A closed-loop sleep modulation method combining transcranial pulsed electrical stimulation and acoustic stimulation

Abstract

The application discloses a closed-loop sleep regulation method combined with transcranial pulse electrical stimulation and sound stimulation, and belongs to the technical field of sleep aid. The method is realized based on a closed-loop sleep slow wave regulation system, and can realize closed-loop sleep slow wave regulation from sleep physiological signal collection, sleep stage judgment, target slow wave identification, phase locking combined stimulation, post-stimulation artifact removal to instant response feedback regulation. The method realizes time-coupled combined stimulation in the window of the same target slow wave event, reduces the deep network response threshold, and thus obtains slow wave synergistic amplification effect with high intensity, high stability and individual universality without increasing the single-mode stimulation intensity and without causing micro-arousal.

Description

Technical Field

[0001] This invention belongs to the field of sleep aid technology, specifically relating to a closed-loop sleep regulation method that combines transcranial pulse electrical stimulation and sound stimulation. Background Technology

[0002] Slow-wave sleep is a typical EEG activity during non-rapid eye movement (NREM) sleep and is closely related to important physiological processes such as bodily recovery, maintenance of nervous system homeostasis, and memory consolidation. Therefore, precise enhancement of slow-wave activity is of significant theoretical and practical value for improving clinical problems such as insufficient deep sleep, inadequate sleep recovery, and insomnia. However, existing sleep enhancement interventions mostly rely on single modalities, such as simple auditory stimulation or simple electrical stimulation, which generally suffer from limitations in intensity, large individual differences, unstable stimulation effects, and difficulty in achieving precise phase matching with slow waves. Therefore, there is an urgent need for a multimodal combined stimulation scheme that combines high precision, high stability, and strong controllability to overcome the effectiveness limitations of single-modal stimulation, achieve precise closed-loop intervention targeting individual slow-wave events, and thus effectively enhance slow-wave activity and improve sleep quality.

[0003] Sound stimulation primarily acts on the auditory cortex and related sensory processing areas, exhibiting a diffuse spatial effect and relatively insufficient target selectivity. While single sound stimuli can induce reactive slow waves or enhance local slow wave activity under certain conditions, their efficiency in regulating the core slow wave network is limited, making it difficult to achieve stable, continuous, and high-intensity precise slow wave regulation. Sound stimulation inherently carries the risk of inducing arousal responses while regulating sleep slow waves, a characteristic that severely restricts its reliability in sleep regulation. A close anatomical and functional connection exists between the auditory system and the thalamic-cortical arousal network. External sound signals, transmitted through the auditory pathway to the medial geniculate body, can further project to key structures for arousal regulation, such as the thalamic reticular nucleus and the thalamic-laminar nucleus, thereby activating the cortical arousal system. Even short-duration, low-intensity sound stimulation may still trigger micro-arousals or transient increases in cortical excitability, leading to impaired sleep continuity.

[0004] Single transcranial pulsed electrical stimulation (TPCS) can apply an external electric field to brain tissue through scalp electrodes and has a certain modulating effect on slow-wave neural activity during sleep. However, it is essentially a transcranial surface stimulation method, where the stimulating current must pass through the scalp, skull, cerebrospinal fluid, and other tissues before entering the brain. Due to the significant differences in conductivity among different tissues and the complex anatomy of the skull, the applied current is prone to shunting, attenuation, and diffusion during conduction. Therefore, the actual range of action usually exhibits a relatively wide spatial distribution rather than being strictly limited to a pre-set target area. Thus, single transcranial electrical stimulation is more likely to produce non-specific modulation of a large area of ​​cortical networks or global slow-wave activity, making it difficult to implement highly spatially selective and precise interventions on specific brain regions or functional networks. Furthermore, single TPCS mainly forms a relatively strong electric field in the superficial cortical region, with limited direct effect on deep structures such as the thalamus, making it difficult to effectively drive or modulate deep slow-wave core circuits.

[0005] Transcranial pulsed electrical stimulation (TPCS) introduces large-amplitude electrical stimulation artifacts into the electroencephalogram (EEG) signal during the output process. These artifacts typically obscure the target slow wave itself and the short-term EEG activity following stimulation, making it difficult for researchers to directly observe the immediate response changes of the target slow wave after combined stimulation. For example, key indicators such as slow wave amplitude, waveform slope, rhythmic continuity, and energy changes in related frequency bands are easily obscured by artifacts and cannot be accurately extracted. Consequently, it is not only difficult to determine whether a single TPCS combined with sound stimulation truly affects the current target slow wave event, but it also limits the dynamic adjustment and optimization of stimulation intensity, duration, and trigger phase based on immediate response results. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a closed-loop sleep regulation method that combines transcranial pulse electrical stimulation and acoustic stimulation.

[0007] The technical problem addressed by this invention is solved as follows:

[0008] A closed-loop sleep regulation method combining transcranial pulsed electrical stimulation and acoustic stimulation is implemented based on a closed-loop sleep slow wave regulation system, including a physiological signal acquisition module, a sleep stage judgment module, a target slow wave identification and phase triggering module, a TPCS and acoustic stimulation control module, an artifact removal module, and an instant response evaluation and parameter adjustment module.

