Sleep deprivation system based on closed-loop acoustic stimulation and photobiomodulation and use method
Through closed-loop sound stimulation and photobiological regulation technology, combined with the 'single pendulum effect' model, the individual recognition algorithm applies stimulation on the falling edge of the slow wave, solving the individual differences and nighttime changes of the sleep regulation algorithm in the existing technology, realizing sleep deprivation The effect of the sleep quality is improved.
Patent Information
- Application Number
- CN202410431426.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-11
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2044-04-11
AI Technical Summary
The existing sleep regulation algorithms have individual differences and nighttime changes when identifying slow wave signals, resulting in inaccurate detection results and lack of individualized adjustments and physiological feedback mechanisms, which cannot effectively achieve sleep deprivation.
Using a sleep deprivation system based on closed-loop sound stimulation and photobiological regulation, combined with the 'single pendulum effect' model, inverse stimulation is applied at the falling edge of the slow wave, and the slow wave DOWN state is accurately located through an individualized slow wave recognition algorithm, using auditory and photobiological stimulation Combined technical means to achieve sleep deprivation.
By accurately controlling the coupling of the brain region in the sleep, interrupting the continuity of slow wave signals, reducing the stability of the sleep structure, achieving a specific sleep deprivation effect, and improving the reduction of sleep quality.
Smart Images

Figure CN118253009B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of bioinformatics, and in particular relates to an individualized slow wave recognition algorithm and a sleep deprivation system based on closed-loop acoustic stimulation and photobiological regulation technology and a method of using the same. Background Art
[0002] Sleep is essential to our physical and mental health. Slow-wave sleep is in the non-rapid eye movement (NREM) sleep stage 3 (NREM3, N3). During this period, brain signals alternate between the depolarized UP state and the hyperpolarized DOWN state at a frequency of 0.5 to 2 Hz, forming slow wave oscillations.
[0003] At present, research on regulating sleep slow waves has made significant progress, most of which aims to enhance slow waves in order to improve macroscopic sleep quality or cognitive function. Existing studies have shown that applying stimulation to the UP state of sleep slow waves can effectively enhance slow waves, increase the amplitude of slow waves, and improve sleep; while applying stimulation to the DOWN state of slow waves will inhibit the amplitude of slow waves to a certain extent, and even reduce the number of consecutive slow waves.
[0004] Physical means of regulating sleep include transcranial magnetic stimulation, transcranial electrical stimulation, sensory stimulation (auditory, visual), photobiomodulation, etc. Among them, due to its strong spatial focusing advantage, transcranial magnetic stimulation is currently a widely used physical treatment method for insomnia in clinical practice, but it has the disadvantages of bulky size, high cost, and low comfort; transcranial electrical stimulation can directly cause biphasic, polarity-dependent electrical changes in the cerebral cortex, but will bring stimulation artifacts during sleep monitoring; and auditory stimulation can present specific types of sounds through the auditory pathway, thereby promoting brain relaxation, reducing tension, and regulating breathing and heart rate; photobiomodulation technology is a non-invasive phototherapy control method that uses red light or near-infrared light as a control source. It uses a series of photobiostimulation effects triggered by endogenous chromophores in the organism to adjust the body's physiological and biochemical reactions, and has the advantages of simple implementation and no artifacts. Compared with magnetic stimulation and electrical stimulation, auditory stimulation is more portable and easy to implement. Due to its real-time and continuous stimulation level, it can provide immediate, fast and relatively accurate physiological basis and feedback for regulating brain activity, achieve precise rhythm or oscillation phase tracking at the EEG level, and achieve real-time closed-loop precise regulation. However, auditory stimulation mainly acts on the temporal lobe of the brain.
[0005] The most common manipulation methods for sleep regulation are designed to be independent of the brain state, that is, the stimulation is applied in an open-loop manner. In contrast, closed-loop auditory stimulation (CLAS) has gradually emerged as a promising technology, taking into account the endogenous activity or oscillations of the brain. Researchers selectively target and present tones through the auditory pathway to enhance or suppress specific brain waves during sleep. By assessing brain activity in real time through electroencephalography (EEG), CLAS can accurately identify the specific oscillations and phases of specific sleep stages, and subsequently control the timing and phase of auditory stimulation so that the delivery of auditory stimulation coincides with the specific stage of ongoing sleep oscillations.
[0006] So far, there are many methods for real-time detection of slow wave signals, such as amplitude method, sine fitting, phase-locked loop, terrain template, etc. Different algorithms are based on different operating principles. Some directly use the most obvious physiological characteristics of slow waves for detection, some transform slow wave signals from time domain to frequency domain for detection, and some continue to use formulas and methods in related fields to detect slow wave signals. In general, these algorithms have their own characteristics. Some algorithms detect different types of slow waves, while others only detect and analyze specific slow waves on a certain issue. However, they may all ignore the differences between individuals and the random / inherent changes of slow wave signals at night, which will have a significant impact on the slow wave real-time detection algorithm.
[0007] In terms of the recognition accuracy of the algorithm: the recognition of slow waves cannot be guaranteed, and there will often be problems with false positive test results. For example, the core of most algorithms is to extract the instantaneous features of EEG signals, and then compare the proximity of the detected features with the slow wave features to make predictions for recognition; simple closed-loop algorithms only provide feedback based on instantaneous features, and the detection algorithm does not use the individual as a template benchmark for reference and parameter adjustment.
[0008] In terms of spatial position: the changes in the position of electrodes in the collected brain areas and the differences in the synchronous brain areas during sleep were not taken into account, and no preliminary improvements were made based on physiological feedback.
[0009] Individual characteristics: The inherent / random changes in slow wave signals between individuals over time at night, as well as the unique signal feature differences between individuals were not taken into account by previous algorithms. Summary of the invention
[0010] In order to solve the above technical problems, the present invention proposes a sleep deprivation system based on closed-loop acoustic stimulation and photobiomodulation and a method of use. Based on the "pendulum effect" model of sleep slow waves, reverse stimulation is applied on the falling edge of the slow waves, and a sleep deprivation system is provided. The system combines a personalized slow wave recognition algorithm to accurately locate the slow wave DOWN state, and achieves sleep deprivation through the technical means of combining auditory stimulation with photobiomodulation.
