muscle nerve rehabilitation score based on the fused multidimensional features. The pelvic floor muscle nerve rehabilitation score includes at least the task participation index, motor cortex activation index, brain network synergy index, effective maintenance index, and resting inhibition index. The neurofeedback module is used to grade and assess the subject's brain state based on the pelvic floor muscle neurorehabilitation score, and dynamically adjust the neurofeedback training strategy based on the assessment results to achieve personalized pelvic floor muscle rehabilitation training.

[0052] To facilitate understanding, the following will be combined with... Figures 2-4 The overall technical solution and implementation process of the present invention are illustrated through an embodiment. Figure 3 The time-frequency analysis results under different task conditions are shown in the figure. Active exertion can induce a clearer and more sustained decrease in the power of sensorimotor cortex rhythms in the alpha and beta bands, which persists during task execution; left and right stimulation can obtain relatively brief and scattered alpha band modulation. Figure 4 The source space functional connectivity matrix diagrams are presented under different task conditions. Left and right side stimuli yield a generally similar functional connectivity structure, while active stimulation shows a wider functional coupling of the frontal lobe, temporal lobe, and limbic system.

[0053] It should be understood that the algorithm provided by this invention relies on multiple modules for hardware implementation, specifically including: Module 1: EEG Signal Acquisition; This module is used to acquire the electroencephalogram (EEG) activity signals of subjects during pelvic floor muscle rehabilitation training tasks, and simultaneously record task-related events to provide data and labels for subsequent brain state analysis. The pelvic floor muscle training tasks mainly include left and right pelvic floor muscle stimulation tasks, active pelvic floor muscle contraction tasks, and resting tasks. Left and right stimulation are achieved through artificial physical stimulation by a gynecologist using a professional vaginal examination. Active pelvic floor muscle contraction involves subjects actively locating the pelvic floor muscles and performing contraction movements. Resting tasks involve subjects relaxing without performing any tasks. Neuroscan's EEG equipment is used to simultaneously acquire 64 channels of EEG signals at a sampling rate of 1000Hz. During acquisition, the impedance of the EEG signals is reduced to below 30KΩ.

[0054] Module 2: EEG Data Preprocessing; This module is used to reduce non-neural activity such as artifacts in the raw signal and improve the quality of the EEG signal. The preprocessing process is as follows: First, the raw EEG data is downsampled to 250Hz. Then, an FIR filter is used to perform a 1-60Hz bandpass filter on the data, followed by a 50Hz notch filter. The data is then segmented from -1 to 10 seconds, and abnormal artifact segments are removed. A mean reference is then applied, and baseline correction is performed on the segmented data using baseline data between -500 and 0.

[0055] Module 3: Time-Frequency Feature Analysis Module; This module analyzes the dynamic characteristics of EEG signals during pelvic floor muscle training, focusing on their time-frequency variations over time and frequency. Traditional EEG power analysis primarily reflects the overall power change within a specific frequency band over a certain time range, making it difficult to simultaneously characterize the timing of power changes and their corresponding frequencies. This solution employs time-frequency analysis to obtain the dynamic characteristics of EEG signals over time and frequency. The time-frequency analysis is implemented using Morlet wavelet transform. For each preprocessed segment, the event-related spectral perturbation is calculated. The Morlet wavelet transform formula is as follows: ; in, For the first A fragment of brainwave, For frequency, For time, For frequency The corresponding complex conjugate of the complex Morlet wavelet. The number of wavelet periods varies with frequency according to... Adaptive settings ensure that wavelets corresponding to different analysis frequencies maintain approximately consistent time lengths, achieving a balance between temporal localization and frequency resolution. The time-frequency power of all EEG segments under the same training conditions is averaged. ; in, The number of effective EEG segments under this training condition. For the first A fragment in time and frequency The instantaneous power at the point. Using the average power from -0.5 to 0 seconds before the start of the task as the limit, the event-related spectral perturbation is calculated: ; The baseline before the task serves as a reference for event-related spectral perturbations. Sixty analysis frequency points were linearly set within the 2-40 Hz range. After calculating the ERSP, the focus was on the alpha (8-13 Hz) frequency associated with sensorimotor cortex activation and the beta (13-30 Hz) frequency band associated with motor preparation and execution. These frequencies were used to characterize sensorimotor cortex activity related to motor preparation and execution during pelvic floor muscle training and to compare differences between left-stimulation, right-stimulation, active exertion, and resting states.