[0009] The physiological signal acquisition module collects multi-channel physiological signals during the subject's sleep process in real time, including electroencephalogram (EEG), electrooculogram (EOG), and electromyogram (EMG) signals, and transmits them to the sleep stage determination module in real time.

[0010] After receiving the multichannel physiological signal, the sleep stage determination module inputs it into the pre-trained sleep staging model to make a real-time judgment on the subject's current sleep stage. When the sleep staging model determines that the subject is in stage N3 and the N3 stage determination result remains stable within a preset continuous time window, the target slow wave recognition and phase triggering process is initiated.

[0011] The target slow wave recognition and phase triggering module superimposes and averages the signals from leads FP1, FP2, F3, and F4 in the multi-channel physiological signal to obtain the frontal slow wave composite signal; it then identifies the target slow wave event based on the frontal slow wave composite signal; after identifying the target slow wave event, it performs phase estimation on the subsequent frontal slow wave composite signal; when the frontal slow wave composite signal enters the preset target phase interval, it generates a stimulation trigger command and sends it to the TPCS and sound-based stimulation control module.

[0012] After receiving the stimulation trigger command, the TPCS and sound combined stimulation control module controls the simultaneous output of transcranial pulse electrical stimulation and sound stimulation. The two stimuli work together on the same preset target phase interval to achieve combined stimulation.

[0013] After the combined stimulation is completed, the artifact removal module performs TPCS artifact correction on the EEG signal after the combined stimulation.

[0014] The instant response assessment and parameter adjustment module receives the EEG signal after TPCS artifact correction, assesses the instant response caused by the current combined stimulation, and adjusts the stimulation parameters of the combined stimulation trigger phase, acoustic stimulation, and transcranial pulsed electrical stimulation based on the assessment results.

[0015] Furthermore, the physiological signal acquisition module employs a 32-channel EEG acquisition system; specifically, 28 channels of EEG signals are acquired via a head-mounted EEG cap, with reference electrodes positioned on the left and right earlobes; 2 channels of electrooculography (EOG) electrodes are acquired, with the right EOG electrode positioned above the right outer canthus and the left EOG electrode positioned above the left outer canthus; and 2 channels of electromyography (EMG) electrodes are acquired near the mandible.

[0016] Furthermore, in the sleep stage determination module, the sleep stage model adopts a CNN-BiLSTM model that combines a convolutional neural network and a bidirectional long short-term memory network. The output sleep stage classification results include wakefulness, N1 stage, N2 stage, N3 stage, and REM sleep stage.

[0017] Furthermore, the process of target slow-wave event recognition in the target slow-wave recognition and phase triggering module is as follows:

[0018] Based on the sign change of the frontal slow wave composite signal, the position where the frontal slow wave composite signal changes from positive to negative is defined as the falling zero-crossing point, and the position where it changes from negative to positive is defined as the rising zero-crossing point. The frontal slow wave composite signal between the falling zero-crossing point and the adjacent rising zero-crossing point is extracted as a candidate slow wave segment. For each candidate slow wave segment, its negative peak amplitude and the positive peak amplitude before the negative peak are detected, and the negative half-wave duration and peak-to-peak value are calculated. Candidate slow wave segments that simultaneously meet the conditions of 0.25s ≤ negative half-wave duration ≤ 1.25s, negative peak amplitude below a set threshold, and peak-to-peak value height above a set threshold are selected as target slow wave events.

[0019] Furthermore, the process of phase estimation of the subsequent frontal slow-wave synthesis signal and generation of stimulus triggering command in the target slow-wave recognition and phase triggering module is as follows:

[0020] After downsampling and smoothing the frontal slow-wave composite signal following the target slow-wave event, a Hilbert transform is performed to obtain an analytical signal. The instantaneous phase angle of the analytical signal is extracted and converted into an angle value. Based on the average angle value of the most recent sampling points, it is determined whether the current composite signal has entered the preset target phase interval of 0°–90°. If it has entered the preset target phase interval, a stimulus trigger command is generated and sent to the TPCS and sound-based combined stimulus control module.

[0021] Furthermore, in the TPCS and sound-based combined stimulation control module, the sound stimulation uses short pink noise pulses, with a single sound stimulation duration set to 50ms; the duration of word transcranial pulse electrical stimulation is also set to 50ms; the sound stimulation intensity is determined based on the individual hearing threshold of the subject or the sound intensity tolerated during sleep; the stimulation intensity of transcranial pulse electrical stimulation is based on 1mA; during the combined stimulation process, the intensity of both sound stimulation and transcranial pulse electrical stimulation can be dynamically adjusted within a small range of 10% above and below their respective baseline values ​​based on the immediate response assessment results.

[0022] Furthermore, in the artifact removal module, the specific process of TPCS artifact correction is as follows:

[0023] During the resting state or non-sleep combined stimulation phase, the same stimulation parameters, electrode positions, multi-channel physiological signal acquisition equipment, and lead configuration as the combined stimulation were used to continuously apply 50 transcranial pulse electrical stimulations, while simultaneously acquiring EEG signals. For each EEG signal segment corresponding to a transcranial pulse electrical stimulation, a fixed time window from 50 ms before stimulation to 500 ms after stimulation was extracted, with the stimulation trigger mark as the time zero point. For each EEG signal segment, the mean of the EEG signal after stimulation was calculated by subtracting it from the mean of the EEG signal 50 ms before stimulation, i.e., baseline correction of the EEG signal after stimulation was performed using the short time window before stimulation. Outliers were removed from the baseline-corrected EEG signals, and the remaining effective baseline-corrected EEG signals were superimposed and averaged to construct an individualized artifact template for the subject under the current transcranial pulse electrical stimulation conditions.

