Closed-loop rTMS safety control method and system based on multi-modal neural state

Abstract

The application discloses a closed-loop rTMS safety control method and system based on a multi-modal neural state, relates to the technical field of neural regulation and closed-loop rTMS control, and comprises the following steps: synchronously collecting electroencephalogram (EEG), electromyogram (EMG) and stimulation response signals, extracting EEG state stability indexes, EMG coordination indexes and stimulation response fluctuation indexes; inputting the multi-modal indexes into a closed-loop safety state discrimination model, generating five safety states of permission, delay, downgrading, suspension and recalibration according to multi-criteria; outputting corresponding control instructions according to the state, and automatically recovering and restarting at a low dose after suspension through sliding window monitoring. The system comprises signal collection, feature extraction, safety discrimination, dose control and stimulation execution interface modules. Through multi-modal joint verification and a five-grade classification response mechanism, the application realizes autonomous safety management and control of a closed-loop rTMS stimulation process, and improves abnormal state recognition and classification disposal capability.

Description

Technical Field

[0001] This invention relates to the field of neural modulation and closed-loop rTMS control technology, and in particular to a closed-loop rTMS safety control method and system based on multimodal neural states. Background Technology

[0002] Repetitive transcranial magnetic stimulation (rTMS) is a non-invasive and painless neuromodulation technique that has been widely used in fields such as motor function rehabilitation. In recent years, a closed-loop intervention paradigm that uses electroencephalography (EEG) to detect endogenous rhythmic oscillations in real time and uses these oscillations as biomarkers to drive rTMS stimulation output is emerging as a promising direction in this field. This paradigm is characterized by its ability to couple the stimulation timing with the subject's real-time neural state, offering higher precision compared to open-loop interventions with fixed parameters.

[0003] Post-Movement Beta Synchronization (PMBS) can be stably induced through voluntary movements such as index finger extension, and it is associated with the Primary Motor Cortex (M1) region. It is closely related to the short-term inhibitory state of the cortex mediated by γ-aminobutyric acid (GABA) and has become one of the core driving biomarkers of the closed-loop rTMS regulatory system.

[0004] However, current closed-loop rTMS systems, which use rhythmic oscillations such as PMBS as the core driving signals, have mainly solved two core problems: stimulation triggering timing and dynamic adjustment of stimulation parameters based on biomarker amplitude. However, significant shortcomings remain in the intelligent safety management and control during the stimulation process, and a complete safety control system has not yet been formed. These deficiencies are mainly manifested in four aspects.

[0005] First, the assessment of biomarker homeostasis is limited to a single dimension. Existing systems interpret biomarkers based on a single indicator (such as whether beta power recovers to baseline levels within a fixed time window), failing to comprehensively verify the physiological stability and repeatability of the induced PMBS signal across multiple dimensions. In actual experiments, factors such as noise contamination, poor electrode contact, subject inattention, premature muscle relaxation, or overflow movements can all lead to non-steady-state distortions in the PMBS signal. Existing systems cannot identify these abnormal signal characteristics and will still trigger stimuli according to pre-set logic, applying stimulation to atypical neural states and creating potential risks of neuromodulation.

[0006] Secondly, there is a lack of online cross-validation mechanisms for multiple signals. EEG signals are susceptible to contamination from multiple sources, including eye movement artifacts, electrocardiogram artifacts, electromyography interference, and environmental power frequency noise, making it difficult to guarantee the authenticity and stability of the signals in a single way. Existing closed-loop systems rely solely on EEG spectral changes to determine the "motor execution - PMBS induction" event chain, without combining actual motor behavior signals for collaborative verification. This makes it impossible to identify false positive signals caused by EEG artifacts, resulting in the problem that the system may erroneously trigger stimuli even when the subject has not actually performed effective movement, leading to a decrease in the effectiveness and accuracy of stimulus intervention.

[0007] Third, it lacks the ability to monitor and regulate the closed-loop neural response after stimulation. rTMS stimulation directly affects the cerebral cortex, influencing its excitability. Repeated stimulation may cause a series of neural state changes, such as abnormally increased cortical excitability, prolonged PMBS baseline recovery time, and abnormal fluctuations in motor evoked potentials (MEPs). However, existing systems only follow preset total pulse parameters to complete a fixed stimulation process, and cannot monitor the neural response characteristics after stimulation in real time, lacking a safe closed loop of "detection-judgment-response".

[0008] Fourth, the stimulation parameter control strategy is crude and lacks a graded safety response mechanism. Existing systems mostly use a fixed step adjustment mode based on the PMBS amplitude and motion speed of the previous trial for stimulation intensity adjustment, resulting in a single control direction and insufficient flexibility. At the same time, the system has not constructed a multi-level safety state machine covering all scenarios and lacks a graded response logic of "allow stimulation - delay stimulation - downgrade stimulation - pause stimulation - recalibrate". When faced with sudden situations such as signal abnormalities and neural state imbalances, it cannot execute differentiated safety graded control strategies, and does not support a closed-loop control process of automatic low-dose restart based on signal stability after pausing stimulation.

[0009] The aforementioned technical deficiencies prevent existing closed-loop rTMS systems from achieving autonomous and safe control throughout the entire process in scenarios such as individual difference adaptation, long-term stimulation, and complex task intervention. They still rely on human supervision by operators to ensure the safety of the stimulation process, making it difficult to meet the requirements of large-scale and standardized autonomous and safe control capabilities.

[0010] Based on this, the present invention aims to provide a closed-loop rTMS safety control method and system based on multimodal neural states. It is based on the stability assessment of multimodal neural-electromyography-stimulation response signals and can perform online safety status judgment, graded dose control, and automatic pause and resumption of the closed-loop rTMS stimulation process, so as to fill the engineering gap in the safety control link of the existing closed-loop rTMS system. Summary of the Invention

[0011] The purpose of this invention is to provide a closed-loop rTMS safety control method and system based on multimodal neural states, and to solve the technical problems existing in the safety control link of the existing closed-loop rTMS system.