[0011] One of the technical solutions adopted by the present invention is: a sleep deprivation system based on closed-loop acoustic stimulation and photobiomodulation, comprising: an interactive interface, a control module, an EEG acquisition module, an individualized slow wave recognition algorithm module and a stimulation module;
[0012] The interactive interface is used to input stimulation information, wherein the stimulation information includes a stimulation threshold, the number of stimulations in each slow wave cycle, a stimulation duration corresponding to each stimulation intensity, a delayed stimulation time, and a stimulation address;
[0013] The control module sends a signal acquisition instruction to the EEG acquisition module according to the stimulation information inputted by the interactive interface; and based on the EEG signal uploaded by the EEG acquisition module, the control module calls the individualized slow wave recognition algorithm module to perform slow wave recognition on the EEG signal;
[0014] After the individualized slow wave identification algorithm module matches the moment when the slow wave appears, the control module encapsulates each stimulation address and the corresponding stimulation parameters into multiple stimulation sub-commands according to the stimulation information input through the interactive interface, and then sends them to the stimulation module in sequence according to the order of the stimulation addresses;
[0015] The stimulation module obtains stimulation information according to the received stimulation sub-command, and the stimulation channel corresponding to the stimulation address generates a driving voltage corresponding to the current stimulation sub-command to drive and produce a stimulation effect.
[0016] The stimulation module includes an auditory stimulation unit and a light biostimulation unit; the auditory stimulation unit acts on the temporal lobe area of the brain, and the light biostimulation unit acts on the frontal lobe area of the brain.
[0017] The individualized slow wave identification algorithm module matches the moment when the slow wave appears, specifically the slow wave DOWN state.
[0018] The second technical solution adopted by the present invention is: an application method of a sleep deprivation system based on closed-loop acoustic stimulation and photobiomodulation, comprising:
[0019] S1. Collecting the first EEG data of the test subject using a sleep deprivation system based on closed-loop acoustic stimulation and photobiomodulation, and establishing a personalized template based on the first EEG data;
[0020] S2, collecting the second EEG data of the test subject based on the sleep deprivation system of closed-loop acoustic stimulation and photobiomodulation, and then identifying the duration of the DOWN state of the slow wave falling edge based on the personalized template established in step S1;
[0021] S3. The peak of the slow wave UP state is used as the starting position of the stimulation, the auditory and light stimulation sequence is started, and the stimulation duration is controlled according to the duration T of the stimulation predicted by the duration of the slow wave falling edge DOWN state.
[0022] Beneficial effects of the present invention: In NREM sleep, slow oscillations reflect changes in the resting membrane potential of thalamic and cerebral cortical neurons, which switch between depolarized UP (active) and hyperpolarized DOWN (quiescent) states. In the UP state, thalamocortical cells (relay neurons) and cortical thalamic cells show strong synaptic activity (excitation and inhibition) and bursts of action potentials, while in the DOWN state, their membrane hyperpolarization induces a relative quiescent period. The alternation of UP and DOWN states during NREM sleep produces slow oscillations with a frequency of <1 Hz. The present invention uses closed-loop auditory stimulation and photobiomodulation to act on the slow-wave DOWN state in the NREM stage, and inhibits the conversion of brain neuron signals between the depolarized UP active state and the hyperpolarized DOWN state through sound and light, interrupting the coupling between the main EEG activities of the frontal lobe and temporal lobe brain regions, and can destroy the continuity and intensity of the brain's spontaneous slow wave signals in the NREM sleep stage, thereby reducing the stability of the sleep structure and the sleep quality, and achieving specific sleep deprivation. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] Figure 1 A "single pendulum effect" model of sleep slow waves provided by an embodiment of the present invention;
[0024] Among them, (a) is the case where there is no external force; (b) is the case where the external force is in the same direction as the ball's motion; (c) is the case where the external force is in the opposite direction to the ball's motion;
[0025] Figure 2 A schematic diagram of the brain regions affected by auditory stimulation and photobiomodulation provided by an embodiment of the present invention;
[0026] Figure 3 A schematic diagram of a stimulation sequence of auditory stimulation and light stimulation provided by an embodiment of the present invention;
[0027] Figure 4 A schematic diagram of a sleep deprivation system based on closed-loop acoustic stimulation and photobiomodulation provided by an embodiment of the present invention;
[0028] Figure 5 The international 10-20 EEG electrode placement diagram provided in the embodiment of the present invention;
[0029] Figure 6 The selection of slow wave characteristics and the algorithm for predicting the stimulation time T provided in the embodiment of the present invention;
[0030] Figure 7 A schematic diagram of the relationship between the incremental increase in auditory stimulation intensity and sleep depth provided by an embodiment of the present invention;
[0031] Figure 8 The present invention provides a system implementation process and stimulation result analysis. DETAILED DESCRIPTION
[0032] To facilitate those skilled in the art to understand the technical content of the present invention, the present invention is further explained below with reference to the accompanying drawings.
[0033] The present invention proposes a "single pendulum effect" model of sleep slow waves, such as Figure 1 As shown. Simplifying the slow wave oscillation into the motion trajectory of a simple pendulum, in the absence of external force, the pendulum system will perform simple harmonic motion with constant frequency and amplitude. However, if a force is applied to the pendulum at a specific phase, the amplitude of the pendulum will increase or decrease, depending on the phase of the pendulum when the force is applied. Correspondingly, the system frequency will also change. In the regulation of slow waves, the external force here is equivalent to external stimulation, assuming that the direction of the external force is fixed to the left. Then the stage of the ball moving upward is equivalent to the UP state of the slow wave; similarly, the stage of the ball moving downward is equivalent to the DOWN state of the slow wave. When an external force is applied during the rising stage of the ball (stimulation is given in the UP state), the amplitude of the ball's swing will increase (the amplitude of the slow wave increases); when an external force is applied during the descending stage of the ball (stimulation is given in the DOWN state), the amplitude of the ball's swing will decrease (the amplitude of the slow wave decreases). Therefore, applying stimulation in the UP state of the slow wave can achieve slow wave enhancement and sleep improvement; applying stimulation in the DOWN state of the slow wave can achieve slow wave suppression and sleep deprivation.