[0024] For the EEG signal after combined stimulation, segments of the same length as the individualized artifact template were extracted, and a segmented scaling strategy was used for TPCS artifact correction. The EEG signal after combined stimulation was divided into three time periods: 0-10ms as the stimulation onset period, 10-50ms as the stimulation duration period, and 50-500ms as the post-stimulation decay period.

[0025] Calculate the template scaling factor for each time period:

[0026]

[0027] in, Let t be the scaling factor for the j-th time interval, and t be the time. For the j-th time period, The signal is the baseline-corrected EEG signal after combined stimulation. For individualized artifact templates;

[0028] For each time period, TPCS artifact correction is performed to obtain the correction results. :

[0029]

[0030] The correction results are evaluated using residuals, with evaluation indicators including residual peak-to-peak value, residual root mean square value, and correlation between residuals and individualized artifact templates. If all three evaluation indicators meet the set threshold requirements, the result is considered a valid correction result.

[0031] Furthermore, in the immediate response assessment and parameter adjustment module, the amplitude changes, rise slope changes, and delta band energy changes of the EEG signal after TPCS artifact correction are extracted and compared with the corresponding features of the EEG signal before combined stimulation to determine the enhancement effect of the current combined stimulation on the target slow wave event. If the enhancement effect meets the preset requirements, the current stimulation parameters are maintained; otherwise, the stimulation parameters of the combined stimulation trigger phase, acoustic stimulation, and transcranial pulse electrical stimulation are adjusted.

[0032] The beneficial effects of this invention are:

[0033] The method described in this invention combines bottom-up auditory input with top-down TPCS cortical excitability bias. By implementing temporally coupled joint stimulation within a window of the same target slow-wave event, the response threshold of the deep network is reduced, thereby achieving a slow-wave synergistic amplification effect with high intensity, high stability, and individual universality without increasing the intensity of single-modal stimulation or inducing microarousals.

[0034] The method described in this invention addresses the problems of large amplitude artifacts in TPCS stimulation, which easily cover target slow waves and short-term EEG signals after stimulation. By collecting artifact samples with the same stimulation conditions as the formal sleep experiment before the formal sleep experiment, an individualized artifact template is constructed. Combined with strategies such as stimulation event time alignment, template matching, amplitude correction, and subtraction processing, artifacts are removed from the EEG signals after stimulation in the formal experiment, restoring the short-term effective EEG window after stimulation, and providing a reliable data foundation for slow wave immediate response analysis.

[0035] The method described in this invention possesses the ability to dynamically adjust stimulation parameters based on immediate response assessment. By analyzing the artifact-corrected short-term EEG signals after stimulation, immediate response features such as slow wave amplitude, waveform slope, rhythm continuity, slow wave power, and energy changes in related frequency bands are extracted. The enhancement effect of a single TPCS combined with sound stimulation on the target slow wave event is determined, and parameters such as stimulation intensity, stimulation duration, trigger phase, sound intensity, and sound stimulation sequence are dynamically adjusted based on the immediate response results. This forms a closed-loop control process of "stimulus output - immediate response assessment - parameter optimization adjustment," improving the stability, individual adaptability, and control efficiency of sleep slow wave enhancement. Attached Figure Description

[0036] Figure 1 This is a block diagram of the closed-loop sleep slow-wave modulation system described in this invention;

[0037] Figure 2 This is a schematic diagram of the UP-phase before the rising branch peak of the target slow wave in the method described in this invention;

[0038] Figure 3 This is a schematic diagram illustrating how the artifact removal module in the method of the present invention corrects the electroencephalogram (EEG) signal after stimulation.

[0039] Figure 4 This is a schematic diagram illustrating the synchronous joint control of slow-wave events around the same target in the method described in this invention. Detailed Implementation

[0040] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0041] This embodiment provides a closed-loop sleep regulation method combining transcranial pulsed electrical stimulation and acoustic stimulation, implemented based on a closed-loop slow-wave sleep regulation system, such as... Figure 1 As shown, it includes a physiological signal acquisition module, a sleep stage judgment module, a target slow wave recognition and phase triggering module, a TPCS and sound combined stimulation control module, an artifact removal module, and an instant response evaluation and parameter adjustment module. The physiological signal acquisition module is used to continuously acquire physiological signals such as EEG, EEG, and EMG during the subject's sleep process; the sleep stage judgment module is used to judge the subject's current sleep stage in real time based on the physiological signals, and to open the subsequent target slow wave recognition and stimulation triggering process when the subject is determined to be in N3 stage; the target slow wave recognition and phase triggering module is used to identify a single target slow wave event in the EEG signal in real time, and to judge its current phase state based on the slow wave waveform characteristics, and to generate a stimulation triggering command when the target slow wave enters the preset target phase interval; the TPCS and sound joint stimulation control module is used to control the synchronous output of transcranial pulse electrical stimulation and sound stimulation around the same target slow wave event or to output them according to a preset relative timing sequence according to the stimulation triggering command; the artifact removal module is used to remove TPCS stimulation artifacts in the EEG signal after stimulation to restore the short-term effective EEG signal after stimulation; the instantaneous response evaluation and parameter adjustment module is used to extract instantaneous response features such as slow wave amplitude, rise slope, rhythm continuity, and related frequency band energy based on the EEG signal after artifact removal, and to adjust the subsequent stimulation parameters according to the instantaneous response features. Therefore, the system can realize closed-loop sleep slow wave regulation from sleep physiological signal acquisition, sleep stage judgment, target slow wave identification, phase-locked joint stimulation, post-stimulation artifact removal to immediate response feedback regulation.