[0012] To achieve the above objectives, this invention provides a closed-loop rTMS security control method based on multimodal neural states, comprising the following steps: S0. During the course of a motor task that can induce MRBD / PMBS events, the subject's electroencephalogram (EEG), electromyogram (EMG), and stimulus response signals are collected simultaneously. The stimulus response signals include the MEP amplitude induced by non-therapeutic single-pulse TMS detection and the EEG recovery characteristics within the stimulation interval window. S1. The collected EEG signals, EMG signals, and stimulus response signals are preprocessed and analyzed in time and frequency to extract EEG state stability indicators, EMG synergy indicators, and stimulus response fluctuation indicators, respectively. S2. Input the extracted EEG stability index, EMG synergy index and stimulus response fluctuation index into the closed-loop safety status discrimination model, and generate the closed-loop rTMS safety status of the current trial according to multi-level criterion rules. The safety status includes the allowed stimulation status, delayed stimulation status, degraded stimulation status, paused stimulation status and recalibration status. S3. Output the corresponding rTMS control command according to the safety status: output the stimulation control command according to the current dose level in the allowed stimulation state; suspend the current stimulation in the delayed stimulation state; output the stimulation control command after lowering the dose level by one level according to the preset lowering rule in the de-stressing stimulation state; stop the stimulation output and start the automatic recovery monitoring process in the paused stimulation state; suspend the current stimulation protocol and trigger the parameter re-measurement process in the recalibration state. S4. When in the paused stimulation state, continuously monitor the multimodal state characteristics of subsequent trials. When the safety recovery conditions are met for M consecutive effective trials, reset the dose level to the lowest level and switch to the allowed stimulation state, where M is a positive integer from 10 to 20. The trial-confirmation window is formed within P trials after entering the allowed stimulation state, where P is a positive integer from 10 to 15.

[0013] Preferably, the EEG state stability index includes: the relative deviation of the current trial PMBS peak value from the average PMBS peak value of the most recent N valid trials. The relative deviation of the current PMBS duration from the average PMBS duration of the most recent N valid trials. Spatial concentration of beta-band non-negative energy on sensorimotor cortical electrode groups And the similarity between the current trial frequency spectrum and the individualized historical template. The specific calculation formula is as follows: ; ; ; ; ; ; in, For the first The relative deviation index of the PMBS peak value in each trial. For the first The peak value of PMBS in each trial The average of the peak values ​​of PMBS in the most recent N valid trials. For the first The average duration of PMBS in the most recent N valid trials before each trial. For the first The duration of PMBS in each trial To determine the beta power of the i-th electrode in the sensory motor cortex electrode group. The spectrum during the current PMBS event interval. The template is obtained by averaging the spectrum of the most recent 10 valid PMBS event intervals collected in the previous recalibration process. To avoid constants with a denominator of zero.

[0014] Preferably, the electromyographic synergy index includes: the difference between the electromyographic onset time and the onset time of the electroencephalographic MRBD of the target muscle group. The non-negative deviation of the difference between the peak time of PMBS and the termination time of electromyography relative to the individual baseline delay during the initial calibration phase. and the baseline drift factor of electromyography in non-task channels The specific calculation formula is as follows: ; ; ; ; in, The moment when the amplitude of the electromyographic envelope first continuously exceeds three standard deviations of the baseline noise level for at least 30 ms is defined. The time when the electromyographic envelope amplitude falls back to below twice the baseline noise level and remains there for at least 50 ms is defined as the moment when the electromyographic envelope amplitude falls back to below twice the baseline noise level and remains there for at least 50 ms. The start time of the period in which the beta power continuously drops to 10% below the baseline and lasts for at least 100 ms. The individual baseline recovery delay obtained during the initial calibration phase, The discrete sampling points of the non-task side hand muscle channel within a time window of 500ms and L sampling points; The RMS baseline values ​​are the subjects' continuous 3-minute resting state RMS values ​​recorded during the initial calibration phase. This is the real-time electromyography baseline for the non-task channel. This is the peak time of PMBS. This represents the time difference between the peak time of PMBS and the termination time of electromyography of the target muscle group in the current trial.

[0015] Preferably, the stimulus-response fluctuation index includes: the coefficient of variation of MEP amplitude induced by the most recent K safety-compliant single-pulse TMS probes. The time required for beta power to recover to near baseline after stimulation The slope of recovery time after stimulation in the most recent R trials and the energy ratio of the beta band before and after stimulation The specific calculation expression is as follows: After completing 10 valid stimulus trials, in the next trial, the motor-related... Before the movement-related beta desynchronization (MRBD) phase, a non-therapeutic single-pulse TMS probe with an intensity of 80%–120% of the resting-state motor threshold (RMT; RMT refers to the minimum stimulus intensity that can stably induce the minimum discernible motor evoked potential (MEP) when the subject is in a completely relaxed state of the target muscle using single-pulse TMS stimulation of the motor cortex) is passed, and the MEP amplitude induced by the probe pulse is recorded; the MEP amplitude sequence of the most recent K probes is then taken. calculate: , ; in, Let K be the standard deviation of the MEP amplitude from the most recent K probes. The average of the MEP amplitudes from the most recent K probes. To avoid constants with a denominator of zero; The time required for the beta power to return to near the baseline mean 500 ms before stimulation after each effective stimulus. Take the most recent R trials The recovery time slope was obtained by least squares linear regression of the sequence. : ; Energy ratio of beta band before and after stimulation : ; in, To stimulate beta energy within the first 500ms, This refers to the beta energy within a 500-1000ms window after the stimulus ends. For the most recent The local index in the effective stimulus trial sequence, with a value of [value]. ; For the most recent The effective stimulus test was followed by the local sequence number. The mean, ; For the first The recovery time required for the post-stimulation beta power to recover to near the baseline mean of 500 ms before stimulation in a valid stimulation trial; For the most recent The mean of the recovery time series corresponding to each effective stimulus trial; The number of most recent effective stimulus trials used to calculate the recovery time slope. Number the current effective stimulus trial.