[0034] A series of parameters of sleep slow waves are closely related to sleep quality. Based on the "single pendulum effect" model, the present invention proposes a new idea for achieving sleep deprivation, that is, applying stimulation of the same frequency and opposite direction in the DOWN state of sleep slow waves.
[0035] In view of the fact that auditory stimulation in the prior art mainly acts on the temporal lobe of the brain, in order to compensate for the singleness of auditory stimulation in regulating brain areas, the present invention combines photobiomodulation technology to simultaneously stimulate the frontal lobe of the brain. Researchers generally believe that the frontal lobe is the starting point of sleep slow waves, and this area responds more significantly to regulation. In the present invention, the auditory stimulation input from the natural channel is combined with the direct brain state regulation of photobiomodulation technology to establish a model of coordinated regulation of the primary-higher cortical pathway, and the sleep slow wave DOWN state of the two brain areas is regulated at the same time, further improving the effect of sleep deprivation, such as Figure 2 shown.
[0036] The sound types of auditory stimulation include white noise, pink noise, single tone, etc. The different sound types used will lead to differences in the slow wave regulation effect, which means that choosing the right type of stimulation sound is crucial to effectively regulate slow waves. The power spectral density of white noise is a constant, while the power spectral density of pink noise is inversely proportional to the frequency. Single tones include vowels "a", tones "b", etc. Studies have shown that using white noise, pink noise or single tone as the sound type of auditory stimulation can effectively regulate sleep slow waves, but using pink noise to regulate slow waves can stimulate stronger slow wave reactions. Based on this discovery, the auditory stimulation of the present invention chooses to use pink noise, hoping to achieve a more effective slow wave regulation effect. The volume of the auditory stimulation is determined in a personalized way, which is determined by the subject's sensitivity to the stimulation volume and the current sleep depth during the regulation process. In actual applications, the auditory stimulation device can be placed in an appropriate position next to the subject's head, and auditory stimulation is carried out in the form of a stereo playing sound.
[0037] In photobiomodulation technology, different wavelengths of light are absorbed, scattered, and reflected to different degrees by biological media and tissues. Existing studies have shown that mitochondrial cytochrome c oxidase (CCO) acts as a major photoreceptor in the wavelength range of red light to near-infrared light, which may play an important role in the mechanism of action of PBM therapy. In addition, the beneficial effects of near-infrared light with wavelengths such as 1064, 1068, and 1072nm have been found in various studies, which will be explained by another cellular mechanism: due to the slightly higher degree of light scattering at 800-850nm (the absorption peak of CCO), photons may not penetrate deeper into tissues. However, at 1064-1072nm, CCO absorbs less light than 800-850nm, and due to low hemoglobin absorption and low light scattering, light in this spectrum can easily pass through tissues and can stimulate CCO or ion channels located in deeper structures. In order to enable near-infrared light to effectively penetrate the skull and improve the efficiency of light stimulation, the wavelength of near-infrared light for light stimulation in the present invention is selected to be 1064nm, and the initial value and variation of the light power density of the light stimulation sequence are kept fixed while ensuring safety. The photobiomodulation device (light stimulation module) is divided into two parts: a laser and a current source. The laser probe needs to be close to the frontal lobe of the subject, and the current source is placed in an appropriate position next to the subject.
[0038] In order to explore the effect of stimulation phase on the slow wave regulation effect, a study applied the same auditory stimulation at each phase of the slow wave, and then compared the slow wave amplitude after regulation with that of the control group. The results showed that applying stimulation at the crest and trough of the slow wave can induce the strongest slow wave response (the crest corresponds to enhancement, and the trough corresponds to inhibition), which shows that the most effective phase point for achieving sleep deprivation is at the trough of the slow wave. Therefore, in the present invention, auditory stimulation and light stimulation use a short-term, intermittent step-type stimulation sequence, acting on the entire DOWN state of the slow wave, and giving the maximum sound intensity and light power stimulation at the trough, such as Figure 3 shown.
[0039] Among them, D1 (dB) is the threshold of auditory stimulus intensity, n1 (dB) is the increment of auditory stimulus intensity, D2 (mW / cm2) is the threshold of light stimulus power density, n2 (mW / cm2) is the increment of light stimulus power density, T (s) is the duration of the slow wave falling edge, and the determination of the stimulation parameters will be described in detail later. In each slow wave cycle, the number of auditory stimulation and light stimulation is fixed to 5 times, and the stimulation duration corresponding to each stimulation intensity is 50ms (including the rising and falling edges of 5ms), the stimulation interval t of auditory and light stimulation is the same, and is determined by the duration of the slow wave falling edge, and the auditory and light stimulation intensities corresponding to the zero-crossing moment of the slow wave are the threshold intensities D1 and D2 of the two.
[0040] like Figure 4 As shown, the present invention provides a sleep deprivation system based on closed-loop acoustic stimulation and photobiomodulation, including: an interactive interface, a control module, an EEG acquisition, an individualized slow wave recognition algorithm module and a stimulation module.
[0041] The user can set the threshold of auditory and light stimulation, the number of stimulations in each slow wave cycle, the stimulation duration corresponding to each stimulation intensity, the delayed stimulation time, the stimulation address and other parameters on the interactive interface. These parameters will be integrated into the stimulation command and then sent to the control module;
[0042] The control module receives the stimulation command issued by the user, and sends the signal collection instruction to the EEG collection module. When the EEG signal is transmitted to the control module, the signal preprocessing algorithm starts to run, and the individualized slow wave recognition algorithm is called to identify the slow wave of the EEG signal. When the moment of the slow wave is matched, the control module will encapsulate each stimulation address and the corresponding stimulation parameters into multiple stimulation sub-commands according to the stimulation information issued previously, and then send them to the stimulation module in the order of the stimulation address.