[0056] Module 4: Source Spatial Functional Connectivity Analysis Module; This module is used to supplement the information interaction relationship between different brain regions based on the analysis in the time and frequency domains, and to provide a source-localized EEG functional connectivity analysis method. According to previous research, the execution of pelvic floor muscle movements is not only completed by the sensorimotor cortex alone, but also requires the motor cortex, prefrontal cortex and subcortical regions to work together.

[0057] Source localization utilizes the fsaverage standard template brain (MNI standard space) built into the MNE, employing a pre-computed boundary element model and source space model, combined with the BEM model to construct a forward model, mapping the source space dipole to EEG electrode locations. A baseline window (-1 to 0 seconds) is used to calculate the noise covariance. A minimum norm estimation method is employed to generate an inverse operator, establishing the mapping relationship between scalp signals and cortical source space current density. The inverse operator is applied to segments under all conditions to extract the event sequence signal of the source space for each segment.

[0058] Using the standard-defined Desikan-Killiany brain atlas, the source space signals were divided into different brain regions, including the left and right frontal lobes, left and right parietal lobes, left and right sensorimotor areas, left and right temporal lobes, left and right occipital areas, and left and right limbic areas. For each segment of the source time series, the average time series was extracted according to the different brain regions, and output as the averaged time series in the form of brain region ROIs.

[0059] Based on the ROI time series, an orthogonalized power envelope correlation method was employed. This method effectively reduces EEG signal source tracing issues such as volume conduction effects and zero-delay leakage interference. Therefore, this mature method was used in the source tracing analysis of pelvic floor muscle EEG data. A correlation matrix at the ROI level was constructed based on different frequency bands: theta (4-8Hz), alpha (8-13Hz), beta (13-30Hz), and gamma (30-60Hz). The orthogonalized power envelope correlation method was calculated as follows: Performing a Hilbert transform on the filtered signal yields the analytic signal: ; in, This is a real signal. This represents the imaginary part after the Hilbert transform.

[0060] Calculate the instantaneous power of the analytic signal: ; Taking the logarithmic transform of the power yields the log power envelope, where Use a small constant to prevent log(0): To reduce zero-delay leakage and volumetric conduction effects, each pair of ROIs was orthogonalized.

[0061] For ROI i and ROI j, calculate the orthogonal components in both directions: , ; Calculate the orthogonalized logarithmic power envelope based on orthogonal analytic signals: , ; z-score normalization of the log power envelope: ; Calculate the Pearson correlation coefficient in both directions. , ; Symmetry reduction (eliminating directionality): ; The correlation coefficient r is converted into a Fisher z-value, which approximately follows a normal distribution, facilitating subsequent statistical tests. ; Module 5: Feature Fusion and Modeling; This module fuses time-frequency features and functional connectivity features to address the limitation of single EEG indicators in fully describing the pelvic floor muscle movement state. Time-frequency features are used to describe local motor cortex activation, while functional connectivity features are used to describe coordinated changes in brain regions. This approach allows for a comprehensive characterization of pelvic floor muscle neural activity that includes temporal, frequency, and network information.

[0062] Module Six: Pelvic Floor Muscle Brain State Assessment Module; This module is the core of this technical solution. It not only identifies whether the current subject's state is under left-hand or right-hand stimulation, active exertion or rest, but also comprehensively assesses whether the current brain state is in an effective rehabilitation state. By constructing multiple evaluation indicators such as task participation index, motor cortex activation level, brain network synergy level, effective training dose, and resting inhibition ability, it achieves a quantitative assessment of the Pelvic Floor Neural Rehabilitation Score (PFNRS). .

[0063] The Task Participation Index (TPI) primarily assesses whether a subject has transitioned from a resting state to effective pelvic floor muscle neural control. It compares changes in brain activity between the resting and active exertion states, using the ERSP results calculated by the aforementioned modules, to evaluate whether the subject has entered a task-oriented state. A high TPI indicates that the patient can mobilize the neural resources of the pelvic floor muscles, while a low TPI indicates that the patient cannot. The expression for TPI is: .

[0064] The Motor Activation Index (MAI) focuses on the decrease in task-state ERD (Extreme Response Discontinuity) of the motor cortex compared to baseline, i.e., pelvic floor muscle-related motor desynchronization, and is used to assess whether the patient's motor cortex is responding effectively. The expression for MAI is: .