[0042] The method described in this embodiment specifically includes the following steps:

[0043] The physiological signal acquisition module continuously collects multichannel physiological signals from the subject during sleep. Specifically, this embodiment uses a 32-channel EEG acquisition system, with electrodes positioned according to the international EEG 10-20 electrode standard. Specifically, a head-mounted EEG cap collects 28 channels of EEG signals, with reference electrodes positioned on the left and right earlobes; two-channel electrooculography (EOG) electrodes collect EOG signals, with the right EOG electrode located above the right outer canthus and the left EOG electrode located above the left outer canthus; and two-channel electromyography (EMG) electrodes collect EMG signals near the mandible. The EEG signals reflect slow-wave activity and related brain rhythm changes during sleep, the EOG signals assist in determining eye movement activity and sleep stage changes, and the EMG signals assist in determining muscle tone levels and wakefulness. The physiological signal acquisition module transmits the collected EEG, EOG, and EMG signals to the sleep stage determination module in real time.

[0044] After receiving the multichannel physiological signal, the sleep stage determination module inputs the multichannel physiological signal into a pre-trained sleep staging model to make a real-time judgment on the subject's current sleep stage. The sleep staging model can be a machine learning model or a deep learning model trained based on multimodal signals of EEG, EEG, and EMG, used to output the sleep stage classification result corresponding to the current time window. In this embodiment, the sleep staging model can adopt a CNN-BiLSTM model that combines a convolutional neural network and a bidirectional long short-term memory network. The convolutional neural network is used to extract local time-frequency features in the multichannel physiological signal, and the bidirectional long short-term memory network is used to model the sleep stage transition relationship between adjacent time windows. The model output includes at least one classification result among wakefulness, N1 stage, N2 stage, N3 stage, and REM sleep stage. When the sleep stage determination module determines that the subject is not in N3 stage according to the output result of the sleep staging model, the system does not open the subsequent target slow wave recognition and joint stimulus triggering process; when the sleep stage determination module determines that the subject is in N3 stage, and the N3 stage determination result remains stable within a preset continuous time window, the system opens the target slow wave recognition and phase triggering process.

[0045] Once the system opens the target slow wave identification and phase triggering process, the target slow wave identification and phase triggering module superimposes and averages the signals from leads FP1, FP2, F3, and F4 to obtain a frontal slow wave composite signal, and then filters target slow wave events based on the frontal slow wave composite signal. Specifically, firstly, the sign change of the frontal slow wave composite signal is calculated. The position where the signal changes from positive to negative is defined as the falling zero-crossing point, and the position where the signal changes from negative to positive is defined as the rising zero-crossing point. The frontal slow wave composite signal between the falling zero-crossing point and the adjacent rising zero-crossing point is extracted as a candidate slow wave segment. For each candidate slow wave segment, its negative peak amplitude and the positive peak amplitude before the negative peak are detected, and the duration of the negative half-wave and the peak-to-peak value are calculated. The duration of the negative half-wave can be determined by the time interval between the falling zero-crossing point and the adjacent rising zero-crossing point. The slow wave duration condition is set to be less than 0.25s or greater than 1.25s. Candidate slow wave segments that meet the slow wave duration condition are selected. For the selected candidate slow wave events, if they simultaneously meet the preset negative peak amplitude condition and the peak-to-peak value condition, they are determined to be target slow wave events that meet the preset slow wave conditions.

[0046] The target slow-wave recognition and phase triggering module performs phase estimation on the identified target slow-wave events. Specifically, after downsampling and smoothing the frontal slow-wave composite signal following the target slow-wave event, a Hilbert transform is performed to obtain an analytical signal; the instantaneous phase angle of the analytical signal is extracted and converted into an angle value. Subsequently, the target slow-wave recognition and phase triggering module determines whether the current composite signal enters the preset target phase interval of 0°–90° based on the average angle values ​​of the most recent sampling points. If it enters the preset target phase interval, it is determined that it has entered the 0°–90° enhanced phase window during the transition from the Down state to the Up state, a stimulus triggering command is generated, and the stimulus triggering command is sent to the TPCS and sound-based joint stimulation control module. Figure 2 As shown, the system can identify the UP-phase before the rising branch peak of the target slow wave, i.e., the 0°-90° phase interval, thereby effectively locating the preset enhanced phase window; the black dot in the figure indicates the identification position within the phase window, and the arrow indicates the stimulation trigger after the target phase is identified.

