Closed-loop control electrical stimulation system for targeting parafacial nerve nucleus to induce slow wave sleep

Abstract

Disclosed is a closed-loop control electrical stimulation system for targeting the parafacial nerve nucleus to induce slow wave sleep, comprising an electrical stimulation module for electrically stimulating the parafacial nerve nucleus of a target to be detected. An electroencephalogram signal of said target is acquired, and signal processing is performed on the electroencephalogram signal to obtain a sleep state recognition module of a current sleep stage of said target. The sleep state recognition module uses a general time series network, which comprises a convolutional neural network unit, a fast Fourier transform unit, a circadian time unit, and a result output unit. According to the present invention, the sleep state can be determined, and an effect of slow wave sleep induced by the previous electrical stimulation of the parafacial nerve nucleus is evaluated in a closed loop. When the obtained sleep state is determined to be awake or rapid eye movement sleep, it indicates that the parafacial nerve nucleus needs to be electrically stimulated at this time. Whether the next electrical stimulation is to be performed can be continuously and cyclically determined on the basis of the effect of the previous electrical stimulation, and a fast-induced and long-lasting slow wave sleep effect is obtained after multiple closed-loop cycles.

Description

A closed-loop control electrical stimulation system for inducing slow wave sleep by targeting the facial parabrachial nucleus TECHNICAL FIELD

[0001] The present application belongs to the field of medical devices, and particularly relates to a closed-loop control electrical stimulation system for inducing slow wave sleep by targeting the facial parabrachial nucleus. BACKGROUND

[0002] Maintaining slow wave sleep for a certain duration at night is a prerequisite for achieving optimal physiological, psychological and cognitive functions. Studies on animals and humans have shown that slow wave sleep can be promoted by activating gamma-aminobutyric acid neurons in the mammalian pontine facial parabrachial nucleus. After activation, the facial parabrachial nucleus neurons synapses release gamma-aminobutyric acid to the parabrachial nucleus neurons, which in turn project and release glutamate to the cortical projection neurons of the basal forebrain. Activating gamma-aminobutyric acid neurons in the facial parabrachial nucleus can effectively trigger slow wave sleep and regulate cortical electroencephalogram activity. However, there is currently no technology to directly induce and maintain slow wave sleep, making it difficult to clinically translate this target.

[0003] In the prior art, slow wave sleep is induced by using optical genetic regulation: through stereotactic surgery, a light fiber is implanted on one side of the facial parabrachial nucleus of a VGAT-ChR2-EYFP mouse (implantation coordinates: AP = -5.5 mm; ML = 1.4 mm; DV = -4.2 mm), and a stimulator (PG4000A, Cygnus Technology, Inc., Delaware Water Gap, PA) is used to stimulate using a light pulse sequence (wavelength 473 nm, frequency 40 Hz). The disadvantage of optical genetic regulation is that long-term irradiation of light-sensitive channel proteins can cause protein inactivation, and slow wave sleep cannot be maintained for a long time.

[0004] Slow wave sleep can also be induced by using chemical genetic regulation: through stereotactic surgery, a hSyn-DIO-hM4D-mCherry-AAV virus is injected to selectively express hM4Di receptors in the gamma-aminobutyric acid neurons of the facial parabrachial nucleus of a Vgat-IRES-cre mouse (injection coordinates: AP = -5.3 mm, ML = ± 0.7 mm, DV = -3.4 mm). The receptor can bind with the drug clozapine-N-oxide to activate neurons expressing the receptor. By injecting clozapine-N-oxide into the peritoneal cavity of the mouse, gamma-aminobutyric acid neurons in the facial parabrachial nucleus can be specifically activated to induce slow wave sleep. However, the disadvantage of chemical genetic regulation is that clozapine-N-oxide takes more than half an hour to take effect, and slow wave sleep cannot be induced immediately.

[0005] SUMMARY

[0006] The technical purpose of the present application is to provide a closed-loop control electric stimulation system for targeting facial nucleus to induce slow wave sleep, so as to solve the problem of slow wave sleep.

[0007] To solve the above problems, the technical scheme of the present application is:

[0008] A closed-loop control electric stimulation system for targeting facial nucleus to induce slow wave sleep, comprising:

[0009] An electric stimulation module configured to perform electric stimulation on the facial nucleus of the target to be detected;

[0010] A sleep state recognition module configured to collect electroencephalogram signals of the whole brain of the target to be detected, and perform signal processing on the electroencephalogram signals to obtain the sleep stage of the target to be detected;

[0011] The sleep state recognition module adopts a general time series network, which includes a convolutional neural network unit, a fast Fourier transform unit, a day-night time unit, and a result output unit.