[0016] Preferably, the multi-level judgment rule in S2 is as follows: when , , , , , , , , , When both conditions are met, it is determined to be a permissible stimulus state; When the mild overstepping criterion of the delayed stimulation state is met, it is determined to be a delayed stimulation state. Specifically, when any of the collected EEG stability indicators or EMG synergy indicators oversteps the limit but does not reach the severe overstepping threshold, and the stimulus response fluctuation indicator is still within the allowable stimulation range, it is determined to be a delayed stimulation state. A reduced-order stimulus state is defined as one that meets the cumulative out-of-bounds criterion, specifically: if the same EEG stability index or EMG synergy index exceeds the limit three or more times in the last five consecutive trials, or In the last five detections, the cumulative increase exceeded 25% of the initial value, or The interval has been continuously deviated from in the last 5 trials. When this occurs, it is determined to be a reduced-order stimulus state; A paused stimulation state is determined when the severe out-of-bounds criterion for the paused stimulation state is met or when an artificial safety abort signal is received. Specifically, when... ,or and The increase exceeded 50%, or or ,or The stimulation is considered to be paused when at least 3 consecutive trials or 5 consecutive trials fail to detect a valid MRBD / PMBS, or when any of the following artificial safety stop signals are received: subject discomfort, significant coil position shift, or abnormal device safety interlock: the stimulation is paused. When the cumulative number of effective stimulation trials reaches the preset upper limit, the deviation of the resting motor threshold exceeds 10%, or the electrode contact quality is lower than the qualified threshold, it is determined to be in a recalibration state. Specifically, when the cumulative number of effective stimulation trials reaches 80, or the current RMT deviates from the initial calibrated RMT by more than 10% RMT, or the impedance of any EEG channel exceeds 20kΩ, the contact quality of the key dry electrode channel is lower than the qualified threshold of the equipment, or the EMG baseline RMS changes by more than 3 times compared with the calibration, it is determined to be in a recalibration state.

[0017] Preferably, the downgrade rule includes: shifting the stimulation dose parameter one level in the reverse direction along a preset dose level table; the dose level table includes 5 levels: 60%RMT, 65%RMT, 70%RMT, 75%RMT, and 80%RMT, with a step of 5%RMT between adjacent levels; when the current level is 60%RMT, the control state is switched to the paused stimulation state.

[0018] Preferably, the automatic recovery monitoring process includes: S31A: Do not output rTMS stimulation in the paused stimulation state, but continue to induce MRBD and PMBS events with a motor task that can induce MRBD / PMBS events, and extract multimodal state features according to S1. S32A. Set a sliding window of length M=10. In 10 consecutive valid trials within the sliding window, all EEG state stability indices and EMG synergy indices in each trial meet the permissible stimulus state criterion. It fell back to 0.30 or below, and Return to When the interval is within the safe recovery condition, the dose level is reset to 60%RMT, and the control state is switched to the permissive stimulation state. S33A. Within P=10 trials after automatic recovery, the threshold in the stimulus state criterion is allowed to tighten as follows: , , , , , , If any indicator in the window exceeds the limit, the system will immediately revert to the paused stimulation state.

[0019] Preferably, the re-measurement process includes: S31B, pause rTMS stimulation output; S32B. Re-measure RMT according to the existing RMT measurement method: the minimum stimulation intensity corresponding to at least 5 out of 10 TMS pulses induced by MEP not less than 50μV. S33B. Collect the time-spectral data of MRBD / PMBS event intervals from 15 valid trials. After removing any abnormal trials where the EEG stability index exceeded the limit, take the average time-spectral data of the remaining trials as the new PMBS baseline template. ; S34B, Recalibrate electromyography baseline Update the impedance record; S35B, after calibration, the dose level is reset to 60%RMT, the control state is switched to the allowable stimulation state and enters the trial-confirmation window.

[0020] Preferably, the dosage progression rule is also included: under permissible stimulation conditions, when all 15 consecutive effective stimulation trials of L=15 times meet the permissible stimulation condition criteria, the dosage level is increased by one level in the positive direction; when the dosage level reaches 80% RMT and still meets the permissible stimulation condition criteria, the dosage level is kept at the highest level and no further increase is made.

[0021] This invention also provides a closed-loop rTMS safety control system based on multimodal neural states, comprising: The multimodal signal acquisition module is used to simultaneously acquire electroencephalogram (EEG) signals, electromyogram (EMG) signals, and stimulus-response signals during the subject's performance of a motor task; The multimodal feature extraction module is used to preprocess the acquired multimodal signals and extract multimodal indicators such as EEG state stability indicators, electromyographic coordination indicators, and stimulus response fluctuation indicators. The closed-loop safety status determination module is used to output the closed-loop rTMS safety status according to multimodal indicators and preset multi-level judgment rules. The dose control module is used to determine the control process of allowing stimulation, delaying stimulation, down-order stimulation, pausing stimulation, or recalibrating based on the safety status, and to complete the sliding window safety recovery monitoring and dose increment after automatic pause; The stimulation execution interface module is used to convert the control commands generated by the dose control module into physical trigger signals that drive the rTMS stimulator.

[0022] Therefore, the above-mentioned closed-loop rTMS security control method and system based on multimodal neural states has the following beneficial effects: (1) This invention no longer uses a single trigger criterion of "whether PMBS power returns to baseline". Instead, it conducts a comprehensive quantitative assessment of three multimodal indicators: EEG stability, EMG synergy, and stimulus response fluctuation. This comprehensively verifies the neurophysiological state, motor execution state, and signal quality of the current trial, and engineered judgment to determine whether the current trial meets the safe stimulation conditions. This effectively avoids the risk of accidentally triggering stimulation to atypical neural states, artifact contamination states, or ineffective motor states. This invention improves the online safety control capability of the closed-loop rTMS system, but it is not directly equivalent to improving clinical efficacy.

[0023] (2) The present invention constructs a five-level safety state machine including permitted stimulation, delayed stimulation, downgraded stimulation, paused stimulation, and recalibration, and sets calculable threshold conditions for the criteria corresponding to each level, expanding the traditional control response from "whether to trigger" to "how to grade the response", so that the closed-loop rTMS system can perform dose downgrade, pause or recalibration operations that match the severity of the abnormality when an abnormal event occurs, thus forming a complete engineered safety closed loop.