[0065] The Brain Coordination Index (BCI) is used to assess the degree of brain region coordination, considering that pelvic floor muscle activity involves more than just the motor cortex. It compares the strength of functional connectivity in task-oriented and resting states. The focus is on the prefrontal cortex, sensorimotor cortex, and default mode network, particularly in the beta band. Enhanced functional connectivity in these areas indicates increased brain network coordination during active pelvic floor muscle contraction. The BCI is expressed as follows: .

[0066] The Effective Maintenance Index (EMI) is used because the active contraction of the pelvic floor muscles is a continuous process, not an instantaneous one. Therefore, it is necessary to assess the duration of sustained brain activation, not just the degree of activation. The EMI is measured as the ratio of the effective ERD duration to the total task duration. The expression for EMI is: .

[0067] The Rest Inhibition Index (RII) is used to control for ERD (Excessive Deficit Damage) at rest and ensure the reliability of the task-oriented assessment. For example, many patients with pelvic floor dysfunction experience persistent tension, resulting in high levels of ERD and functional connectivity activation at rest, hindering effective relaxation. Therefore, the RII is considered to calibrate the rehabilitation score. The expression for RII is: .

[0068] Module Seven: Neurofeedback-Guided Rehabilitation Training Module; This module dynamically adjusts training feedback strategies based on the brain state assessment results provided in the preceding modules, promoting patients to enter an effective brain-pelvic floor muscle collaborative rehabilitation mode. Different training states are determined through pelvic floor muscle neurorehabilitation scores, and then corresponding feedback strategies are implemented. Specifically, by converting each assessment score into a scale of 1-100, rules are established based on the score range to achieve a comprehensive evaluation of indicators such as the neural involvement process, motor cortex activation level, brain network synergy level, and effective maintenance level during the patient's rehabilitation training process. This promotes more specific and personalized rehabilitation training guided by different brain state assessments.

[0069] The rehabilitation status assessment and feedback strategies are shown in the table below:

[0070] Thus, the method and system provided by this invention integrate time-frequency characteristics and functional connectivity characteristics during pelvic floor muscle rehabilitation, while considering both local brain region activity and inter-brain region coordination. It proposes a comprehensive quantitative score for the pelvic floor muscle rehabilitation process, including task participation, motor cortex activation, brain region synergy, effective maintenance, and resting inhibition scores, providing a more comprehensive characterization of the pelvic floor muscle rehabilitation process. This application develops personalized feedback strategies based on brain state index scores. When motor cortex activation is insufficient, it enhances motor cortex imagery; when the effective maintenance index is insufficient, it increases training duration to maintain contraction and force exertion; and when the resting inhibition index is insufficient, it increases relaxation training and promotes breathing regulation. By adjusting the feedback strategy according to different brain states, the adaptability and effectiveness of the rehabilitation process can be better improved.

[0071] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 5As shown, the computer device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and model predictions. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The model predictions of the computer device store static and dynamic information data. The network interface of the computer device is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the steps in the above method embodiments.

[0072] Those skilled in the art will understand that Figure 5 The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the computer device to which the present invention is applied. A specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0073] Corresponding to the above embodiments, this invention also provides a computer storage medium containing one or more program instructions. These one or more program instructions are used to execute the method described above.

[0074] The present invention also provides a computer program product, the computer program product including a computer program, the computer program being stored on a non-transitory computer-readable storage medium, and the computer being able to perform the above-described method when the computer program is executed by a processor.

[0075] In this embodiment of the invention, the processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0076] The various methods, steps, and logic diagrams disclosed in the embodiments of this invention can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this invention can be directly implemented by a hardware decoding processor, or implemented by a combination of hardware and software modules in the decoding processor. The software modules can reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The processor reads information from the storage medium and, in conjunction with its hardware, completes the steps of the above methods.

[0077] The storage medium can be memory, such as volatile memory or non-volatile memory, or may include both volatile and non-volatile memory.

[0078] Among them, non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory.

[0079] Volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous linked dynamic random access memory (Synchlink DRAM, SLDRAM), and direct memory bus RAM (DRRAM).