[0012] The convolutional neural network unit is configured to receive the electroencephalogram signals for convolution calculation to obtain convolution calculation results; the fast Fourier transform unit is configured to receive the electroencephalogram signals for fast Fourier calculation to obtain fast Fourier transform results; the day-night time unit is configured to output day-night rhythm time information; and the result output unit is configured to receive the convolution calculation results, the fast Fourier transform results, and the day-night rhythm time information and combine them to obtain a three-dimensional vector, and determine the current sleep stage according to the three-dimensional vector.

[0013] Specifically, the electric stimulation module includes an electric stimulation generation unit and an electrode.

[0014] The electric stimulation generation unit is electrically connected to the electrode and configured to output electric stimulation to the electrode, each electric stimulation including 1000 pulses with an intensity of 20uA and a pulse width of 5ms.

[0015] The electrode adopts two strands of nickel-titanium alloy twisted into a conductive wire with a length of 4mm, and the surface of the electrode is wrapped with an insulating layer, while the tail of the electrode is not wrapped with an insulating layer.

[0016] The convolutional neural network unit includes eight convolutional layers, which are:

[0017] The first convolutional layer adopts 64 convolutional kernels with a length of 9 to perform convolution operation on the electroencephalogram signals with a step size of 1.

[0018] The second convolutional layer adopts 64 convolutional kernels with a length of 9 to perform convolution operation on the data output by the first convolutional layer with a step size of 2.

[0019] The third convolutional layer uses 64 convolutional kernels with a length of 9 to perform convolutional operation on the data output by the second convolutional layer with a step of 2;

[0020] The fourth convolutional layer uses 64 convolutional kernels with a length of 9 to perform convolutional operation on the data output by the third convolutional layer with a step of 2;

[0021] The fifth convolutional layer uses 64 convolutional kernels with a length of 7 to perform convolutional operation on the data output by the fourth convolutional layer with a step of 1;

[0022] The sixth convolutional layer uses 64 convolutional kernels with a length of 7 to perform convolutional operation on the data output by the fifth convolutional layer with a step of 2;

[0023] The seventh convolutional layer uses 64 convolutional kernels with a length of 7 to perform convolutional operation on the data output by the sixth convolutional layer with a step of 2;

[0024] The eighth convolutional layer uses 64 convolutional kernels with a length of 7 to perform convolutional operation on the data output by the seventh convolutional layer with a step of 2.

[0025] Further preferably, a rectified linear unit activation layer configured to achieve nonlinear conversion of signals and a batch normalization layer configured to prevent gradient disappearance are further provided after each convolutional layer.

[0026] Further preferably, a first jump connection and a second jump connection are further provided;

[0027] The first jump connection is located between the output end of the first convolutional layer and the input end of the fifth convolutional layer, and is used to capture coarse features.

[0028] The second jump connection is located between the output end of the fifth convolutional layer and the input end of the result output unit, and is used to capture detailed features.

[0029] The fast Fourier transform spectrum in the fast Fourier transform unit ranges from 1 to 12 Hz, and the spectrum is divided into 22 intervals, each of which is 0.5 Hz.

[0030] Specifically, the result output unit includes a fully connected network layer, a Softmax function and an Argmax function;

[0031] The fully connected network layer is configured to receive and combine the convolutional calculation result, the fast Fourier transform result and the circadian rhythm time information and then output;

[0032] The Softmax function is configured to convert the information output by the fully connected network layer into a three-dimensional vector, which represents a vector of sleep stage probability distribution;

[0033] The argmax function is configured to receive a three-dimensional vector, determine the sleep stage with the highest output probability.

[0034] Further preferably, a master control module is further provided, and the master control module is in signal connection with the electric stimulation module and the sleep state recognition module respectively.

[0035] The master control module is configured to receive the current sleep stage output from the sleep state recognition module, determine the current sleep stage, if the current sleep stage is not a slow wave sleep stage, control the electric stimulation module to output corresponding electric stimulation, so that the target to be detected enters the slow wave sleep state, otherwise, no electric stimulation is sent and the sleep state of the target to be detected is continuously determined.

[0036] Compared with the prior art, the present application has the following advantages and positive effects:

[0037] The present application can determine the sleep state, and is used for closed-loop evaluation of the effect of the last time of electric stimulation of the facial nucleus to induce slow wave sleep, when the determined sleep state is wakefulness or rapid eye movement sleep, it is indicated that the facial nucleus needs to be electrically stimulated at this time, that is, the ideal slow wave sleep is induced. Based on the effect of the last time of electric stimulation, it is continuously determined whether the next time of electric stimulation is needed, and after a plurality of closed-loop cycles, the effect of rapid induction and long-lasting slow wave sleep is obtained. BRIEF DESCRIPTION OF DRAWINGS

[0038] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are for purposes of illustration only and are not considered a limitation of the present application.