[0024] (3) The present invention sets up an automatic recovery monitoring process with a sliding window length M=10, a trial-confirmation window length P=10, and a dose progression cumulative length L=15, so that the system can continue to evaluate the stability of the "neuro-electromyography-stimulation" response based on the induced task even when the system is paused, and automatically re-enter low-dose stimulation and increase it according to the dose progression rule after confirming that the safe recovery conditions are met, thus avoiding excessive dependence on the operator during long-term stimulation.

[0025] (4) The dose control of the present invention is based on the basic control law of "stable - progressive, abnormal - downgrade, severe - pause, cumulative - calibration". It is decoupled from the specific stimulation frequency, stimulation site, target brain region and specific rTMS stimulation parameters. It can be used as an independent external safety control layer, which is compatible with various closed-loop rTMS system architectures driven by neural rhythm oscillation. It has strong engineering compatibility and portability and is applicable to a wide range of scenarios.

[0026] (5) The safety control method of the present invention uses the stability of multimodal signals as the sole criterion. It can run online without relying on offline calculation steps such as image registration, brain power source localization or hemispheric causality analysis. The requirements for system computing resources and hardware configuration are comparable to those of conventional closed-loop rTMS systems, and it has the feasibility of direct engineering implementation.

[0027] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0028] Figure 1This is an overall flowchart of the closed-loop rTMS security control method based on multimodal neural states according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the multimodal state feature extraction process according to Embodiment 1 of the present invention; Figure 3 This is a schematic diagram of the state transition of the five-level closed-loop safety state machine according to Embodiment 1 of the present invention; Figure 4 This is a schematic diagram of the automatic pause and automatic resume monitoring process according to Embodiment 1 of the present invention; Figure 5 This is a block diagram of the module composition of the closed-loop rTMS safety control system based on multimodal neural states according to Embodiment 3 of the present invention. Detailed Implementation

[0029] The following detailed description of embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0030] The rTMS stimulator, EEG acquisition device, EMG acquisition device and their synchronization control device involved in this invention are conventional devices in the field, including but not limited to transcranial magnetic stimulators or repetitive transcranial magnetic stimulators with external trigger interfaces, 64-lead wet or dry electrode EEG acquisition systems, and EMG acquisition systems with bio-amplification and analog-to-digital conversion functions; this invention does not limit the specific model, manufacturer and hardware parameters of the hardware devices.

[0031] Example 1: A closed-loop rTMS security control method based on multimodal neural states.

[0032] This embodiment provides a closed-loop rTMS security control method based on multimodal neural states, such as... Figure 1 As shown, it includes the following steps: Step 1: Synchronous acquisition of multimodal signals.

[0033] During the subjects’ performance of a motor task that could induce MRBD / PMBS events, the Neuroscan Quickcap 64 wet electrode EEG cap was used to place the EEG electrodes according to the 10-20 international standard lead system. The sampling rate was set to 1000 Hz, the reference electrodes were bilateral mastoids, the electrode impedance was reduced to below 10 kΩ, and the notch filter was set to 50 Hz. Simultaneously, Ag-AgCl surface self-adhesive disc electrodes (10 mm in diameter) were used to record the electromyography (EMG) signals of the target muscle group (first dorsal interosseous muscle FDI). The signals were then amplified and filtered by a bio-amplifier with a bandpass filter from 5 to 2000 Hz and digitized by an analog-to-digital converter. In addition, during the closed-loop rTMS stimulation intervals, single-pulse TMS probes with 80% RMT intensity were transmitted at a preset cycle (one pulse after every 10 effective stimulations) to record the MEP amplitude induced by the probe pulses.

[0034] This embodiment employs an index finger extension exercise task paradigm: each trial includes one index finger extension exercise, with trial intervals randomized to 3.25–3.75 s; each group contains 20 trials, with a 5-minute interval between groups. This invention does not limit the specific form of the task paradigm; any exercise task paradigm capable of stably inducing MRBD and PMBS events is applicable.

[0035] Step 2: Multimodal state feature extraction.

[0036] like Figure 2 As shown, the collected multimodal signals were preprocessed and subjected to time-frequency analysis to extract EEG state stability indicators, electromyographic synergy indicators, and stimulus response fluctuation indicators.

[0037] Further, extraction of EEG state stability indicators: After performing bandpass filtering (1–45 Hz), power frequency notch filtering (50 Hz), and independent component analysis to remove artifacts from electrooculography (EOG) and electrocardiography (ECG), a Slepian sliding time window with a length of 500 ms and an overlap rate of 50% was used to estimate the power of the beta band (13–30 Hz) of 10 sensoromotor cortical electrodes (C3, C1, Cz, C2, C4, CP3, CP1, CPz, CP2, CP4) using multi-window spectral estimation. The baseline normalized beta power of each electrode was obtained according to formulas (1)–(3). (1) (2) (3) In the formula, For the first The first of the valid trials The sensorimotor cortex EEG channels in time beta power at (beta power and) Figure 2 In The meaning of power representation is the same, derived from the first The estimated time-spectrum power of each channel is obtained by integrating in the 13–30 Hz band; The first window obtained by Slepian multi-window spectral estimation The number of valid trials, the first Each brainwave channel in time ,frequency Time-spectral power estimate at; For the first The number of valid trials, the first The average beta power of each EEG channel within a 500ms baseline window before the motor event is used to characterize the individualized baseline of the trial and the channel. The first after baseline correction The number of valid trials, the first The normalized beta power of each EEG channel reflects the change in beta energy relative to the pre-motor baseline; Indicates the valid trial number. This indicates the EEG channel number of the sensorimotor cortex. Indicates time, Indicates frequency; Using the average beta power within 500ms before the start of the task as the baseline, the period after the task in which the normalized beta power is continuously higher than 0.10 for a duration of not less than 200ms is defined as the PMBS event interval. And extract the PMBS peak value within this interval. (Maximum power point), PMBS peak time and PMBS duration .