[0047] After receiving the stimulation trigger command, the TPCS and sound-based combined stimulation control module controls the simultaneous output of transcranial pulsed electrical stimulation and sound stimulation, so that the two stimuli work together on the same preset target phase interval. The sound stimulation uses short-duration pink noise pulses, with a single sound stimulation duration of 50ms; the TPCS stimulation uses short-duration transcranial pulsed electrical stimulation, with a single TPCS stimulation event duration of 50ms. In the stimulation electrodes, channel Fz serves as the anode, and channel Az serves as the cathode. Channel Fz is located in the midfrontal region, corresponding to the frontal slow-wave characterization region used for target slow-wave recognition. Channel Az serves as the loop electrode, forming the transcranial pulsed electrical stimulation pathway together with channel Fz. Specifically, the sound stimulation intensity can be determined based on the individual subject's hearing threshold, tolerable sound intensity during sleep, or pre-experiment results; the TPCS stimulation intensity is based on 1mA. During the formal combined stimulation process, the intensity of both the sound stimulus and the TPCS stimulus can be dynamically adjusted within a small range of 10% above and below their respective baseline values ​​based on the immediate response assessment results. That is, the sound stimulus intensity can be adjusted within the range of 90% to 110% of the baseline sound intensity, and the TPCS stimulus intensity can be adjusted within the range of 0.9mA to 1.1mA.

[0048] After the stimulus output is completed, the artifact removal module performs TPCS artifact correction on the post-stimulation EEG signal, such as... Figure 3As shown. Since TPCS stimulation generates large-amplitude electrical stimulation artifacts in EEG signals, these artifacts may cover short-term EEG activity during and after stimulation, affecting subsequent real-time analysis of frontal slow-wave complex signals. Therefore, this embodiment employs a TPCS artifact removal method based on segmented scaling artifact templates to correct the short-term EEG window after stimulation, recovering effective EEG signals usable for real-time response assessment. Specifically, before formal sleep-related stimulation, TPCS artifact samples are collected in the subject's resting state or during informal stimulation phases. During this phase, stimulation parameters, electrode positions, acquisition equipment, and lead configurations consistent with the formal experiment are used, with 50 consecutive TPCS stimulations applied and EEG signals recorded simultaneously. For each stimulation-related EEG signal segment, a fixed time window from 50 ms before stimulation to 500 ms after stimulation is extracted, with the stimulation trigger marker as the time zero point. For each EEG signal segment, the mean of the post-stimulation EEG signal is calculated by subtracting the mean of the pre-stimulation 50 ms EEG signal, i.e., baseline correction is performed on the post-stimulation EEG signal using the pre-stimulation short-term window to remove DC offset and slow drift. Segments exhibiting significant body movement, electromyographic contamination, signal saturation, abnormal peak values, or unstable trigger markers were discarded. The remaining baseline-corrected valid EEG signal segments were then averaged channel-by-channel to construct an individualized artifact template for the subject under the current TPCS stimulation condition.

[0049] During formal sleep-related combined stimulation (TPCS), the system extracts a segment of EEG signal of the same length as the template after each TPCS stimulation to remove artifacts using a segmented scaling strategy. Specifically, the system uses the stimulation trigger moment as the zero point and extracts the EEG signal within 500ms after stimulation as the main artifact correction window. This window is divided into three time periods: 0-10ms is the stimulation onset period, mainly corresponding to the spike artifacts at the moment the stimulation begins; 10-50ms is the stimulation duration period, mainly corresponding to the main artifacts during the TPCS stimulation output; and 50-500ms is the post-stimulation decay period, mainly corresponding to the tail decay artifacts after the stimulation ends. The system calculates the template scaling factor for each of the three time periods and performs segmented artifact template subtraction to improve the correction matching degree for different time periods.

[0050] For each time segment, the system calculates the template scaling factor corresponding to that segment. Let the EEG segment to be corrected after baseline correction be... Individualized artifact templates, after baseline correction, are The time interval of the j-th artifact is The scaling factor for that time period. It can be calculated using least squares fitting:

[0051]

[0052] This coefficient is used to represent the ratio between the actual artifact amplitude and the pre-constructed template amplitude in the current stimulus fragment. When When >1, it indicates that the artifact amplitude of the current segment within this time period is higher than the template; when When <1, it indicates that the current artifact amplitude is lower than the template; when When the value is approximately 1, it indicates that the current artifact's amplitude is basically the same as the template amplitude. Subsequently, the system performs artifact template subtraction within the corresponding time period:

[0053]

[0054] After artifact template subtraction is completed, the system processes the correction results. Residual evaluation is performed. Residual evaluation primarily uses three indicators: First, the residual peak-to-peak value, which is the difference between the maximum and minimum amplitude values ​​of the corrected result, used to determine if significant residual spikes still exist; second, the residual root mean square value, used to determine if the residual energy of the corrected result is still too high; and third, the correlation between the residual and the original artifact template, used to determine if the corrected result still retains a template morphology similar to the TPCS artifact. If all the above indicators meet the preset requirements, the segment is marked as a valid corrected segment and proceeds to subsequent real-time response analysis; if any residual evaluation indicator does not meet the preset requirements, the stimulus segment is considered to still have significant artifact residue or insufficient correction quality. The system directly marks the segment as an abnormal segment and removes it from subsequent slow-wave real-time response analysis, enhancement effect evaluation, and stimulus parameter update processes, without further fine-tuning or correction. The present invention does not simply use a fixed artifact template for direct subtraction, but rather performs segmented scaling, smooth splicing, and residual verification on the artifact template based on the actual artifact amplitude and time pattern of each TPCS stimulation segment, thereby more accurately removing TPCS stimulation artifacts and restoring the short-term effective EEG window after stimulation.

