CN-122004903-A - Brain electrical rhythm disturbance detection and sleep-aiding regulation and control method based on coherent network map

CN122004903ACN 122004903 ACN122004903 ACN 122004903ACN-122004903-A

Abstract

The invention provides an electroencephalogram rhythm disturbance detection and sleep-aiding regulation and control method based on a coherent network map, and relates to the technical field of sleep-aiding regulation and control. The method encodes the space-time synchronism of the brain electricity into the topological feature of the graph, can capture the topological disturbance of early rhythms before arousal occurs, has higher sensitivity and stronger anti-interference capability compared with single-channel feature detection, and realizes advanced early warning of the risk of rhythm collapse. And the phase equation of the full brain coupled oscillator is constructed based on a generalized kura Mo Tuo model, and phase prediction is optimized by combining the communication weight of a slow wave coherent network. The slow wave steady state manifold memory base line is updated by adopting an exponential moving average method, so that the natural dynamic change of the slow wave rhythm in the sleep period can be adapted, the detection error caused by the fixed base line is eliminated, and the accuracy of disturbance detection is improved.

Inventors

YU WENCHANG
WANG YONGQI
LOU LI
CHEN DEGANG
LIN HONG

Assignees

无锡特文思达健康科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260323

Claims (9)

1. The brain electrical rhythm disturbance detection and sleep-aiding regulation method based on the coherent network map is characterized by comprising the following steps of: S1, acquiring a multichannel original brain electrical signal, and obtaining brain electrolysis analysis signals of a slow wave frequency band and a high frequency band through artifact removal, frequency band separation and analysis signal conversion; s2, extracting the instantaneous phase of the channels based on the brain analysis signals, calculating the phase locking value between the channels, and constructing a dynamic coherent network map through distance punishment correction; S3, calculating a graph Laplace matrix based on the dynamic coherent network map, updating a slow wave steady state manifold memory baseline, and simultaneously extracting transient topological characteristics of a high frequency band; S4, calculating the Riemann distance between the transient topological feature of the high-frequency band and the steady-state manifold memory baseline of the slow wave, deducing the arousal posterior probability based on the Riemann distance, and when the arousal probability exceeds a preset threshold, calculating the time delay of the uplink phase of the slow wave; S5, based on the arousal posterior probability and the slow wave uplink phase time delay, generating a closed-loop stimulation parameter vector with phase locked by combining the user individuation slow wave dominant frequency, and performing stimulation according to the closed-loop stimulation parameter vector to complete sleep-aiding regulation.
2. The method for detecting and assisting sleep in disturbance of brain electrical rhythm according to claim 1, wherein step S1 specifically comprises: Obtaining scalp microvolts analog brain electrical signals, performing analog-to-digital conversion according to a preset sampling rate, and performing windowing processing on continuous signals according to a preset time window length T to obtain a multichannel brain electrical space-time matrix C is the number of channels; Spatial filter matrix based on pre-training For multichannel brain electricity space-time matrix Performing linear projection operation, and stripping electro-oculogram and myoelectricity space artifacts to obtain an electroencephalogram space-time matrix after artifact removal ; Electroencephalogram space-time matrix adopting zero-phase-shift Chebyshev II type band-pass filter Frequency band separation is carried out, and real-value brain electrical signals of 0.5-4Hz slow wave frequency band and 8-13Hz high frequency band are respectively extracted Applying Hilbert transform to the filtered real-valued EEG signals of each frequency band to obtain Hilbert transform result ; Filtering the filtered real-value brain electrical signal And corresponding Hilbert transform results Combining the complex-form analysis signals to obtain band-specific analysis signal tensor F, respectively taking slow wave and high frequency to correspondingly obtain a slow wave frequency band analysis signal And high frequency band resolved signals 。
3. The method for detecting and assisting sleep in disturbance of brain electrical rhythm based on coherent network map according to claim 2, wherein step S2 specifically comprises: resolving a signal tensor based on the band specific Extracting the instantaneous phase of each channel resolved signal ; Calculating phase-locked value between any two channels i and g ; Distance penalty term for introducing electrode physical three-dimensional coordinates : Wherein the ith acquisition channel corresponds to the p-th scalp electrode, the g acquisition channel corresponds to the q-th scalp electrode Pi Dianji, p and q are electrode indexes, and the value range is 、 ; The Euclidean distance between the p electrode and the q electrode; The distance coefficient is preset; Phase-locked value Distance penalty term Multiplying to obtain corrected connectivity weight ; Generating dynamic coherent network adjacency matrix , ; Each electroencephalogram acquisition channel is taken as a node set V, and dynamic coherent network adjacency matrix is used The elements in the method are edge weight sets among nodes Constructing and obtaining a coherent network map , 。