[0039] Fig. 1 is a structural block diagram of the sleep state recognition module of the present application;

[0040] Fig. 2 is a signal flow block diagram of a closed-loop control electric stimulation system for inducing slow wave sleep by targeting the facial nucleus of the present application;

[0041] Fig. 3 is a structural block diagram of a closed-loop control electric stimulation system for inducing slow wave sleep by targeting the facial nucleus of the present application. DETAILED DESCRIPTION

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, specific embodiments of the present application will be described below with reference to the drawings. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings and other embodiments according to these drawings without creative labor.

[0043] For simplicity and concision of the drawings, only parts relating to the present application are shown in the drawings, which do not represent the actual structure of the product. In addition, for simplicity and convenience of understanding the drawings, in some drawings, only one of the parts having the same structure or function is shown schematically, or only one of them is marked. In this text, "one" not only means "only one", but also means "more than one".

[0044] The present application is further described in detail below with reference to the accompanying drawings and specific embodiments. The advantages and features of the present application will be more apparent from the following description and claims.

[0045] Embodiment

[0046] Referring to FIGS. 1 to 3, the present embodiment of a closed-loop control electrical stimulation system for inducing slow wave sleep targeting the facial nucleus includes three major modules: first, an electrical stimulation module for electrical stimulation to the facial nucleus of the target to be detected. Second, a sleep state recognition module for collecting electroencephalogram signals of the whole brain of the target to be detected, and processing the electroencephalogram signals to obtain the current sleep stage of the target to be detected. Third, a master control module, which is in signal connection with the above-mentioned electrical stimulation module and sleep state recognition module, can receive the current sleep stage of the target to be detected output from the sleep state recognition module, judge the current sleep stage, and if it is not a slow wave sleep stage (the sleep stage includes wakefulness, slow wave sleep, and rapid eye movement sleep), control the electrical stimulation module to output corresponding electrical stimulation to make the target to be detected enter the slow wave sleep state, otherwise, do not send electrical stimulation and continue to judge the sleep state of the target to be detected.

[0047] Specifically, the electrical stimulation module includes an electrical stimulation generation unit and an electrode. The electrical stimulation generation unit is electrically connected with the electrode and can output electrical stimulation to the electrode for electrical stimulation to the electrode at the facial nucleus position to achieve evoking and maintaining slow wave sleep. Each electrical stimulation includes 1000 pulses, the intensity is 20uA, and the pulse width is 5ms. In order to make the electrode more effectively stimulate the target point when implanted, the electrode in the present embodiment is twisted into a conductive wire with two strands of nickel-titanium alloy, the length is 4mm, and then the insulating material is sprayed onto the surface of the conductive wire at the part where the insulating layer is needed to form an insulating layer, so that the electrode only discharges at the tail discharge end. As an example, the experimenter implants the electrode near the facial nucleus of the mouse (target point coordinates: AP = -5.3mm, ML = ±0.7mm, DV = -3.4mm), uses the pulse of the electrode for electrical stimulation, and evokes slow wave sleep by electrical stimulation.

[0048] Then, the sleep state recognition module is described in detail. The sleep state recognition module can determine the sleep state of a mouse (a target to be detected) from whole brain electroencephalogram signals. The whole brain electroencephalogram signals can be collected and recorded by skull pin electrodes fixed near the anterior fontanel and the posterior fontanel. In this embodiment, a convolutional neural network is composed of a general time series network. The general time series network specifically includes the following structures: a convolutional neural network unit, a fast Fourier transform unit, a day-night time unit, and a result output unit. In order to make the general time series network suitable for closed feedback loop applications, the electroencephalogram signal time period window is set to 10 seconds. The electroencephalogram signals in this period are used as an input vector. Assuming that z is the input vector, the input vector is input into a d-dimensional convolution kernel κ. The output of the τth input vector after passing through the convolution layer network is a vector that satisfies the following formula:

