Sleep state recognition device and implantable closed-loop stimulation system suitable for sleep state

Abstract

The invention provides a sleep state recognition device and an implantable closed-loop stimulation system suitable for a sleep state, and is applied to the technical field of medical instruments. The device comprises a processor which is used for preliminarily judging whether a user is in a sleep state or not according to user state information and / or clock readings; when the user is in the suspected sleep state, acquiring a local field potential LFP collected by an electrode and an acceleration collected by a triaxial accelerometer; judging whether the user is in the sleep state again according to the local field potential LFP and the acceleration; and when the user is in a sleep state, determining a sleep stage according to the local field potential LFP. The system comprises a sleep state recognition device and a controller, wherein the controller is used for carrying out proportional regulation by adopting a stimulation amplitude corresponding to a sleep stage and outputting a stimulation signal. The sleep state of the user is accurately recognized, the stimulation parameters are intelligently adjusted, and the stimulation signals are output, so that the deep part of the brain of the user is automatically stimulated, individualized treatment is achieved, and the treatment effect is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical devices, and in particular to a sleep state recognition device and an implantable closed-loop stimulation system suitable for the sleep state. Background Art

[0002] Implantable neurostimulation systems include Deep Brain Stimulation (DBS), which has significant therapeutic effects on a variety of intractable neurological diseases. The pulse generator and electrodes are implanted in the body, and the pulse generator transmits electrical pulses to specific areas of the brain through the electrodes to control the symptoms of the disease. The external device can communicate with the pulse generator and adjust the stimulation parameters of the pulse generator to achieve different stimulation effects. For example, the stimulation position can be changed by adjusting the positive and negative poles of the contact point, and the amplitude, pulse width, and frequency can be modified to change the range of stimulation. The doctor can perform programming based on experience and the patient's response to determine a set of fixed parameters for continuous stimulation.

[0003] However, the physiological state of patients fluctuates. For example, Parkinson's patients have different physiological states related to medication, sleep and other conditions, and their stimulation needs and intensities vary. If continuous stimulation consumes a lot of power, especially for non-rechargeable products, continuous high-frequency stimulation will affect the normal neural circuits to a certain extent, which may have a certain impact on the patient's cognition, behavior, emotions and other aspects. Summary of the invention

[0004] In view of this, the present invention provides a sleep state recognition device, comprising a processor;

[0005] The processor is used to preliminarily determine whether the user is in a sleep state based on user status information and / or clock indications, wherein the user status information includes at least one of acceleration, biorhythm, electrocardiogram, and electroencephalogram; when the user is in a suspected sleep state, obtain the local field potential LFP collected by the electrodes of the implantable closed-loop stimulation system and the acceleration collected by the three-axis accelerometer of the implantable closed-loop stimulation system; determine again whether the user is in a sleep state based on the local field potential LFP and the acceleration; when the user is in a sleep state, determine the sleep stage based on the local field potential LFP.

[0006] Optionally, preliminarily determining whether the user is in a sleeping state includes:

[0007] Determine whether the clock indication is within a preset time period;

[0008] When the clock indication is within a preset time period, determining that the user is in a suspected sleeping state;

[0009] When the clock indication is not within the preset time period, it is determined that the user is not in the sleeping state.

[0010] Optionally, preliminarily determining whether the user is in a sleeping state includes:

[0011] At preset intervals, the acceleration is used to preliminarily determine whether the user is in a sleeping state, wherein:

[0012] Extract time domain features of the three-axis acceleration, including the acceleration mean and acceleration standard deviation;

[0013] Inputting the acceleration mean and the acceleration standard deviation into two pre-trained sleep state prediction models to perform sleep state probability prediction, and multiplying the two prediction results to obtain a sleep probability value;

[0014] Preliminarily determining whether the user is in a sleeping state according to the sleeping probability value;

[0015] When the sleep probability value is greater than or equal to a preset probability value, it is determined that the user is in a suspected sleep state;

[0016] When the sleep probability value is less than a preset probability value, it is determined that the user is not in a sleep state.

[0017] Optionally, determining again whether the user is in a sleeping state includes:

[0018] Determining whether the local field potential LFP has a frequency division artifact;

[0019] When the local field potential LFP does not have a frequency division artifact, time domain feature extraction is performed on the three-axis acceleration to obtain the acceleration mean and acceleration standard deviation;

[0020] Inputting the acceleration mean and the acceleration standard deviation into two pre-trained sleep state prediction models to perform sleep state probability prediction, and multiplying the two prediction results to obtain a sleep probability value;

[0021] determining again whether the user is in a sleeping state according to the sleeping probability value;

[0022] When the sleep probability value is greater than or equal to a preset probability value, it is determined that the user is in a sleep state;

[0023] When the sleep probability value is less than a preset probability value, it is determined that the user is not in a sleep state.