[0043] The stimulation module receives the stimulation sub-commands sent by the control module in sequence, obtains the stimulation information, and the stimulation channel corresponding to the stimulation address generates the driving voltage corresponding to the current stimulation sub-command, driving the earphones and laser to generate auditory and light stimulation.
[0044] The specific implementation process of EEG acquisition is as follows:
[0045] EEG collection was divided into two consecutive nights, namely the first baseline night and the second experimental night.
[0046] During the first night of real-time EEG acquisition, the subject's hearing threshold will be determined. Sound stimulation will be used to observe the response to sound in the EEG activity during the subject's NREM sleep stage 2. When a K complex wave appears, the sound volume will be used as the subject's initial hearing threshold volume. The EEG data will then be fully collected. After the first night of sleep data collection is completed, the data will be sent to a professional sleep physician for sleep stage scoring and slow wave marking during sleep. The slow waves will be detected using an offline slow wave detection algorithm and compared with the slow waves marked by the sleep physician. Appropriate slow wave data will be selected, slow wave features will be extracted, and templates will be established using the extracted slow wave features to provide available templates. On the second night, real-time EEG data will be collected and analyzed.
[0047] A 32-channel EEG acquisition system was used (the electrodes were placed according to the international EEG 10-20 electrode standard, such as Figure 5 As shown in the figure, a head-mounted EEG cap is used to collect 28 channels of EEG (reference electrodes are located on the left and right earlobes); 2 channels of electrooculogram (EOG), where the right EOG electrode is located above the right outer canthus and the left EOG electrode is located above the left outer canthus; and 2 channels of electromyography (EMG) signals near the mandible. Among them, the system sampling rate is 500Hz, and filters are used for high-pass filtering (0.3Hz cutoff frequency) and low-pass filtering (70Hz cutoff frequency), and notch filtering is performed at 50Hz to remove power frequency noise. After pre-processing such as artifact removal and downsampling, the next step of real-time processing is carried out.
[0048] The EEG acquisition area is placed in accordance with the standard electrode position on both the first and second nights. The difference is: on the first night, no EEG of any channel is selected, and the EEG signal of each channel is fully collected. On the second night, we will use the EEG data of the same subject on the first night to find the areas with more slow wave origins and the best synchronization; these areas may appear in the frontal lobe, may appear in the temporal lobes on both sides, or may appear in multiple areas at the same time. These features will be added to the personal slow wave template, and the data will be actively collected according to the template position on the second night. For example, after analyzing the data of the first night, it was found that the area around the Fz electrode in the frontal lobe area is the area with the most slow wave origins, and the slow wave synchronization around several of its electrodes is very large. We will set Fz as the seed point in the frontal lobe area on the second night, and extract the target signal and detect the upcoming slow wave by averaging the voltage on the surrounding electrodes centered on the seed point. Similarly, seed points can also be set in multiple areas.
[0049] The process of creating a personalized template is:
[0050] Find the origin of slow waves and the area of slow wave synchronization and add them to the template; the slow wave oscillation signals on multiple channels are not precisely synchronized in all channels, and there is a measurable time lag relationship between them. Offline extraction of each slow wave signal of the baseline sleep night for negative envelope analysis can determine the origin of the slow wave (i.e. the starting point of a single slow wave in a certain brain area) within the life cycle of a single slow wave on all channels at the same time, and comprehensively count the origin information of all slow waves to find one or more regional points where the slow wave origin appears more. In order to count the local regional features during the slow wave propagation process, the slow wave data in multiple channels are subjected to coherence analysis. If a high degree of coherence is generated between an EEG channel and the origin EEG channel and the two electrodes are adjacent in position, it is considered that the slow wave has passed through the channel during the propagation process. The multiple channels detected will converge into one area on the EEG topography, and this area represents the synchronization area of the slow wave.
[0051] In order to embed the extracted baseline night slow wave data into the real-time detection algorithm, considering the changes of slow wave signals between individuals over time at night and the unique signal feature differences between individuals, the detection algorithm needs to use individuals as template benchmarks for reference and parameter adjustment; ten representative slow wave physiological features were initially selected, and the features can also be expanded in quantity and type, as shown in Table 1:
[0052] Table 1 Ten representative slow wave physiological characteristics
[0053]
[0054] The weights of slow wave physiological characteristics are generated using a multivariate linear regression model. Sleep depth is used as a macro indicator of nighttime slow wave changes. Sleep depth is simply defined as the logarithmic ratio between delta and beta power. The changes in sleep depth over time are monitored throughout the N3 phase of the baseline night. After traversing the entire N3 data, the maximum and minimum sleep depth values are recorded. According to the interval between the maximum and minimum values of the subjects, the sleep depth of the subjects is divided into 11 intervals to measure the intensity of sleep depth, such as the intervals less than the minimum value, 9 different intervals (between the minimum and maximum values), and the interval greater than the maximum value. When the sleep depth is detected in the N3 stage of a sleep cycle of the subject, every time the sleep depth change has a large deviation, the time point is recorded in the template, and the time axis between the time point of the last sleep depth change record and the time point of this record is used to intercept the EEG data of the N3 sleep stage of the subject for a period of time, and the features of multiple slow waves in this data are extracted; the weight of each feature is analyzed using a multivariate linear regression model to find out the weight of each feature, and these features and the corresponding time axis are added to the personal template. Similarly, when the next change comes, the weight of the feature within a period of time is also calculated until all the slow wave data of the N3 stage of the baseline night are traversed. The slow wave weight is applied in the real-time detection process of the algorithm. The algorithm will select the sleep depth retained by the template in the same timeline range according to the actual monitored sleep depth, select the feature weight corresponding to the sleep depth for the feature that has been detected at the current moment, and finally use the multivariate linear regression model again to make a numerical prediction of the stimulation time T. Due to the change of sleep depth, feature value and feature weight, the predicted value T will be in dynamic change in the real-time process, such as Figure 6 shown.
[0055] A large deviation in sleep depth change specifically refers to a range across one interval or multiple intervals; the EEG data of the subject's N3 sleep stage over a period of time is captured, specifically the time between two sleep depth changes.