[0055] After receiving the artifact-removed EEG signal, the immediate response assessment and parameter adjustment module evaluates the immediate EEG response caused by the current combined stimulation. Specifically, the module extracts changes in slow wave amplitude, slow wave rise slope, and delta band energy, and compares these features with the corresponding features of natural slow wave events under baseline conditions to determine whether the current combined TPCS and sound stimulation enhances the target slow wave event. When the immediate response assessment result indicates that the slow wave enhancement effect meets the preset requirements, the system can maintain the current stimulation parameters; when the immediate response assessment result indicates that the slow wave enhancement effect is insufficient, the system can adjust subsequent stimulation parameters within a preset safety range, such as adjusting the sound stimulation intensity or stimulation trigger phase; when the immediate response assessment result suggests that the stimulation may cause micro-arousals or sleep disturbances, the system can reduce the stimulation intensity, pause stimulation, or wait for subsequent target slow wave events that meet the conditions.

[0056] Through the above-described embodiments, this invention can automatically identify a single target slow-wave event when the subject is in N3 stage sleep, and trigger TPCS and sound-based combined stimulation when the target slow wave enters a preset enhanced phase interval. Simultaneously, after stimulation, effective EEG signals are recovered through artifact removal, and subsequent stimulation parameters are adjusted based on the immediate EEG response results. Thus, this invention forms a closed-loop sleep slow-wave regulation process of "physiological signal acquisition—sleep stage determination—target slow-wave identification and phase triggering—TPCS and sound-based combined stimulation—artifact removal—immediate response evaluation and parameter adjustment," thereby improving the temporal accuracy, regulation stability, and individual adaptability of slow-wave enhancement.

[0057] This invention proposes a technical solution for the simultaneous joint regulation of TPCS and auditory stimulation around the same target slow-wave event. Existing sleep slow-wave regulation methods mostly employ single stimulation methods, such as simple auditory stimulation or simple transcranial electrical stimulation. Simple auditory stimulation mainly transmits external sensory input to the thalamus-cortex network through the auditory pathway. This pathway is relatively long and easily affected by factors such as individual auditory sensitivity, sleep depth, thalamic sensory gating state, and arousal threshold, making it difficult to stably convert the stimulus input into effective slow-wave enhancement. While simple TPCS can directly regulate the membrane potential and excitability level of neurons in the target cortical region through short-term pulsed electric fields, its effect is affected by skull impedance, cerebrospinal fluid shunting, electric field diffusion, and individual anatomical differences, making it difficult to consistently achieve high-intensity slow-wave enhancement when used alone. This invention does not simply apply the two stimulation methods in parallel, but rather uses a single target slow-wave event as the regulation object, such as... Figure 4 As shown, when the TPCS (Transcranial Thromboplasty System) enters the preset enhanced phase window, it outputs TPCS and auditory stimulation synchronously or near synchronously. This couples the "bottom-up" sensory input provided by the auditory stimulation with the "top-down" cortical excitability modulation provided by the TPCS in time, creating convergence on the functional network. Thus, the TPCS enhances the target cortical network's response to auditory input, while the auditory stimulation provides induction and rhythmic shaping input for the current slow wave rise process. Both work together to promote more neurons to participate in synchronized firing. Compared to single stimulation methods, this invention can improve slow wave amplitude, rise slope, and rhythmic stability under a lower single-modal stimulation load, reducing regulatory fluctuations caused by ineffective stimulation and individual differences, thereby achieving a stronger, more stable, and more consistent sleep slow wave enhancement effect.

[0058] This invention provides a TPCS artifact removal method based on segmented scaling artifact templates to address the problem of large amplitude and short duration of electrical stimulation artifacts during transcranial pulsed electrical stimulation (TCS), which easily cover key EEG windows after stimulation. In existing TCS and EEG simultaneous recording schemes, stimulation artifacts often obscure the target slow wave and its immediate post-stimulation response, making it difficult to accurately extract key indicators such as slow wave amplitude, rise slope, and delta band energy, and also difficult to determine whether a single combined stimulation truly affects the current target slow wave event. Before formal sleep combined stimulation, this invention collects multiple TPCS artifact samples under the same stimulation parameters, electrode positions, acquisition equipment, and lead configuration as the formal experiment. These samples are aligned and averaged using the stimulation trigger time as the zero point to construct an individualized artifact template. During formal stimulation, the system extracts EEG segments after each stimulation and divides the artifact window into a stimulation initiation segment, a stimulation duration segment, and a post-stimulation decay segment. The template scaling factor is calculated for each corresponding time period, and then segmented template subtraction is performed to accommodate differences in artifact amplitude and decay patterns in different stimulation segments. After template subtraction, the system evaluates the correction quality using indicators such as residual peak-to-peak value, residual root mean square value, and correlation between residuals and artifact templates. If the requirements are met, the segment is used as a valid EEG window for subsequent immediate response analysis; if the requirements are not met, the segment is directly removed to avoid residual artifacts affecting the enhancement effect assessment and parameter adjustment. Therefore, this invention improves the availability and reliability of short-term EEG signals after stimulation, providing a more accurate data foundation for single slow-wave response assessment and closed-loop optimization.