4. The method for detecting and assisting sleep in brain electrical rhythm disturbance based on coherent network map according to claim 3, wherein step S3 specifically comprises: based on the dynamic coherent network adjacency matrix Calculating a graphics matrix Based on the graph degree matrix Dynamic coherent network adjacency matrix Solving a graph Laplace matrix , Respectively calculating adjacent matrixes of the slow wave frequency band and the high frequency band to obtain a slow wave frequency band graph Laplacian matrix And high frequency band graph Laplace matrix ; Exponential sliding average method for slow wave steady-state manifold memory base line Updating: Wherein, the As a forgetting factor, A slow wave steady state manifold memory baseline updated for the t-th time window, Slow wave steady state manifold memory baseline, initial value for t-1 time window Taking an average matrix of a Laplacian matrix of the slow wave frequency band chart of the healthy crowd in the deep sleep period; Extracting a high-frequency band graph Laplacian matrix of a current time window in an arousal monitoring window with the same length as an electroencephalogram time window As a transient topological feature.
5. The method for detecting and assisting sleep in brain electrical rhythm disturbance based on coherent network map according to claim 4, wherein in step S4, a Riemann distance between transient topological feature of high frequency band and slow wave steady state manifold memory baseline is calculated, specifically comprising: baseline of slow wave steady state manifold memory And high frequency band graph Laplace matrix Mapping to symmetric positive flow space, where the high frequency band is mapped to a laplace matrix By adding Is converted into a symmetric positive definite matrix, Calculating affine invariant Riemann distance of the identity matrix and the identity matrix 。
6. The method for detecting and aiding sleep regulation and control of brain electrical rhythm disturbance based on coherent network map according to claim 5, wherein in step S4, the probability of arousal posterior is inferred based on the Riemann distance, specifically comprising: will affine invariant Riemann distance Inputting a pre-trained Bayes logistic regression model, and outputting the model as arousal posterior probability P, wherein the likelihood function of the Bayes logistic regression model is as follows: Wherein, the method comprises the steps of, As a function of the sigmoid, As the weight coefficient of the light-emitting diode, As a result of the bias term, Representing the occurrence of an arousal event.
7. The method for detecting and aiding sleep regulation and control of brain electrical rhythm disturbance based on coherent network map according to claim 6, wherein when the arousal posterior probability P is greater than a preset threshold, determining that there is a risk of rhythm collapse, and analyzing signals based on slow wave frequency bands Extracting a whole brain slow wave signal, and constructing a whole brain core oscillator phase equation based on a generalized kura Mo Tuo model: Wherein, the A phase change rate for the kth channel oscillator; Is the natural frequency of the kth channel slow wave; is the coupling strength; An adjacency matrix for the slow wave band; solving the numerical solution of the model by adopting a Longer-Kutta method with the step length of 1ms to obtain the comprehensive phase of the full brain core oscillator , Searching for a lead Least positive real number of (2) As an uplink phase time delay.
8. The method for detecting and assisting sleep in brain electrical rhythm disturbance based on coherent network map according to claim 7, wherein step S5 specifically comprises: Based on the arousal posterior probability P and the uplink phase time delay Generating a phase-locked closed-loop stimulation parameter vector by combining the user-individualized slow wave dominant frequency iSWF , ; Indicating a trigger delay that is to be triggered, The intensity of the stimulus is indicated and, Representing the stimulation frequency; The user individuation slow wave main frequency iSWF is obtained by carrying out frequency spectrum analysis on a sleep period slow wave signal of a user passing time period; Delaying triggers in stimulation parameter vectors Assigning as uplink time delay Will stimulate the frequency Assigning a value to a user's personalized slow wave dominant frequency iSWF, determining stimulus intensity , ; Is the maximum safe stimulation intensity of hardware.
9. The method for detecting and aiding sleep regulation and control of brain electrical rhythm disturbance based on coherent network map according to claim 8, further comprising, after stimulation according to a closed-loop stimulation parameter vector: collecting coherent network map of next time window , Taking the value as the length T of a time window, and calculating the affine invariant Riemann distance between the Laplace matrix of the high-frequency band chart and the current slow wave steady-state manifold memory base line in the time window Simultaneously extracting the slow wave average phase-locked value corresponding to the time window ; Construction of sleep homeostasis maintenance cost function : Wherein, the Weight coefficients for affine invariant Riemann distances; a weight coefficient which is the average phase-locked value of the slow wave; Regularization coefficients for the stimulus parameter vector; Is that The L2 norm of (2); maintaining cost function in sleep homeostasis The gradient descent direction of (2) is reversely fine-tuned to the coupling strength K, and the self-adaptive optimization of closed-loop regulation and control is completed.