[0024] Optionally, determining the sleep stage according to the local field potential LFP comprises:

[0025] Performing power spectrum estimation and denoising on the local field potential LFP;

[0026] The denoised power spectrum is combined into multiple frequency band energies and normalized;

[0027] Performing dimension reduction on the normalized multiple frequency band energies to obtain multiple dimension reduction features, wherein the multiple dimension reduction features include a first dimension reduction feature w1, a second dimension reduction feature w2, a third dimension reduction feature w3, and a fourth dimension reduction feature w4;

[0028] The multiple dimension reduction features are input into the pre-built ARIMA model to obtain the predicted value set W ′ ;

[0029] According to the predicted value group W ′ Determine sleep stages, which include wakefulness, rapid eye movement, and non-rapid eye movement.

[0030] Optionally, the predicted value group W ′ Including the first dimension reduction feature prediction value w1 ′ , the second dimension reduction feature prediction value w2 ′ , the third dimension reduction feature prediction value w3 ′ And the fourth dimension reduction feature prediction value w4 ′ , the first dimension reduction feature prediction value w1 ′ , the second dimension reduction feature prediction value w2 ′ , the third dimension reduction feature prediction value w3 ′ And the fourth dimension reduction feature prediction value w4 ′ Each contains multiple predicted values;

[0031] According to the predicted value group W ′ Determine sleep stages, including:

[0032] Calculate the mean of the predicted values ​​of each dimension reduction feature;

[0033] The first optimal classification threshold and the first dimension reduction feature prediction value w1 are preset ′ The mean values ​​were compared to determine the rapid eye movement period or wakefulness period and the non-rapid eye movement period;

[0034] When it is determined to be a rapid eye movement period or a wakeful period, the second optimal classification threshold and the second dimension reduction feature prediction value w2 are preset ′ The mean values ​​were compared to distinguish between REM and wakefulness.

[0035] The present invention also provides an implantable closed-loop stimulation system suitable for sleep state, comprising the sleep state recognition device and a controller;

[0036] The controller is used to proportionally adjust the output stimulation signal using the stimulation amplitude corresponding to the sleep stage, wherein different sleep stages correspond to different stimulation amplitudes.

[0037] Optionally, when it is preliminarily determined that the user is not in a sleeping state, it is determined whether it is nighttime according to the clock indication, and if it is nighttime, a stimulation signal is outputted using a preset stimulation amplitude S0;

[0038] When it is determined again that the user is not in the sleeping state, a stimulation signal is outputted using the preset stimulation amplitude S0.

[0039] Optionally, when it is determined that the local field potential LFP has a frequency division artifact, a preset stimulation amplitude S0 is used to output a stimulation signal.

[0040] Optionally, for the awake period in the sleep staging, a stimulation signal is outputted using a preset stimulation amplitude S0;

[0041] For the non-rapid eye movement period in the sleep stage, the corresponding stimulation amplitude S1 of the output stimulation signal is:

[0042] S1=-a3×(N1-N3)+N1

[0043] Among them, a3 is the sleep state coefficient of the non-rapid eye movement period, N1 is the upper limit of the stimulation of the non-rapid eye movement period, and N3 is the lower limit of the stimulation of the non-rapid eye movement period;

[0044] For the rapid eye movement stage in sleep, the corresponding stimulation amplitude S2 of the output stimulation signal is:

[0045] S2=a4×(PT)+T

[0046] Among them, a4 is the sleep state coefficient of the rapid eye movement period, P is the upper limit of the rapid eye movement stimulation, and T is the lower limit of the rapid eye movement stimulation.

[0047] The sleep state recognition device provided by the present application combines user state information (such as acceleration, biological rhythm, electrocardiogram, electroencephalogram) and clock indications for preliminary judgment, and can quickly judge whether the user is likely to be in a sleep state, and further judge by obtaining the local field potential LFP and acceleration collected by the electrodes of the implantable closed-loop stimulation system, further improving the accuracy and reliability of the judgment. This dual judgment mechanism helps to reduce misjudgments and missed judgments, and ensure the accuracy of subsequent sleep staging.

[0048] The implantable closed-loop stimulation system for sleeping provided by the present application determines that the user is in a sleeping state and performs sleep staging when the processor determines that the user is in a sleeping state, and after the sleep staging is performed, the controller determines the corresponding stimulation mode for the user according to the sleep staging result, and outputs the corresponding stimulation signal according to the stimulation mode, thereby improving the accuracy of the stimulation parameters. By accurately identifying the user's sleep state and staging according to the sleep state recognition device, the controller can intelligently adjust the stimulation parameters and output the stimulation signal to achieve automatic stimulation of the user's deep brain and optimize the treatment effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0050] Figure 1 It is a schematic diagram of a process of a processor in a sleep state recognition device identifying a sleep state in an embodiment of the present invention;

[0051] Figure 2 A flowchart of a method for a processor in a sleep state recognition device to preliminarily determine whether a user is in a sleep state in an embodiment of the present invention;

[0052] Figure 3 It is a flowchart of another method for a processor in a sleep state recognition device to preliminarily determine whether a user is in a sleep state in an embodiment of the present invention;

[0053] Figure 4 It is a schematic diagram of a flow chart of a processor in a sleep state recognition device in an embodiment of the present invention determining again whether a user is in a sleep state;

[0054] Figure 5 The figure is a schematic diagram of a process of determining sleep stages by a processor in a sleep state recognition device in an embodiment of the present invention. DETAILED DESCRIPTION

[0055] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0056] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicating the orientation or positional relationship, are based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", and "third" are used for descriptive purposes only, and cannot be understood as indicating or implying relative importance.