[0059] This invention provides a closed-loop parameter optimization method based on real-time EEG response, enabling combined stimulation to move beyond fixed-parameter open-ended outputs and instead adaptively adjust subsequent stimulation strategies based on actual EEG changes after each or several stimulations. Existing slow-wave enhancement techniques typically employ preset sound intensity, electrical stimulation intensity, or fixed stimulation sequence, making it difficult to adapt to dynamic differences among different subjects, sleep cycles, and slow-wave events. This can easily lead to problems such as insufficient stimulation, unstable enhancement effects, or excessive stimulation causing sleep disturbances. This invention, after the combined TPCS and sound stimulation output, first removes TPCS stimulation artifacts by segmenting and scaling artifact templates to restore the short-term effective EEG window after stimulation. Then, it extracts real-time response indicators such as slow-wave amplitude, rise slope, and delta band energy, and compares them with individualized baseline values ​​corresponding to baseline natural slow-wave events to determine whether the current combined stimulation produces an effective enhancement effect. Based on this, the system dynamically adjusts parameters such as sound stimulus intensity according to the real-time enhancement score: when the enhancement effect reaches the preset standard, the current parameters are maintained; when the enhancement effect is insufficient, the stimulus intensity is slightly increased within a safe range; when the enhancement effect does not improve or decreases, it reverts to the previous effective parameters. Thus, this invention forms a closed-loop control process of "joint stimulus output—artifact correction—real-time response analysis—parameter optimization adjustment—next joint stimulus output," which can improve the individual adaptability of stimulus parameter settings, the stability of slow-wave enhancement effects, and the safety of the sleep intervention process.

Claims

1. A closed-loop sleep regulation method combining transcranial pulsed electrical stimulation and acoustic stimulation, characterized in that, Based on a closed-loop sleep slow wave regulation system, it includes a physiological signal acquisition module, a sleep stage judgment module, a target slow wave identification and phase triggering module, a TPCS and sound joint stimulation control module, an artifact removal module, and an instant response evaluation and parameter adjustment module. The physiological signal acquisition module collects multi-channel physiological signals during the subject's sleep process in real time, including electroencephalogram (EEG), electrooculogram (EOG), and electromyogram (EMG) signals, and transmits them to the sleep stage determination module in real time. After receiving the multichannel physiological signal, the sleep stage determination module inputs it into the pre-trained sleep staging model to make a real-time judgment on the subject's current sleep stage. When the sleep staging model determines that the subject is in stage N3 and the N3 stage determination result remains stable within a preset continuous time window, the target slow wave recognition and phase triggering process is initiated. The target slow wave recognition and phase triggering module superimposes and averages the signals from leads FP1, FP2, F3, and F4 in the multi-channel physiological signal to obtain the frontal slow wave composite signal; it then identifies the target slow wave event based on the frontal slow wave composite signal; after identifying the target slow wave event, it performs phase estimation on the subsequent frontal slow wave composite signal; when the frontal slow wave composite signal enters the preset target phase interval, it generates a stimulation trigger command and sends it to the TPCS and sound-based stimulation control module. After receiving the stimulation trigger command, the TPCS and sound combined stimulation control module controls the simultaneous output of transcranial pulse electrical stimulation and sound stimulation. The two stimuli work together on the same preset target phase interval to achieve combined stimulation. After the combined stimulation is completed, the artifact removal module performs TPCS artifact correction on the EEG signal after the combined stimulation. The instant response assessment and parameter adjustment module receives the EEG signal after TPCS artifact correction, assesses the instant response caused by the current combined stimulation, and adjusts the stimulation parameters of the combined stimulation trigger phase, acoustic stimulation, and transcranial pulsed electrical stimulation based on the assessment results.

2. The closed-loop sleep regulation method combining transcranial pulsed electrical stimulation and acoustic stimulation according to claim 1, characterized in that, The physiological signal acquisition module uses a 32-channel EEG acquisition system; 28 channels of EEG signals are acquired through a head-mounted EEG cap, with reference electrodes placed on the left and right earlobes; 2 channels of EOG electrodes are acquired, with the right EOG electrode located above the right outer canthus and the left EOG electrode located above the left outer canthus; and 2 channels of electromyography electrodes are acquired near the mandible.

3. The closed-loop sleep regulation method combining transcranial pulsed electrical stimulation and acoustic stimulation according to claim 1, characterized in that, In the sleep stage determination module, the sleep stage classification model adopts a CNN-BiLSTM model that combines a convolutional neural network and a bidirectional long short-term memory network. The output sleep stage classification results include wakefulness, N1 stage, N2 stage, N3 stage and REM sleep stage.