Description

Brain electrical rhythm disturbance detection and sleep-aiding regulation and control method based on coherent network map Technical Field The invention relates to the technical field of sleep-aiding regulation and control, in particular to an electroencephalogram rhythm disturbance detection and sleep-aiding regulation and control method based on a coherent network map. Background Sleep disorder becomes a global high-rise public health problem, chronic insomnia can cause organic injury of nervous, endocrine and metabolic systems, conventional drug treatment has the risks of addiction, residual effect and long-term side effects, and non-invasive electroencephalogram closed-loop nerve regulation is a core research and development direction of the current noninvasive sleep-aiding technology. The brain electrical slow wave rhythm is a core biomarker of deep sleep, the phase synchronism among the whole brain channels directly determines the maintenance capability of sleep stable state, the core of sleep fragmentation induces arousal, and early topological disturbance of the brain electrical rhythm can occur before occurrence. The current closed-loop sleep-aiding technology is used for detecting one-dimensional characteristics such as amplitude, phase and the like of multi-focus single-channel electroencephalogram, is insufficient in utilization of rhythm coherence among multiple channels and dynamic topology change of a brain network, is difficult to realize advanced detection of micro-consciousness and accurate phase locking regulation and control, and needs to develop a high-sensitivity self-adaptive closed-loop sleep-aiding technical scheme based on dynamic characteristics of the brain network. Firstly, the existing mainstream scheme is mainly based on single-channel electroencephalogram characteristics for detection, only one-dimensional time domain and frequency domain attributes such as amplitude, frequency and the like can be captured, rhythm synchronicity changes among all brain multiple channels can not be represented, only the synchronous changes can be identified after arousal occurs, advanced early warning can not be realized, single-channel signals are easily interfered by electro-oculography and myoelectricity artifacts, and false detection and omission rate of arousal events are high. Secondly, the existing phase locking stimulation scheme relies on single-channel phase detection, the coupling relation among all brain channels is not considered, phase prediction is easily affected by local noise, the time delay resolving precision of an uplink phase is insufficient, the stimulation cannot stably lock an optimal regulation window of a slow wave uplink phase, and the sleep-aiding regulation effect has large individual difference and poor stability. Thirdly, the existing closed-loop regulation and control scheme mostly adopts fixed baselines and fixed stimulation parameters, natural dynamic changes of slow wave rhythms in a sleep period cannot be adapted, self-adaptive optimization cannot be performed according to individual rhythms of users, nerve adaptation effects are easy to occur after long-term use, and regulation and control effects continuously decay. Fourth, the prior art does not establish quantitative association between brain network topology difference and sleep steady state, can not construct a differentiable closed loop optimization objective function, is difficult to realize end-to-end iterative optimization of regulation parameters, and can not balance regulation effect and stimulation safety. Disclosure of Invention The invention mainly aims to provide an electroencephalogram rhythm disturbance detection and sleep-aiding regulation and control method based on a coherent network map, which constructs a dynamic coherent network map based on multichannel electroencephalogram signals, encodes the space-time synchronicity of the electroencephalogram into map topological features, can capture early-stage rhythm topological disturbance before arousal occurs, has higher sensitivity and stronger anti-interference capability compared with single-channel feature detection, and realizes advanced early warning of rhythm collapse risk. Secondly, a phase equation of the full brain coupled oscillator is constructed based on a generalized kura Mo Tuo model, phase prediction is optimized by combining the communication weight of a slow wave coherent network, compared with single-channel phase detection, the time delay resolving precision of an uplink phase is higher, the exogenous stimulation is ensured to be accurately locked with an optimal regulation window of the uplink phase of the slow wave, and the enhancing effect of the slow wave synchronism is greatly improved. Thirdly, the slow wave steady state manifold memory base line is updated by adopting an exponential moving average method, so that the natural dynamic change of the slow wave rhythm in the sleep period can be adapted, the detection e