[0049] Referring to FIGS. 1 and 2, the convolutional neural network unit is essentially a plurality of convolution layers connected in sequence, which is used to perform convolution calculation on the received electroencephalogram signals to obtain a convolution calculation result. Specifically, the first convolution layer uses 64 convolution kernels with a length of 9 to perform convolution operation on the electroencephalogram signals with a step size of 1. The second convolution layer uses 64 convolution kernels with a length of 9 to perform convolution operation on the data (128 sampling points) output by the first convolution layer with a step size of 2. The third convolution layer uses 64 convolution kernels with a length of 9 to perform convolution operation on the data (128 sampling points) output by the second convolution layer with a step size of 2. The fourth convolution layer uses 64 convolution kernels with a length of 9 to perform convolution operation on the data (128 sampling points) output by the third convolution layer with a step size of 2. The fifth convolution layer uses 64 convolution kernels with a length of 7 to perform convolution operation on the data (128 sampling points) output by the fourth convolution layer with a step size of 1. The sixth convolution layer uses 64 convolution kernels with a length of 7 to perform convolution operation on the data (128 sampling points) output by the fifth convolution layer with a step size of 2. The seventh convolution layer uses 64 convolution kernels with a length of 7 to perform convolution operation on the data (128 sampling points) output by the sixth convolution layer with a step size of 2. The eighth convolution layer uses 64 convolution kernels with a length of 7 to perform convolution operation on the data (128 sampling points) output by the seventh convolution layer with a step size of 2.

[0050] Further, a rectified linear unit (ReLU) and a batch normalization (BN) are further arranged at the back of each convolution layer. The rectified linear unit is used to realize nonlinear conversion of signals, and the batch normalization is used to prevent the gradient vanishing problem in the deep learning process. In addition, the neural network is further provided with a first skip connection and a second skip connection. The first skip connection is arranged between the output end of the first convolution layer and the input end of the fifth convolution layer, and is used to capture rough features. The second skip connection is arranged between the output end of the fifth convolution layer and the input end of the result output unit, and is used to capture detailed features.

[0051] Referring to FIG. 2, the fast Fourier transform unit is used to receive the electroencephalogram signal to perform fast Fourier calculation, and obtain a fast Fourier transform result. Specifically, the fast Fourier transform spectrum range is 1-12 Hz, which covers the key characteristic oscillation activities of the delta wave (1-4 Hz) and theta wave (6-9 Hz) frequency bands. The spectrum is divided into 22 intervals, each interval being 0.5 Hz. The circadian time unit is used to output circadian rhythm time information.

[0052] Referring to FIG. 2, the result output unit is used to receive the convolution calculation result, the fast Fourier transform result and the circadian rhythm time information and merge them to obtain a three-dimensional vector, and determine the current sleep stage according to the three-dimensional vector. Specifically, the result output unit includes a fully connected network layer, a Softmax function and an Argmax function. The fully connected network layer is used to receive the convolution calculation result, the fast Fourier transform result and the circadian rhythm time information and merge them to output. Since the task is multi-class classification, the Softmax function is used to receive the information output by the fully connected network layer and convert it into a three-dimensional vector, which represents the vector of sleep stage probability distribution. Finally, the Argmax function is used to receive the three-dimensional vector to determine the sleep stage with the highest output probability.

[0053] When the closed-loop deep brain stimulation based on the general time series network is performed, the principle is as follows: the electric stimulation of the facial nucleus is performed, and the whole brain electroencephalogram signal is acquired at the same time of the stimulation. The acquired whole brain electroencephalogram signal is processed, the convolution neural network, the fast Fourier transform result of the electroencephalogram signal and the output of the circadian rhythm time are merged, and are converted into a three-dimensional vector through the fully connected neural network. The vector corresponds to the probability of each sleep stage. The stage with the maximum probability is taken as the judgment of the sleep stage. If it is not the slow wave sleep stage, the stimulation device sends an electric pulse to the facial nucleus and continues to judge the sleep state. Otherwise, no electric pulse is sent and the sleep state is continued to be judged, so as to realize the closed-loop control.

[0054] The embodiments of the present application are explained in detail above with reference to the accompanying drawings, but the present application is not limited to the above-described embodiments. Even if various changes are made to the present application, if the changes fall within the scope of the claims of the present application and equivalents thereof, they are still within the protective scope of the present application.

Claims

1. A closed-loop control electrical stimulation system targeting parafacial nucleus to induce slow wave sleep, characterized in that, The method comprises the following steps: an electric stimulation module configured to electrically stimulate the facial nucleus of the target to be detected; a sleep state recognition module configured to collect electroencephalogram signals of the whole brain of the target to be detected, and to process the electroencephalogram signals to obtain the sleep stage of the target to be detected; wherein the sleep state recognition module adopts a general time sequence network, and the general time sequence network comprises a convolutional neural network unit, a fast Fourier transform unit, a day-night time unit and a result output unit; the convolutional neural network unit is configured to receive the electroencephalogram signals for convolution calculation to obtain a convolution calculation result; the fast Fourier transform unit is configured to receive the electroencephalogram signals for fast Fourier calculation to obtain a fast Fourier transform result; the day-night time unit is configured to output day-night rhythm time information; and the result output unit is configured to receive the convolution calculation result, the fast Fourier transform result and the day-night rhythm time information and combine them to obtain a three-dimensional vector, and to determine the current sleep stage according to the three-dimensional vector.