[0056] The expression of the multiple linear regression model is as follows:
[0057] T=β 0 +β 1 x 1 +β 2 x 2 +…+β k x k
[0058] In offline analysis, T is the independent variable, which represents the array of time spent from the peak to the trough of each slow wave during offline detection. 1 ,x 2 ,…,x kis also an independent variable. Each x represents a different value of the same feature of multiple slow waves. Different x represents an array composed of the features of multiple slow waves. 0 is the intercept, β 1 ,β 2 ,…,β k It is the weight coefficient and also the dependent variable, which represents the generation of the slow wave feature weight under the sleep depth for a period of time.
[0059] In real-time analysis: T is the dependent variable, which indicates the time required for the output variable to be stimulated, x 1 ,x 2 ,…,x k is the independent variable, each x represents an array of multiple slow wave features, β 0 is the intercept, β 1 ,β 2 ,…,β k is the weight coefficient of the independent variable x, which represents the weight of the slow wave feature generated offline.
[0060] Based on the slow wave activity throughout the baseline night, the template will collect the following five categories of features:
[0061] Slow wave origin and synchronization area: used to determine the brain area involved in slow wave detection on the night of the experiment. Other brain areas are collected normally during the EEG signal collection process, but the collected data do not actively participate in the calculation of real-time slow wave detection.
[0062] As time changes throughout the night, the slow wave feature weights and dynamic time axis records play an important role in the real-time slow wave detection process. Over a period of time, each feature will get its own weight. Features with large weights will be marked as specific features, while features with small weights will be marked as non-specific features. In the real-time detection process, in order to avoid the appearance of a large number of false positive slow waves, it is necessary to judge whether the slow wave feature appears in a reasonable interval to make a rejection or acceptance judgment. The specific feature has a large change amount, and the interval is set to be unilateral, that is, the feature is greater than the minimum value or less than the maximum value. The threshold set; and the non-specific feature has a small change amount, the more stable the interval is set, and it is set to be bilateral, that is, it satisfies the interval set by both the minimum value and the maximum value. This processing method determines whether some irregular, erroneous, and false positive slow wave signals are excluded during the operation of the slow wave algorithm, thereby ensuring the robustness of the detection program to a certain extent. When making real-time predictions, a multivariate linear regression model is used to predict the time T of the slow wave DOWN state. Different templates under different sleep depths and dynamic time axes will select different feature weight values, and different weights and slow wave feature values affect the size of the predicted time T.
[0063] Sleep depth, which is defined by the logarithmic ratio between delta and beta power, characterizes the individual's macroscopic sleep situation in a short period of time. It can be used to preliminarily reflect the subject's sleep quality, as well as control the volume intensity and offline feature weight generation based on the linear / nonlinear relationship between the depth size.
[0064] Classic thresholds and zero-crossing points: Based on the data of the same subject's baseline night on the previous night, new individual thresholds are extracted and added to the template so that individualized thresholds can be applied in the classic real-time detection algorithm to preliminarily determine the presence of a slow wave; the zero-crossing point and peak data template of the subject on the previous baseline night are extracted and added to the subject's individual template.
[0065] The power of the local area of the forehead and the global power. The power of the local area of the forehead is obtained by the average power of the EEG data of several channels involved in the slow wave detection of the forehead, while the global power is obtained by the average power of the EEG channels of the whole brain. The ratio of the local area power to the global power reflects whether the target area has a strong change relative to the global power. It can be used to determine whether slow waves appear in large numbers during this time period, and whether the sleep in the local area is deepened / weakened.
[0066] The implementation process of the slow wave identification algorithm is as follows;
[0067] In the overall implementation of the algorithm, a detection strategy combining two methods is adopted to achieve the positioning of the entire slow wave physiological morphology and the control of the instantaneous zero-crossing point or UP state phase; coupled with a unique verification method, it ensures that the zero-crossing point phase is tracked in real time only after the continuous and effective slow wave physiological characteristics appear. The method is described as follows:
[0068] Identify the entire slow wave: First, verify the slow wave in the real-time detection process according to the classic slow wave detection standard, that is, when the three judgment conditions are met at the same time:
[0069] 1. The interval between the two zero crossing points (the zero crossing point of the DOWN state and the zero crossing point of the UP state) is between 0.3 and 1 second;
[0070] 2. The most negative peak value between the two zero crossing points must be smaller than -80μV;
[0071] 3. The amplitude between the most negative peak and the upcoming maximum peak must be greater than 140μV.
[0072] The thresholds of 0.3~1s, -80μV, and 140μV set for the three conditions of the classic detection program are empirical values. In actual operation, due to the differences in the hardware conditions of EEG acquisition and the individual subjects, it is often necessary to redefine the range of the thresholds. Therefore, based on the baseline data of the same subject on the previous night, we used an offline slow wave detection algorithm to detect several slow waves in the subject's N3 stage throughout the night, extracted the three features required by the classic detection program from the several detected slow waves, averaged each extracted feature to obtain the average value, so as to use it as a new threshold in the subject's detection program, and added these thresholds to the template so that individualized thresholds can be applied in the real-time detection process.
[0073] In order to avoid the appearance of a large number of false positive slow waves, after the initial detection of the slow wave, the algorithm will subsequently analyze the structural morphology of the slow wave. The analysis results will determine whether the slow wave detected at this moment is qualified, and whether the features listed in Table 1 are within the reasonable interval value of the slow wave features (this reasonable interval range is selected from the baseline night of the previous night). If all are met, it is considered to be accepted. If you choose to accept, you will continue to the next stage of the algorithm to accurately locate the zero point or UP state phase.
[0074] The accurate positioning of the zero-crossing point will continue to use the previous TOPOSO method, apply the zero-crossing point and peak features extracted from the template to the real-time data stream, and use the similar quantity and sign polarity relationship between the personalized template and the voltage amplitude to determine the zero-crossing point of the slow wave. Details can be found in the published patent application CN117045964A.
[0075] When the algorithm locks on a phase of the zero point or UP state, it will immediately identify the current sleep depth and quickly call the personalized template to compare the sleep depth corresponding to the current time with the data recorded in the template. After the match is completed, the corresponding feature weight will be found. This feature weight will match all the features in Table 1. At this point, all features have been matched with the corresponding weights and the expression of the multivariate linear regression model has been formed. The stimulation time T is immediately calculated as the predicted time for the slow wave from the peak of the UP state to the trough of the DOWN state during real-time detection.