[0080] The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0081] Those skilled in the art will recognize that, in one or more of the examples above, the functions described in this invention can be implemented using a combination of hardware and software. When applied as software, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of computer programs from one place to another. Storage media can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0082] The above specific embodiments further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made on the basis of the technical solution of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for assessing the brain state of pelvic floor muscle movement based on EEG fusion characteristics, characterized in that, The method includes: Multichannel electroencephalogram (EEG) signals were collected from the subjects during the pelvic floor muscle rehabilitation training task, and the EEG signals were preprocessed. Time-frequency feature analysis was performed on the preprocessed EEG signals to extract time-frequency features related to pelvic floor muscle movement; source-space functional connectivity analysis was performed on the preprocessed EEG signals to extract functional connectivity features between different brain regions. The time-frequency features and the functional connectivity features are fused together, and a pelvic floor muscle brain state assessment model is constructed based on the fused multidimensional features. The pelvic floor muscle nerve rehabilitation score is calculated through the brain state assessment model. The pelvic floor muscle nerve rehabilitation score includes at least the task participation index, motor cortex activation index, brain network synergy index, effective maintenance index, and resting inhibition index. The subjects' brain status was graded and assessed based on the pelvic floor muscle neurorehabilitation score, and the neurofeedback training strategy was dynamically adjusted based on the assessment results.

2. The method according to claim 1, characterized in that, Multichannel electroencephalogram (EEG) signals were collected from subjects during pelvic floor muscle rehabilitation training tasks, and the EEG signals were preprocessed, specifically including: The pelvic floor muscle rehabilitation training task includes a left pelvic floor muscle stimulation task, a right pelvic floor muscle stimulation task, an active pelvic floor muscle contraction task, and a resting task. The left and right pelvic floor muscle stimulation tasks are achieved through artificial physical stimulation by vaginal digital examination by a gynecological rehabilitation therapist. The active pelvic floor muscle contraction task is achieved by the subject actively contracting the pelvic floor muscles, and the resting task is achieved by the subject maintaining a relaxed state. EEG signals are acquired synchronously using a multi-channel EEG acquisition device, with a sampling rate not lower than a preset sampling frequency; the preprocessing includes, in sequence: downsampling, bandpass filtering, power frequency notch filtering, segmentation, artifact removal, average reference, and baseline correction.

3. The method according to claim 1, characterized in that, Time-frequency feature analysis was performed on the preprocessed EEG signals to extract time-frequency features related to pelvic floor muscle movement, specifically including: The Morlet wavelet transform was applied to each segment of the single-channel EEG signal to obtain the instantaneous power in the time-frequency domain. Using the signal power of the baseline window before the task as a reference, calculate the event-related spectral perturbation; The power variation features of the alpha band associated with sensorimotor cortex activation and the beta band associated with motor preparation and execution are extracted from the event-related spectral perturbation and used as the time-frequency features.

4. The method according to claim 1, characterized in that, Source-space functional connectivity analysis was performed on the preprocessed EEG signals to extract functional connectivity features between different brain regions, specifically including: A forward model is constructed based on the standard template brain and the boundary element model. The minimum norm estimation method is used to generate the inverse solution operator, which maps the scalp EEG signal to the current density of the cortical source space to obtain the source space time series signal. The source space was divided into multiple brain regions (ROIs) using standard brain atlases, and the average time series of each ROI was extracted. The Hilbert transform is applied to the average time series of each ROI to obtain the analytical signal, the instantaneous power is calculated and a logarithmic transform is performed to obtain the power envelope; The power envelope of each ROI pair is orthogonalized to reduce volume conduction effect and zero-delay leakage interference. The orthogonalized Pearson correlation coefficient is calculated and converted into Fisher z value after symmetry processing to obtain the ROI horizontal functional connectivity matrix under each frequency band, which serves as the functional connectivity feature.

5. The method according to claim 1, characterized in that, The time-frequency features and the functional connectivity features are fused together, and a pelvic floor muscle brain state assessment model is constructed based on the fused multidimensional features. The pelvic floor muscle neurological rehabilitation score is calculated using the brain state assessment model, including: The time-frequency features are used to describe the activation level of local motor cortex, and the functional connectivity features are used to describe the coordinated changes between different brain regions. The pelvic floor muscle neurorehabilitation score (PFNRS) is calculated using the following formula: ; in, The task engagement index assesses whether a subject has entered an effective task state by comparing changes in brain activity power between the resting state and the task state. The motor cortex activation index is used to assess the response intensity of the motor cortex by evaluating the degree of task-related desynchronization. It is a brain network synergy index that assesses the degree of brain region synergy by comparing the strength of functional connectivity between the task state and the resting state. To effectively maintain the index, the duration of sustained brain state activation was assessed by the ratio of effective desynchronization duration to total task duration. The resting inhibition index assesses a subject's relaxation ability by measuring the degree of activation in the resting state. , , , , These are the weighting coefficients for each index.