[0038] Take N=10 of the most recent valid trials, and calculate the relative deviation of the PMBS peak value according to formulas (4)-(9). PMBS duration deviation index Spatial concentration index Similarity index of spectrum The reason for choosing N=10 is that in the conventional closed-loop rTMS intervention paradigm, the number of trials per group is 20. N=10 can ensure statistical robustness while ensuring that the historical window does not cross groups to avoid bias introduced by inter-group drift.

[0039] Let the peak value of PMBS in the current trial be The set of PMBS peak values ​​from the most recent N valid trials is Where N is 10. To make To accurately reflect the deviation of the current trial from the historical stable template, the historical mean is calculated first. Then calculate the relative deviation of the current trial. : (4) (5) Let the duration of the current trial PMBS be... PMBS duration deviation index Calculate using the following formula: (6) (7) The beta power vectors of sensorimotor cortical electrodes (including 10 electrodes: C3, C1, Cz, C2, C4, CP3, CP1, CPz, CP2, and CP4) were selected. Where M=10. First, apply the following to the vector: Norm normalization makes Then calculate the spatial concentration index using the following formula. : (8) The average time-spectrum data of the most recent 10 PMBS event intervals, collected during the previous recalibration process and deemed valid, is used as the historical template. ; Calculate the spectrum of the current trial. and Cosine similarity within the PMBS event interval is used as an index of spectral morphology similarity. : (9) The above normalization process does not change the relative trends of each indicator and is only used to ensure the stability of online calculation.

[0040] Further, extraction of electromyographic synergistic indicators: After performing bandpass filtering, full-wave rectification, and low-pass envelope extraction with a time constant of 50 ms on the electromyographic (EMG) signal, the moment when the envelope amplitude first continuously exceeds three standard deviations above the baseline noise level for a duration of not less than 30 ms was defined as the EMG onset time. The time when the envelope amplitude falls back to below twice the baseline noise level and remains there for at least 50 ms is defined as the electromyography termination time. The thresholds for determining the start and end of electromyography (EMG) (3 times / 2 times the standard deviation, minimum duration of 30ms / 50ms) are set according to commonly used thresholding methods in the field of EMG signal processing, which can effectively suppress spurious start and end determinations caused by background noise.

[0041] Simultaneously, the onset time of the electroencephalogram (MEG) MRBD corresponding to the onset of electromyography (EMG) is recorded as... The peak time of PMBS is recorded as . The criteria for determination are: the beta power continuously drops to 10% below the baseline and lasts for at least 100ms, and the start time of this period is recorded as _____. .

[0042] The difference between the electromyographic onset time and the MRBD onset time of the target muscle group is calculated according to formulas (10)-(12). The non-negative deviation of the difference between the peak time of PMBS and the termination time of electromyography relative to the individual baseline delay during the initial calibration phase. The PMBS peak typically lags behind the termination of electromyography, therefore Do not adopt directly The absolute time difference is used, while the individual reference delay relative to the initial calibration phase is adopted. The deviation amount is used to avoid misjudging normal physiological lag as abnormal.

[0043] In physiological motor tasks, the onset of electromyography (EMG) typically occurs approximately 30–80 ms after the onset of MRBD, and the peak of PMBS typically occurs approximately 300–600 ms after the termination of EMG; therefore, this embodiment will... Individual baseline delay under the same task paradigm as the initial calibration phase Compare, and with As a delay deviation, this embodiment will allow the stimulus criterion in... and Setting the threshold to 80ms and setting the upper limit of the threshold in the delayed stimulus criterion to 150ms can avoid misjudging normal PMBS lag as abnormal.

[0044] (10) (11) (12) The baseline electromyography of the muscle channels on the non-task side of the hand was calculated according to formula (13). During the initial calibration phase, the RMS baseline values ​​of the subjects were recorded continuously for 3 minutes while they were at rest. When running online As a factor of the baseline drift of electromyography (EMG), the upper limit of the permissible stimulation criterion is set to 1.5 times the resting EMG background.

[0045] (13) Further, stimulus-response fluctuation index extraction: During closed-loop rTMS control, after every 10 effective stimulation trials, a non-therapeutic single-pulse TMS probe is inserted before the MRBD phase of the next trial. The probe intensity is set to 80%–120% RMT, preferably the lowest intensity that can stably induce a discernible MEP without exceeding the limits of existing safety guidelines. This probe pulse is only used to read the subject's current cortical excitability state and does not constitute therapeutic rTMS stimulation. In the initiation phase before K probes are completed, the effective MEP recording from the initial calibration phase can be used to supplement the window, or the MEP amplitude variation coefficient induced by the single-pulse TMS probe can be temporarily not triggered. The criterion for downgrading / pausing.

[0046] Acquire the MEP amplitude sequence induced by the probe pulse, and take the nearest one. The MEP amplitude sequence of the second detection is calculated according to formula (14). . The upper limit of permissible stimulation is set to 0.30, and the severe overstepping threshold is set to 0.50, which serve as the engineering safety thresholds in this embodiment. The above thresholds are not clinical efficacy data or universal physiological constants. In practical applications, adaptive calibration can be performed based on the individual MEP distribution during the initial calibration phase, ethical approval requirements, and device safety specifications.

[0047] , (14) For each effective stimulus, determine the time required for the beta power to return from the end of stimulation to near the baseline mean value 500 ms prior to stimulation. Take the most recent R=5 valid trials The slope of the sequence is calculated according to formula (15). (Unit: s / trial). The maximum allowable stimulation duration is set to 0.05 s / trial, and the severe overshoot threshold is set to 0.10 s / trial. The absolute value increases by more than 50% compared to the initial value; the above thresholds are used for online safety control and are not used as indicators for judging clinical efficacy. The calculation expression is as follows: (15) The energy ratio of the beta band before and after stimulation is calculated according to formula (16). . Allow stimulation interval set to This corresponds to the engineering safety expectation that the beta energy does not shift significantly immediately after stimulation; Values ​​exceeding 1.50 (i.e., an immediate beta energy increase of more than 50% compared to before stimulation) or below 0.50 (i.e., an immediate beta energy decrease of more than 50% compared to before stimulation) are considered serious violations.

[0048] (16).

[0049] Step 3: Closed-loop safety status determination.