4. The closed-loop sleep regulation method combining transcranial pulsed electrical stimulation and acoustic stimulation according to claim 1, characterized in that, The process of target slow-wave event recognition in the target slow-wave recognition and phase triggering module is as follows: Based on the sign change of the frontal slow wave composite signal, the position where the frontal slow wave composite signal changes from positive to negative is defined as the falling zero-crossing point, and the position where it changes from negative to positive is defined as the rising zero-crossing point. The frontal slow wave composite signal between the falling zero-crossing point and the adjacent rising zero-crossing point is extracted as a candidate slow wave segment. For each candidate slow wave segment, its negative peak amplitude and the positive peak amplitude before the negative peak are detected, and the negative half-wave duration and peak-to-peak value are calculated. Candidate slow wave segments that simultaneously meet the conditions of 0.25s ≤ negative half-wave duration ≤ 1.25s, negative peak amplitude below a set threshold, and peak-to-peak value height above a set threshold are selected as target slow wave events.

5. The closed-loop sleep regulation method combining transcranial pulsed electrical stimulation and acoustic stimulation according to claim 1, characterized in that, The process of phase estimation and stimulus triggering command generation of the subsequent frontal slow-wave synthesis signal in the target slow-wave recognition and phase triggering module is as follows: After downsampling and smoothing the frontal slow wave composite signal following the target slow wave event, a Hilbert transform is performed to obtain the analytical signal; the instantaneous phase angle of the analytical signal is extracted and converted into an angle value; based on the average angle value of the most recent sampling points, it is determined whether the current composite signal has entered the preset target phase interval of 0°–90°. If the target phase interval is entered, a stimulus trigger command is generated and sent to the TPCS and sound-based combined stimulus control module.

6. The closed-loop sleep regulation method combining transcranial pulsed electrical stimulation and acoustic stimulation according to claim 1, characterized in that, In the TPCS and sound-based combined stimulation control module, the sound stimulation uses short pink noise pulses, with a single sound stimulation duration set to 50ms; the duration of word transcranial pulse electrical stimulation is also set to 50ms. The baseline sound intensity is determined based on the individual hearing threshold of the subject or the sound intensity that can be tolerated during sleep. The stimulation intensity of transcranial pulsed electrical stimulation is based on 1mA. During the combined stimulation process, the intensity of both sound stimulation and transcranial pulsed electrical stimulation can be dynamically adjusted within a small range of 10% above and below their respective baseline values ​​based on the immediate response assessment results.

7. The closed-loop sleep regulation method combining transcranial pulsed electrical stimulation and acoustic stimulation according to claim 1, characterized in that, In the artifact removal module, the specific process of TPCS artifact correction is as follows: During the resting state or non-sleep combined stimulation phase, the same stimulation parameters, electrode positions, multi-channel physiological signal acquisition equipment, and lead configuration as the combined stimulation were used to continuously apply 50 transcranial pulse electrical stimulations, while simultaneously acquiring EEG signals. For each EEG signal segment corresponding to a transcranial pulse electrical stimulation, a fixed time window from 50 ms before stimulation to 500 ms after stimulation was extracted, with the stimulation trigger mark as the time zero point. For each EEG signal segment, the mean of the EEG signal after stimulation was calculated by subtracting it from the mean of the EEG signal 50 ms before stimulation, i.e., baseline correction of the EEG signal after stimulation was performed using the short time window before stimulation. Outliers were removed from the baseline-corrected EEG signals, and the remaining effective baseline-corrected EEG signals were superimposed and averaged to construct an individualized artifact template for the subject under the current transcranial pulse electrical stimulation conditions. For the EEG signal after combined stimulation, segments of the same length as the individualized artifact template were extracted, and a segmented scaling strategy was used for TPCS artifact correction. The EEG signal after combined stimulation was divided into three time periods: 0-10ms as the stimulation onset period, 10-50ms as the stimulation duration period, and 50-500ms as the post-stimulation decay period. Calculate the template scaling factor for each time period: in, Let t be the scaling factor for the j-th time interval, and t be the time. For the j-th time period, The signal is the baseline-corrected EEG signal after combined stimulation. For individualized artifact templates; For each time period, TPCS artifact correction is performed to obtain the correction results. : The correction results are evaluated using residuals, with evaluation indicators including residual peak-to-peak value, residual root mean square value, and correlation between residuals and individualized artifact templates. If all three evaluation indicators meet the set threshold requirements, the result is considered a valid correction result.

8. The closed-loop sleep regulation method combining transcranial pulsed electrical stimulation and acoustic stimulation according to claim 1, characterized in that, In the immediate response assessment and parameter adjustment module, the amplitude changes, rise slope changes, and delta band energy changes of the EEG signal after TPCS artifact correction are extracted and compared with the corresponding features of the EEG signal before combined stimulation to determine the enhancement effect of the current combined stimulation on the target slow wave event. If the enhancement effect meets the preset requirements, the current stimulation parameters are maintained; otherwise, the stimulation parameters of the combined stimulation trigger phase, acoustic stimulation, and transcranial pulse electrical stimulation are adjusted.

Images