[0076] The process of determining the stimulation parameters is:
[0077] After the algorithm starts stimulating, there will be a period of system delay and delay caused by the method until the auditory stimulation and light stimulation start from the peak point of the slow wave, and finally the stimulation stops when it reaches the trough of the slow wave; when a stimulation sequence is successfully completed, the system control module will end the stimulation and will not perform any operations within the next 3 seconds. After 3 seconds, the algorithm will start again and detect slow waves. The prediction of the parameters required during the stimulation period includes: the action time T(s) from the start point of the stimulation to the trough, and the increment n1(dB) of the volume determined by the sleep depth.
[0078] The number of auditory and light stimuli is also determined by the duration T(s) of the slow wave falling edge. The prediction of the stimulation duration T(s) is jointly predicted by multiple detected features. The specific weight value of each feature will be determined according to the individual template on the baseline night.
[0079] The intensity of the auditory stimulus is determined by the sleep depth and the strength of the slow wave oscillations within a time window. The sleep depth is simply defined as: the logarithmic ratio between delta and beta power, and the incremental increase of the pitch is linearly modulated by the sleep depth, so that stronger (or softer) auditory stimuli are played during deep (or shallow) sleep, such as Figure 7 as shown.
[0080] The implementation process of the system is as follows Figure 8 As shown, the following process is included:
[0081] Night 1 (Baseline Night):
[0082] The subjects were given an EEG cap and the electrodes were fixed: a 32-channel EEG acquisition system was used to collect 28 channels of EEG data using a head-mounted EEG cap, with reference electrodes located on the left and right earlobes; the right and left EOG signals were recorded simultaneously, with the right EOG electrode placed above the right outer canthus and the left EOG electrode placed above the left outer canthus; the electrode for collecting electromyography (EMG) signals was placed near the mandible, aligned with the EEG signal. Figure 1 Starts with automatic sleep staging.
[0083] After the EEG cap and electrodes are placed, start EEG acquisition.
[0084] First, in order to determine the subject's hearing threshold volume D1 (dB), during the EEG synchronous acquisition process, the subject's EEG signal was observed. When it was determined that the subject was in NREM sleep stage 2, auditory stimulation with a continuously increasing decibel value was applied until a clear K complex wave was observed in the EEG signal. The decibel value at this time was recorded as the subject's hearing threshold volume D1 (dB).
[0085] After the first night of sleep data collection is completed, the data will be sent to a professional sleep physician for sleep stage scoring and slow wave marking during sleep. Then, an offline slow wave detection algorithm will be used to detect slow waves and compare them with the slow waves marked by the sleep physician for slow wave screening. In order to be able to screen out the slow waves observed manually and the slow waves detected by the offline algorithm based on the digital characteristics of the EEG, the features of the slow waves are extracted and integrated into the slow wave algorithm, and the algorithm is used for auxiliary detection. The slow waves marked by the offline detection algorithm and the sleep physician are used as the slow wave data set, so as to extract the slow wave features and establish a personal template.
[0086] Based on the slow wave activity throughout the night on the baseline night, the following five types of features are collected and added to the template:
[0087] 1. The origin and synchronization area of slow waves were determined to determine the brain area for slow wave detection on the second night of the experiment.
[0088] 2. As time changes throughout the night, the record of slow wave feature weights will set the interval variation range of the features, which determines whether to exclude some irregular, erroneous, and false-positive slow wave signals during the operation of the slow wave algorithm, thereby ensuring the robustness of the detection program to a certain extent; and when making real-time predictions, due to different sleep depths and dynamic time axis templates will select different feature weight values, and the difference in weight values and the instantaneous size of the slow wave feature value affect the multivariate linear regression model set feature values and feature weights to predict the time T that the slow wave DOWN state will last.
[0089] 3. Sleep depth, which characterizes an individual’s macroscopic sleep status in a shorter period of time, can be used to preliminarily reflect the subject’s sleep quality.
[0090] 4. Classic thresholds and zero crossing points. The three classic thresholds are obtained by averaging the test values of the subjects from the previous night, and are used by the slow wave detection algorithm to preliminarily judge the occurrence of slow waves. The zero crossing point and peak point data templates are also obtained from the baseline data of the previous night, and are used by the slow wave detection algorithm to judge the real-time phase of the zero crossing point or the UP state.
[0091] 5. The local area power and global power of the forehead reflect whether the target area has stronger slow wave changes relative to the global area.
[0092] Second Night (Experimental Night):
[0093] The subjects were given an EEG cap, electrodes were placed, and an infrared laser was fixed so that light stimulation could directly illuminate the subject's frontal lobe.