[0050] like Figure 3 As shown, the closed-loop safety status discrimination module performs five-level safety status discrimination for the current trial based on formula (17) and the specific threshold settings listed in Table 1. Figure 2 Multimodal state feature vectors Specifically, this involves extracting the corresponding EEG state stability indicators, electromyographic coordination indicators, and stimulus-response fluctuation indicators from the multimodal state features, represented as follows: Multimodal state feature vectors After transpose operation Then, as in equation (17) The amount of input.

[0051] (17) Table 1: Specific Threshold Settings

[0052] The state transition rules can be summarized as follows: Permissive stimulus state → Satisfying mild boundary violation condition → Delayed stimulus state; Any state → satisfying the cumulative out-of-bounds condition → reduced-order stimulus state; If the cumulative exceedance still occurs at the lowest level of the reduced-level stimulation state, then stimulation will be paused. Any state → meets the serious boundary violation condition → pause stimulation state; Pause stimulation → Meet safe recovery conditions → Allow stimulation (lowest level + trial confirmation window); Allow stimulation state → The cumulative number of effective stimulation trials reaches the preset limit or the device / physiological baseline drift exceeds the limit → Recalibration state.

[0053] Step 4: Output of graded stimulus control instructions.

[0054] The safety status output by the closed-loop safety status determination module is transmitted to the dose control module. Based on the current dose level and the safety status, the dose control module executes the following control actions: Under permissible stimulation conditions, the rTMS stimulation control command is output according to the current dose level; at the same time, the cumulative counter is incremented by 1. When the cumulative count reaches L=15 valid stimulation trials that meet the permissible stimulation criteria, the dose level is increased by one level and the cumulative counter is reset to zero. In the delayed stimulus state, the stimulus of the current trial is suspended, and reassessment is awaited in the next trial; the cumulative counter does not increment. In the de-stressing stimulation state, after moving one level in the reverse direction of the dose level table, a stimulation control command is output and the cumulative counter is cleared; if the current level is already the lowest level (60%RMT) in the dose level table, then the stimulation pause state is entered. In the paused stimulation state, the stimulation output is stopped, and step 5 (automatic pause and automatic resumption) is initiated. Under recalibration status, suspend the current stimulation protocol, remeasure the RMT and PMBS baseline templates according to the recalibration procedure, and reset the dose level to the lowest level after recalibration is completed.

[0055] Step 5: Automatic pause and automatic resume.

[0056] like Figure 4 As shown, in the paused stimulation state, the system does not output rTMS stimulation, but continues to induce MRBD and PMBS events in the original motor task paradigm, and continues to extract multimodal state features according to step 2.

[0057] Set a sliding window of length M=10. When all EEG stability and EMG synergy indices for 10 consecutive valid trials within the window meet the permissible stimulus criteria, and... It fell back to 0.30 or below, and Return to When the range is within the specified range, the safe recovery condition is considered met, the dose level is reset to the lowest level 60%RMT in the dose level table, and the control state is switched to the permissive stimulation state. Within P=10 trials after automatic recovery, the threshold in the permissible stimulus criterion is tightened as follows: , , , , , , A "test-confirm" window is formed. If any indicator in this window exceeds the limit, the system will directly revert to the paused stimulation state. Step 5: Automatic pause and automatic resumption. The selection of M=10, P=10, and L=15 in this embodiment is based on the following: M=10 ensures that the determination of the resumption of the pause state has sufficient statistical robustness without excessively prolonging the pause time; P=10 constitutes a trial and confirmation period of exactly half a group compared to the number of trials within the group (20), which is convenient for compatibility with existing group-based stimulation paradigms; L=15 ensures that the dose upregulation occurs after 15 consecutive stable and effective stimuli, corresponding to engineering experience of a stable state lasting more than two-thirds of the group. Those skilled in the art can adjust the specific values ​​of M, P, and L according to the specific application scenario.

[0058] Example 2: A specific implementation of a dosage level table.

[0059] In one implementation, the dosage level table is organized along the "stimulus intensity safety gating" dimension, including five levels: 60% RMT, 65% RMT, 70% RMT, 75% RMT, and 80% RMT. Figure 1 The "n%" includes these 5 levels, with a 5% RMT increment between adjacent levels. The output stimulation intensity under permissible stimulation conditions is the RMT value corresponding to the current level, and it must not exceed the stricter upper limit specified in the upper-level treatment plan, equipment manual, or ethical approval document; when de-stimulating, the level moves one level in the opposite direction; when increasing the dose, the level moves one level in the forward direction; the initial level and the level after pause and recovery are both 60% RMT.

[0060] In another implementation, the dose level table is organized along the dimension of "single stimulation duration," including four levels: 2s, 3s, 4s, and 5s, with a 1s increment between adjacent levels. Here, "single stimulation pulse train duration" refers to the total duration of a single triggered rTMS pulse train, rather than the pulse width of a single magnetic pulse.

[0061] In another implementation, the dosage level table is organized along the dimension of "the proportion of effective stimulation trials per unit time", including four levels: 5, 10, 15, and 20 maximum number of stimulations allowed per 20 trials.

[0062] It should be understood that the dosage level organization forms given in the above three embodiments can be used alone or in combination (i.e., organizing two-dimensional or three-dimensional level tables along two or three dimensions at the same time), and are only used as examples; those skilled in the art can adjust the specific dimensions, number of levels and values ​​of the dosage level table according to the actual application scenario without departing from the spirit of the present invention.

[0063] Example 3: Closed-loop rTMS safety control system based on multimodal neural states.

[0064] like Figure 5 As shown, this embodiment provides a closed-loop rTMS safety control system based on multimodal neural states, including a multimodal signal acquisition module, a multimodal feature extraction module, a closed-loop safety state discrimination module, a dose control module, and a stimulus execution interface module. The specific functions of each module are consistent with the corresponding steps in Embodiment 1.

[0065] Example 4: Computer-readable storage medium.