2. The closed-loop control electrical stimulation system targeting the parafacial nucleus to induce slow-wave sleep of claim 1, wherein, the electric stimulation module comprises an electric stimulation generation unit and an electrode; the electric stimulation generation unit is electrically connected with the electrode and is configured to output electric stimulation to the electrode, and each time the electric stimulation comprises 1000 pulses with an intensity of 20uA and a pulse width of 5ms; the electrode adopts two strands of nickel-titanium alloy twisted into a conductive wire with a length of 4mm, and the surface of the electrode is wrapped with an insulating layer, while the tail end of the electrode is not wrapped with the insulating layer.

3. The closed-loop control electrical stimulation system targeting the parafacial nucleus to induce slow-wave sleep of claim 1, wherein, the convolutional neural network unit comprises eight convolution layers in sequence: a first convolution layer adopting 64 convolution kernels with a length of 9 to perform convolution operation on the electroencephalogram signals with a step length of 1; a second convolution layer adopting 64 convolution kernels with a length of 9 to perform convolution operation on the data output by the first convolution layer with a step length of 2; a third convolution layer adopting 64 convolution kernels with a length of 9 to perform convolution operation on the data output by the second convolution layer with a step length of 2; a fourth convolution layer adopting 64 convolution kernels with a length of 9 to perform convolution operation on the data output by the third convolution layer with a step length of 2; a fifth convolution layer adopting 64 convolution kernels with a length of 7 to perform convolution operation on the data output by the fourth convolution layer with a step length of 1; a sixth convolution layer adopting 64 convolution kernels with a length of 7 to perform convolution operation on the data output by the fifth convolution layer with a step length of 2; a seventh convolution layer adopting 64 convolution kernels with a length of 7 to perform convolution operation on the data output by the sixth convolution layer with a step length of 2; an eighth convolution layer adopting 64 convolution kernels with a length of 7 to perform convolution operation on the data output by the seventh convolution layer with a step length of 2.

4. The closed-loop control electrical stimulation system targeting the parafacial nucleus to induce slow-wave sleep of claim 3, wherein, a rectified linear unit activation layer and a batch normalization layer are further arranged behind each convolution layer, the rectified linear unit activation layer is configured to realize nonlinear conversion of signals, and the batch normalization layer is configured to prevent gradient disappearance.

5. The closed-loop control electrical stimulation system targeting the parafacial nucleus to induce slow-wave sleep according to claim 3 or 4, characterized in that, a first jump connection and a second jump connection are further arranged; the first jump connection is located between the output end of the first convolution layer and the input end of the fifth convolution layer to capture rough features. The second jump connection is located between the output end of the fifth convolutional layer and the input end of the result output unit, and is used to capture detailed features.

6. The closed-loop control electrical stimulation system targeting the parafacial nucleus to induce slow-wave sleep of claim 1, wherein, The fast Fourier transform spectrum range in the fast Fourier transform unit is 1-12 Hz, and the spectrum is divided into 22 intervals, each interval being 0.5 Hz.

7. The closed-loop control electrical stimulation system targeting the parafacial nucleus to induce slow-wave sleep of claim 1, wherein, The result output unit comprises a fully connected network layer, a Softmax function and an Argmax function. The fully connected network layer is configured to receive and combine the convolution calculation result, the fast Fourier transform result and the circadian rhythm time information and then output. The Softmax function is configured to receive the information output by the fully connected network layer and convert it into a three-dimensional vector, which represents a vector of sleep stage probability distribution. The Argmax function is configured to receive the three-dimensional vector and determine the sleep stage with the highest output probability.

8. The closed-loop control electrical stimulation system targeting the parafacial nucleus to induce slow-wave sleep of claim 2, wherein, A main control module is further provided, which is in signal connection with the electric stimulation module and the sleep state recognition module respectively. The main control module is configured to receive the current sleep stage output by the sleep state recognition module, determine the current sleep stage, and if it is not a slow wave sleep stage, control the electric stimulation module to output corresponding electric stimulation so that the target to be detected enters a slow wave sleep state, otherwise, no electric stimulation is sent and the sleep state of the target to be detected is continuously determined, thereby realizing closed-loop control.

Images