[0094] Start EEG acquisition, determine the main area for acquisition, and use the EEG data collected in this area for subsequent detection; before starting the algorithm to detect slow waves, the real-time data needs to be preprocessed and artifacts detected to ensure the standardization and availability of the data; then, after observing a large number of slow wave signals in the subject's EEG signal, manually start the slow wave detection algorithm, use the personalized sleep template obtained the previous night to assist the slow wave algorithm in identification, and record the subject's sleep depth in an instant. Use the sleep depth and several features of slow waves under the dynamic time axis to find the feature weights corresponding to several slow wave feature values recorded in the first baseline night, mark which features are specific features and which are non-specific features at this sleep depth according to the feature weights, and set different detection intervals for different features; when the slow wave algorithm preliminarily detects a slow wave through the classical algorithm, determine whether the slow wave feature appears in a reasonable interval, so as to make a decision to reject or accept the slow wave detection algorithm at that moment as true. If the algorithm detection result is accepted as true, these features and corresponding feature weights will continue to be used in the real-time detection process, and the multivariate linear regression model will be used to calculate the predicted time T for auditory stimulation and light stimulation to accurately act on the slow wave falling edge DOWN state. Finally, after determining the real-time phase information of the zero point or rising UP state, after a period of system response and method delay, the peak of the slow wave UP state is used as the starting position of the stimulation, and the auditory and light stimulation sequence is started. The stimulation action time is controlled according to the predicted stimulation maintenance time T to generate a stimulation sequence; for example, the total time of the entire slow wave DOWN state is predicted to be T (s), and at this sleep depth, a stimulation sequence contains 5 sound / light stimulations. Then the stimulation interval time can be obtained by subtracting the total time of 5 sound / light stimulations from the total time, so as to obtain the total interval time, and then divide this interval time by 4 intervals to obtain the time t of each sound / light interval. In the sound stimulation sequence, the maximum volume of the sound stimulation is determined by the sleep depth at this moment, and the minimum volume is determined by the auditory threshold D1 of the baseline night. The maximum sound intensity is subtracted from the minimum volume to obtain the total sound intensity increment, which is then divided by the number of 4 increment intervals to obtain the sound intensity increment n1 of each auditory stimulation. In the light stimulation sequence, the light stimulation power density threshold D2 and the light stimulation power density increment n2 are both fixed values. After determining the parameters required for these stimulation sequences, a stimulation sequence can be generated for each slow wave DOWN state. According to the order of this stimulation sequence, the sound stimulation and the light stimulation are played synchronously. After the last stimulation in the stimulation sequence is completed, the algorithm will be closed for 3s and be in the refractory period, and will be reopened after 3s to prepare for the next slow wave detection.
[0095] The main parameters of the stimulation sequence are:
[0096] 1. Hearing threshold D1 (dB): determined on the baseline night;
[0097] 2. Auditory stimulation sound intensity increment n1 (dB): determined based on the instantaneous change in the quantified value of sleep depth.
[0098] 3. Light stimulation power density threshold D2 (mW / cm2): fixed at 160mW / cm2. This energy intensity is one-fifth of the maximum allowable exposure to the skin. The light dose used is within the safety threshold and will not cause thermal damage to human skin.
[0099] 4. Light stimulation power density increment n2 (mW / cm2): fixed at 20mW / cm2.
[0100] 5. The interval time t(s) between auditory stimulation and light stimulation sequences: The time interval between sequences is automatically assigned according to the predicted time T.
[0101] Analysis of stimulation results
[0102] This embodiment identifies slow waves according to an individualized slow wave detection algorithm and uses auditory stimulation and photobiomodulation to accurately act on the slow wave DOWN state of specific brain areas. It uses a continuously changing stimulation sequence and uses the same-frequency and opposite stimulation to accurately act on the DOWN phase of the slow wave during the real-time propagation of the slow wave, thereby interrupting the coupling of EEG activities between brain areas, and can destroy the continuity and intensity of spontaneous slow wave signals of the brain during the NREM sleep stage, thereby reducing the stability of the sleep structure and the sleep quality, and achieving sleep deprivation in the NREM sleep stage.
[0103] In the data analysis of the stimulation results, the electroencephalographic physiological signals during sleep during the two nights were compared, and the electroencephalographic data of -3s and +3s before and after the stimulation began were also analyzed. Using the electroencephalographic signals as the representation, it was found that after the stimulation occurred, the low-frequency power in the sleep electroencephalogram decreased and the sleep structure changed.
[0104] First, all recorded stimulation epochs were averaged, and the power spectral density (PSD) of delta and SO was analyzed throughout the night. It was found that the power spectral density of delta and SO was smaller under stimulation conditions compared with the baseline night. This result was expected before the experiment, because previous studies have pointed out that under destructive stimulation conditions, stimulation will reduce the duration and proportion of the N3 stage, and may also reduce the continuity of the N2 and N3 stages, which is highly consistent with our research results.
[0105] Then, the auditory power and light power presented during real-time stimulation were classified post hoc, and the stimulation was divided into a low-dose stimulation group and a high-dose stimulation group. The results showed that compared with the low-dose stimulation group, the high-dose stimulation group significantly suppressed the power spectral density of delta and SO from the start of stimulation to the end of stimulation. It was further found that the greater the dose given at the falling edge of the slow wave, the greater the inhibitory effect on delta and SO.
[0106] Finally, the sleep structure of the subjects changed. Under the stimulation conditions of the experimental night, compared with the baseline night, the proportion of the N3 stage in their sleep stages decreased and the duration was also shorter, and the percentage of the N3 sleep stage in the whole night's sleep stages decreased; and under the stimulation conditions, compared with the N1 stage on the baseline night, the duration of the N1 stage on the experimental night was significantly longer than that on the baseline night for the subjects who were subjected to auditory stimulation and photobiomodulation; in addition, the difference in the N2 stage between the experimental night and the baseline night was also compared, and it was found that no significant difference was observed between the two nights for the same subject.
[0107] Therefore, the destructive auditory and light stimulations applied during the experimental night reduced the sleep time and percentage of the N3 sleep stage in the overall sleep stage, but increased the sleep time and percentage of the N1 sleep stage. Moreover, when auditory stimulations that were appropriately higher than the basic auditory threshold but lower than the auditory arousal threshold were used to act on the DOWN state of the slow wave decline, the instantaneous delta and SO power could be significantly reduced, which may arouse the subjects to a higher arousal index and reduce sleep quality. To a certain extent, it can be considered that the destructive stimulations applied during the experimental night affected and changed the sleep structure, disrupted the electrophysiological dynamic state of the spontaneous slow waves during propagation, and achieved sleep suppression and sleep deprivation.
[0108] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the principles of the present invention, and should be understood that the scope of protection of the present invention is not limited to such specific statements and embodiments. For those skilled in the art, the present invention may have various changes and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of the claims of the present invention.