[0066] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the computer program is loaded and executed by a processor, it implements the steps of the closed-loop rTMS security control method based on multimodal neural states in Embodiment 1. The storage medium can be a common computer-readable storage medium in the art, such as a read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0067] Compared to existing closed-loop rTMS systems driven by EEG or PMBS signals, this invention does not redefine the physiological meaning of neural markers such as PMBS and IHI, nor does it change the original stimulation frequency, stimulation site, and target brain region settings. By introducing multimodal stability assessment and a graded safety state machine, the stimulation triggering criterion is expanded from a single indicator to a joint criterion of three types of multimodal indicators, improving the ability to identify atypical neural states, artifact contamination states, and ineffective movement states. The control response is expanded from "whether to trigger" to a five-level mechanism of "allow stimulation - delayed stimulation - downgraded stimulation - pause stimulation - recalibration", realizing a safety control scheme of "graded response + automatic pause + automatic recovery + recalibration". This enables the system to have autonomous graded handling capabilities when abnormal events occur, reducing the dependence on manual supervision by the operator. Through dose progression rules and a "trial and error" window, the stimulation dose always gradually increases from the "lowest effective dose verified as safe", which conforms to the general principle of dose safety in the field of non-invasive neuromodulation.

[0068] Therefore, this invention employs the aforementioned closed-loop rTMS safety control method and system based on multimodal neural states, simultaneously acquiring EEG, EMG, and stimulus response signals while the subject performs a motor task. Addressing the limitation of traditional methods that rely on a single biomarker for assessment, four EEG stability indicators are extracted to resolve the issues of noise contamination and the inability to identify abnormal signals under non-steady-state distortions. To address the lack of multi-signal cross-validation, EMG synergistic indicators are introduced to effectively identify and eliminate false-positive triggers caused by EEG artifacts, preventing erroneous stimulation during ineffective movement. To address the lack of post-stimulation neural response monitoring, stimulus response fluctuation indicators are calculated to achieve real-time closed-loop tracking of cortical excitability trends and quantify accumulated risk. Based on this, a five-level safety state machine is constructed to achieve graded response. After a pause, a sliding window recovery mechanism and a trial-confirmation window with tightened thresholds are introduced to ensure that the dose always gradually increases from the lowest, already verified safe level during automatic restart, balancing safety and efficacy. This invention forms a complete safety closed loop covering pre-stimulation, during-stimulation, and post-stimulation through multimodal and multi-dimensional autonomous collaborative judgment, significantly reducing reliance on manual supervision and ensuring the safety, effectiveness, and engineering adaptability of long-term interventions.

[0069] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A closed-loop rTMS security control method based on multimodal neural states, characterized in that, Includes the following steps: S1. The collected EEG signals, EMG signals, and stimulus response signals are preprocessed and subjected to time-frequency analysis to extract multimodal state features, including EEG state stability indicators, EMG synergy indicators, and stimulus response fluctuation indicators. S2. Input the extracted EEG stability index, EMG synergy index and stimulus response fluctuation index into the closed-loop safety status discrimination model, and generate the closed-loop rTMS safety status of the current trial according to multi-level criterion rules. The safety status includes the allowed stimulation status, delayed stimulation status, degraded stimulation status, paused stimulation status and recalibration status. S3. Output the corresponding rTMS control command according to the safety status: output the stimulation control command according to the current dose level in the allowed stimulation state; suspend the current stimulation in the delayed stimulation state; output the stimulation control command after lowering the dose level by one level according to the preset lowering rule in the de-stressed stimulation state; stop the stimulation output and start the automatic recovery monitoring process in the paused stimulation state. Suspend the current stimulus protocol and trigger the parameter re-measurement process under recalibration status; S4. When in the paused stimulation state, continuously monitor the multimodal state characteristics of subsequent trials. After M consecutive effective trials meet the safe recovery conditions, reset the dose level to the lowest level and switch to the allowed stimulation state.

2. The closed-loop rTMS security control method based on multimodal neural states according to claim 1, characterized in that, EEG state stability indicators include: the relative deviation of the current trial PMBS peak value from the mean of the PMBS peak values ​​of the most recent N valid trials. The relative deviation of the current PMBS duration from the average PMBS duration of the most recent N valid trials. Spatial concentration of beta-band non-negative energy on sensorimotor cortical electrode groups And the similarity between the current trial frequency spectrum and the individualized historical template. The specific calculation formula is as follows: ； ； ； ； ； ； in, For the first The relative deviation index of the PMBS peak value in each trial. For the first The peak value of PMBS in each trial The average of the peak values ​​of PMBS in the most recent N valid trials. For the first The average duration of PMBS in the most recent N valid trials before each trial. For the first The duration of PMBS in each trial To determine the beta power of the i-th electrode in the sensory motor cortex electrode group. The spectrum during the current PMBS event interval. The template is obtained by averaging the spectrum of the most recent 10 valid PMBS event intervals collected in the previous recalibration process. To avoid constants with a denominator of zero.

3. The closed-loop rTMS security control method based on multimodal neural states according to claim 2, characterized in that, Electromyographic synergy indicators include: the difference between the electromyographic onset time of the target muscle group and the onset time of the electroencephalogram (MEG) MRBD. The non-negative deviation of the difference between the peak time of PMBS and the termination time of electromyography relative to the individual baseline delay during the initial calibration phase. and the baseline drift factor of electromyography in non-task channels The specific calculation formula is as follows: ； ； ； ； in, The moment when the amplitude of the electromyographic envelope first continuously exceeds three standard deviations of the baseline noise level for at least 30 ms is defined. The time when the electromyographic envelope amplitude falls back to below twice the baseline noise level and remains there for at least 50 ms is defined as the moment when the electromyographic envelope amplitude falls back to below twice the baseline noise level and remains there for at least 50 ms. The start time of the period in which the beta power continuously drops to 10% below the baseline and lasts for at least 100 ms. The individual baseline recovery delay obtained during the initial calibration phase, The discrete sampling points of the non-task side hand muscle channel within a time window of 500ms and L sampling points; The RMS baseline values ​​are the subjects' continuous 3-minute resting state RMS values ​​recorded during the initial calibration phase. This is the real-time electromyography baseline for the non-task channel. This is the peak time of PMBS. This represents the time difference between the peak time of PMBS and the termination time of electromyography of the target muscle group in the current trial.