6. The method according to claim 5, characterized in that, The method further includes: Each index is normalized to a score of 1-100; Based on the normalized scoring range, the subjects' brain states are divided into multiple levels, which include at least: effective neural control state, formation neural control state, insufficient motor activation state, insufficient brain network coordination state, insufficient sustained control state, insufficient relaxation control state, and low-involvement training state. Each level corresponds to different brain state characteristics and feedback strategies.

7. The method according to claim 6, characterized in that, The subjects' brain status was graded and assessed based on the pelvic floor muscle neurorehabilitation score, and the neurofeedback training strategy was dynamically adjusted based on the assessment results, specifically including: When the assessment indicates an effective state of neural control, reinforce the correct neural pattern, increase reward feedback, and increase the training difficulty and retention time. When assessed as a state of insufficient motor activation, enhance motor cortex recruitment, increase motor imagery cues, and improve feedback intensity in the target area; When assessed as a state of insufficient brain network coordination, promote brain network connectivity, increase the number of task repetitions, and provide overall brain state feedback. When assessed as a state of persistent undercontrol, improve stable control ability, prolong training maintenance time, and reduce transient contraction ratio; When the assessment indicates insufficient relaxation control, improve active relaxation ability, switch relaxation training mode, and increase breathing regulation feedback; When the assessment indicates low training participation, increase training participation, re-guide the task, and increase the frequency of prompts.

8. The method according to claim 1, characterized in that, The training process of the pelvic floor muscle brain state assessment model includes: A training dataset is constructed by collecting multiple sets of EEG signal samples with known task labels. The task labels of the samples include at least left-side stimulation, right-side stimulation, active exertion, and resting. Preprocessing, time-frequency feature analysis and source space functional connectivity analysis are performed sequentially on the EEG signal samples in the training dataset. The time-frequency features and functional connectivity features corresponding to each sample are extracted and fused into a multi-dimensional feature vector. A multi-task neural network model is constructed, which includes a task state classification head, a brain state index regression head, and a graded evaluation result classification head. The task state classification head is used to output the classification probability of four types of tasks, the brain state index regression head is used to output the regression prediction values ​​of TPI, MAI, BCI, EMI, and RII, and the graded evaluation result classification head is used to output the prediction results of brain state level. Using the multidimensional feature vector of the training dataset as input and the corresponding task label as supervision signal, a multi-task joint loss function is defined. The multi-task joint loss function is composed of a weighted sum of task state classification loss, regression loss of each brain state index, and classification loss of the graded evaluation result. The model parameters are iteratively updated using the backpropagation algorithm until the model converges, resulting in a trained pelvic floor muscle and brain state assessment model, which is then used to perform the pelvic floor muscle and brain state assessment and neural feedback phase based on pre-trained features.

9. A pelvic floor muscle motor brain state assessment system based on EEG fusion features, characterized in that, The system includes: The signal acquisition and preprocessing module is used to acquire multi-channel electroencephalogram (EEG) signals during the subject's pelvic floor muscle rehabilitation training task, and to preprocess the EEG signals. The time-frequency feature extraction module is used to perform time-frequency feature analysis on the preprocessed EEG signal. It uses continuous wavelet transform to calculate the event-related spectral perturbation and extracts the power change features of the alpha and beta bands. The functional connectivity analysis module is used to perform source space functional connectivity analysis on preprocessed EEG signals. Based on a standard template brain, a forward model is constructed and an inverse solution operator is generated. Scalp signals are mapped to the cortical source space. After dividing the brain regions into ROIs, the orthogonalized power envelope correlation method is used to calculate the functional connectivity matrix at the ROI level. The feature fusion and evaluation module is used to fuse the time-frequency features with the functional connectivity features, and calculate the pelvic floor muscle nerve rehabilitation score based on the fused multidimensional features. The pelvic floor muscle nerve rehabilitation score includes at least the task participation index, motor cortex activation index, brain network synergy index, effective maintenance index, and resting inhibition index. The neurofeedback module is used to grade and assess the subject's brain state based on the pelvic floor muscle neurorehabilitation score, and dynamically adjust the neurofeedback training strategy based on the assessment results to achieve personalized pelvic floor muscle rehabilitation training.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method as described in any one of claims 1-8.