Claims
1. A sleep deprivation system based on closed-loop acoustic stimulation and photobiomodulation, characterized in that: include: Interactive interface, control module, EEG acquisition module, individualized slow wave identification algorithm module and stimulation module; The interactive interface is used to input stimulation information, and the stimulation information includes a stimulation threshold, the number of stimulations in each slow wave cycle, the stimulation duration corresponding to each stimulation intensity, a delayed stimulation time, and a stimulation address; The control module sends a signal acquisition instruction to the EEG acquisition module according to the stimulation information inputted by the interactive interface; and based on the EEG signal uploaded by the EEG acquisition module, the control module calls the individualized slow wave recognition algorithm module to perform slow wave recognition on the EEG signal; The EEG acquisition module collected EEG signals for two nights. When collecting EEG signals on the first night, the electrodes were placed based on the standard electrode positions to completely collect EEG signals from each channel. The individualized slow wave recognition algorithm module establishes a personalized template based on the EEG signals collected on the first night; When collecting EEG signals on the second night, the electrode positions were determined based on the personalized template, and the electrodes were placed to collect EEG signals of each channel determined based on the personalized template; The individualized slow wave recognition algorithm module recognizes slow waves and verifies individualized slow waves based on the EEG signals collected on the second night and the personalized template; After the individualized slow wave identification algorithm module matches the moment when the slow wave appears, the control module encapsulates each stimulation address and the corresponding stimulation parameters into multiple stimulation sub-commands according to the stimulation information input through the interactive interface, and then sends them to the stimulation module in sequence according to the order of the stimulation addresses; The stimulation module obtains stimulation information according to the received stimulation subcommand, and the stimulation channel corresponding to the stimulation address generates a driving voltage corresponding to the current stimulation subcommand, and drives to produce stimulation; the stimulation module includes an auditory stimulation unit and a light biostimulation unit; the auditory stimulation unit acts on the temporal lobe area of the brain, and the light biostimulation unit acts on the frontal lobe area of the brain; Auditory stimulation and light stimulation adopt a short-term, intermittent step-type stimulation sequence. During the real-time propagation of the slow wave, the same-frequency and opposite stimulation are used to act on each entire DOWN state of the slow wave; the same-frequency and opposite stimulation are applied to each entire DOWN state of the slow wave, specifically: auditory stimulation and light stimulation are applied from the peak point of each entire DOWN state, and finally the stimulation stops after reaching the trough of the slow wave. The number of auditory stimulation and light stimulation is determined by the duration of the falling edge of each entire DOWN state, and the intensity increases with time.
2. A sleep deprivation system based on closed-loop acoustic stimulation and photobiomodulation according to claim 1, characterized in that: The EEG acquisition module includes: a head-mounted EEG cap, 2-channel electrooculogram signal electrodes, and 2-channel electromyogram signal electrodes; the head-mounted EEG cap includes 28 channels of EEG signal electrodes; The 2-channel electrooculogram signal electrodes are located above the right outer canthus and the left outer canthus respectively; The 2-channel EMG signal electrodes are located on the mandible; The 28-channel EEG signal electrodes include 2-channel reference electrodes located on the left and right earlobes respectively; the remaining EEG signal electrodes are placed according to the international EEG 10-20 electrode standard.
3. An application method of the sleep deprivation system based on closed-loop acoustic stimulation and photobiomodulation according to claim 1 or 2, characterized in that: include: S1. Collect the EEG data of the first night from the test subjects using the sleep deprivation system based on closed-loop acoustic stimulation and photobiomodulation, and establish a personalized template based on the EEG data of the first night; when collecting the EEG data of the first night, place the electrodes based on the standard position to completely collect the EEG signals of each channel; S2, collecting EEG data of the second night from the test subject using the sleep deprivation system based on closed-loop acoustic stimulation and photobiomodulation, and then identifying the duration of the DOWN state of the slow wave falling edge based on the personalized template established in step S1; placing the electrodes based on the personalized template when collecting the EEG data of the second night, and collecting EEG data of each channel determined based on the personalized template; S3. Take the peak of the slow wave UP state as the starting position of the stimulation, start the auditory and light stimulation sequence, and control the stimulation duration based on the duration T of the stimulation predicted by the duration of the slow wave falling edge DOWN state; the auditory and light stimulation sequence adopts a short-term, intermittent step-type stimulation sequence.
4. The application method of the sleep deprivation system based on closed-loop acoustic stimulation and photobiomodulation according to claim 3, characterized in that: The personalized template in step S1 specifically collects the following features of slow waves: Slow wave origin, synchronization area, slow wave physiological characteristics, characteristic weights of slow wave physiological characteristics that change over time, sleep depth, local frontal area power and global power; At the same time, the interval time between the zero-crossing point of the DOWN state and the zero-crossing point of the UP state corresponding to each identified slow wave, the most negative peak between the zero-crossing point of the DOWN state and the zero-crossing point of the UP state, and the amplitude between the most negative peak and the upcoming maximum peak are extracted. The average value of these three features is taken as the corresponding zero-crossing time threshold and amplitude threshold in the personalized template, which is used for the preliminary detection of slow waves in the EEG data of the second night.
5. According to the application method of the sleep deprivation system based on closed-loop acoustic stimulation and photobiomodulation as described in claim 4, the slow waves used in step S1 to establish the personalized template are the slow waves jointly marked by the offline slow wave detection algorithm and the sleep physician; if the EEG data collected on the first night meets the required conditions of the slow wave zero-crossing time and amplitude, it is determined to be a slow wave.
6. According to the application method of the sleep deprivation system based on closed-loop acoustic stimulation and photobiomodulation according to claim 5, step S2 updates the zero-crossing time threshold and amplitude threshold of the detection algorithm by the zero-crossing time threshold and amplitude threshold in the personalized template before identifying the slow wave; after the slow wave is preliminarily detected by using the updated zero-crossing time threshold and amplitude threshold, the personalized slow wave is verified according to the personalized template, and the specific process is: according to the slow wave physiological characteristic interval in the personalized template, it is judged whether the slow wave feature appears in the individual interval, so as to make a judgment of rejecting or accepting the slow wave detection algorithm detecting the personalized slow wave at this moment.
7. According to the application method of the sleep deprivation system based on closed-loop acoustic stimulation and photobiomodulation according to claim 6, the specific process of predicting the duration T of the stimulation in step S3 is: according to the sleep depth corresponding to the duration of the DOWN state of the identified slow wave falling edge, according to the corresponding feature weights in the personalized template matching of the slow wave physiological characteristics, a multivariate linear regression model is used to obtain the predicted stimulation duration T.