4. The closed-loop rTMS security control method based on multimodal neural states according to claim 3, characterized in that, Stimulus response variability metrics include: the coefficient of variation of MEP amplitude induced by the most recent K safety-compliant single-pulse TMS probes. The time required for beta power to recover to near baseline after stimulation The slope of recovery time after stimulation in the most recent R trials and the energy ratio of the beta band before and after stimulation The specific calculation expression is as follows: After every 10 effective stimulation trials, a non-therapeutic single-pulse TMS probe with an intensity of 80%–120% RMT is transmitted before the MRBD phase of the next trial, and the MEP amplitude induced by the probe pulse is recorded; the MEP amplitude sequence of the most recent K probes is then taken. calculate: ， ； in, Let K be the standard deviation of the MEP amplitude from the most recent K probes. The average of the MEP amplitudes from the most recent K probes. To avoid constants with a denominator of zero; The time required for the beta power to return to near the baseline mean 500 ms before stimulation after each effective stimulus. Take the most recent R trials The recovery time slope was obtained by least squares linear regression of the sequence. : ； Energy ratio of beta band before and after stimulation : ； in, To stimulate beta energy within the first 500ms, This refers to the beta energy within a 500-1000ms window after the stimulus ends. For the most recent The local index in the effective stimulus trial sequence, with a value of [value]. ; For the most recent The effective stimulus test was followed by the local sequence number. The mean, ; For the first The recovery time required for the post-stimulation beta power to recover to near the baseline mean of 500 ms before stimulation in a valid stimulation trial; For the most recent The mean of the recovery time series corresponding to each effective stimulus trial; The number of most recent effective stimulus trials used to calculate the recovery time slope. Number the current effective stimulus trial.

5. The closed-loop rTMS security control method based on multimodal neural states according to claim 4, characterized in that, The specific rules for multiple criteria in S2 are as follows: when , , , , , , , , , When both conditions are met, it is determined to be a permissible stimulus state; When the mild out-of-bounds criterion of the delayed stimulus state is met, it is determined to be a delayed stimulus state; When the cumulative out-of-bounds criterion for a reduced-order stimulus state is met, it is determined to be a reduced-order stimulus state. When the severe out-of-bounds criterion for the paused stimulation state is met or an artificial safety abort signal is received, the state is determined to be paused stimulation. When the cumulative number of effective stimulation trials reaches the preset upper limit, the deviation of the resting motion threshold exceeds 10%, or the electrode contact quality is lower than the qualified threshold, it is determined to be in a recalibration state.

6. The closed-loop rTMS security control method based on multimodal neural states according to claim 5, characterized in that, The downgrade rule includes: shifting the stimulation dose parameter one level in the reverse direction of the preset dose level table; the dose level table includes 5 levels: 60%RMT, 65%RMT, 70%RMT, 75%RMT, and 80%RMT, with a step of 5%RMT between adjacent levels; when the current level is 60%RMT, switch the control state to pause stimulation state.

7. The closed-loop rTMS security control method based on multimodal neural states according to claim 6, characterized in that, The automatic recovery monitoring process includes: S31A: Do not output rTMS stimulation in the paused stimulation state, but continue to induce MRBD and PMBS events with a motor task that can induce MRBD / PMBS events, and extract multimodal state features according to S1. S32A. Set a sliding window of length M=10. In 10 consecutive valid trials within the sliding window, all EEG state stability indices and EMG synergy indices in each trial meet the permissible stimulus state criterion. It fell back to 0.30 or below, and Return to When the interval is within the safe recovery condition, the dose level is reset to 60%RMT, and the control state is switched to the permissive stimulation state. S33A. Within P=10 trials after automatic recovery, the threshold in the stimulus state criterion is allowed to tighten as follows: , , , , , , If any indicator in the window exceeds the limit, the system will immediately revert to the paused stimulation state.

8. The closed-loop rTMS security control method based on multimodal neural states according to claim 7, characterized in that, The re-measurement process includes: S31B, pause rTMS stimulation output; S32B. Re-measure RMT according to the existing RMT measurement method: the minimum stimulation intensity corresponding to at least 5 out of 10 TMS pulses induced by MEP not less than 50μV. S33B. Collect the time-spectral data of MRBD / PMBS event intervals from 15 valid trials. After removing any abnormal trials where the EEG stability index exceeded the limit, take the average time-spectral data of the remaining trials as the new PMBS baseline template. ; S34B, Recalibrate electromyography baseline Update the impedance record; S35B, after calibration, the dose level is reset to 60%RMT, the control state is switched to the allowable stimulation state and enters the trial-confirmation window.

9. The closed-loop rTMS security control method based on multimodal neural states according to claim 8, characterized in that, It also includes a dose progression rule: under permissible stimulation conditions, when all 15 consecutive effective stimulation trials of L=15 times meet the permissible stimulation condition criteria, the dose level is increased by one level in the positive direction; when the dose level reaches 80% RMT and still meets the permissible stimulation condition criteria, the dose level remains at the highest level and is not increased further.

10. A closed-loop rTMS safety control system based on multimodal neural states, applied to the closed-loop rTMS safety control method based on multimodal neural states as described in any one of claims 1-9, characterized in that, include: The multimodal signal acquisition module is used to simultaneously acquire electroencephalogram (EEG) signals, electromyogram (EMG) signals, and stimulus-response signals during the subject's performance of a motor task; The multimodal feature extraction module is used to preprocess the acquired multimodal signals and extract multimodal indicators such as EEG state stability indicators, electromyographic coordination indicators, and stimulus response fluctuation indicators. The closed-loop safety status determination module is used to output the closed-loop rTMS safety status according to multimodal indicators and preset multi-level judgment rules. The dose control module is used to determine the control process of allowing stimulation, delaying stimulation, down-order stimulation, pausing stimulation, or recalibrating based on the safety status, and to complete the sliding window safety recovery monitoring and dose increment after automatic pause; The stimulation execution interface module is used to convert the control commands generated by the dose control module into physical trigger signals that drive the rTMS stimulator.

Images