[0057] In the description of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, it can also be the internal connection of two components, it can be a wireless connection, or it can be a wired connection. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0058] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0059] An embodiment of the present invention provides an implantable closed-loop stimulation system suitable for a sleep state, the system comprising a sleep state recognition device, electrodes, and a controller. The sleep state recognition device comprises an implantable pulse generator, the implantable pulse generator comprises a processor and a three-axis accelerometer, the processor can recognize the sleep state of a human body, the three-axis accelerometer is used to collect acceleration information; the electrodes are used to collect local field potential (LFP); the controller is used to proportionally adjust the output stimulation signal using a stimulation amplitude corresponding to the sleep stage, wherein different sleep stages correspond to different stimulation amplitudes, the controller is communicatively connected to the processor, and parameters related to the stimulation signal adjustment can be configured according to the recognition result of the processor.

[0060] In some other embodiments, the sleep state recognition device can be integrated in the implantable pulse generator or in the external programmable device, that is, the sleep state recognition can be performed by the internal implantable pulse generator, or by the external programmable device and then send the recognition result to the internal implantable pulse generator, or the internal and external devices can complete it together, such as: acceleration information can be collected by the external device and then sent to the internal implantable pulse generator. Therefore, in some other embodiments, the implantable closed-loop stimulation system also includes an external programmable device.

[0061] An embodiment of the present invention provides a sleep state recognition device, including a processor, such as Figure 1 As shown, the processor is used to perform operations including the following steps:

[0062] S1, preliminarily determining whether the user is in a sleeping state based on user status information and / or clock indication, wherein the user status information includes at least one of acceleration, biological rhythm, electrocardiogram, and electroencephalogram; when it is preliminarily determined that the user is in a suspected sleeping state, the processor executes step S2;

[0063] S2, acquiring the local field potential LFP collected by the electrodes of the implantable closed-loop stimulation system and the acceleration collected by the triaxial accelerometer of the implantable closed-loop stimulation system;

[0064] S3, judging again whether the user is in a sleeping state according to the local field potential LFP and the acceleration; when it is judged again that the user is in a sleeping state, the processor executes step S4;

[0065] S4, determining the sleep stage based on the local field potential LFP.

[0066] like Figure 2 As shown, in step S1, the processor initially determines whether the user is in a sleeping state, specifically including:

[0067] S11a, determining whether the clock indication is within a preset time period. If the clock indication is within the preset time period, executing step S12a; if the clock indication is not within the preset time period, executing step S13a;

[0068] S12a, determining that the user is in a suspected sleeping state;

[0069] S13a, determining that the user is not in a sleeping state.

[0070] The clock indication of the current implantable pulse generator is judged, and the preset time period is the general time period of the user's sleep period, such as 21:00-7:00, and other times are considered to be in the daytime wakefulness period. The preset time period can be configured in the implantable pulse generator through the controller.

[0071] If the clock indication of the current implantable pulse generator is in the time period of 21:00-7:00, it is determined that the user is in a suspected sleeping state, and the processor executes step S2; if the clock indication of the current implantable pulse generator is not in the time period of 21:00-7:00, it is determined that the user is not in a sleeping state, that is, it is determined that the user is in the daytime wakefulness period, and the processor can output the result of not being in a sleeping state and being in the daytime wakefulness period.

[0072] and / or, if Figure 3 As shown, at every preset time, it is preliminarily determined whether the user is in a sleeping state based on the acceleration, wherein:

[0073] S11b, extracting time domain features of the three-axis acceleration, including the acceleration mean and acceleration standard deviation;

[0074] S12b, inputting the acceleration mean and the acceleration standard deviation into two pre-trained sleep state prediction models to perform sleep state probability prediction, and multiplying the two prediction results to obtain a sleep probability value;

[0075] S13b, preliminarily determining whether the user is in a sleeping state according to the sleep probability value, and determining whether the sleep probability value is greater than or equal to a preset probability value; when the sleep probability value is greater than or equal to the preset probability value, the processor executes step S14b; when the sleep probability value is less than the preset probability value, the processor executes step S15b;

[0076] S14b, determining that the user is in a suspected sleeping state;

[0077] S15b, determining that the user is not in a sleeping state.

[0078] The preset probability value can be pre-set in the implantable pulse generator by the controller. For example, the processor preliminarily determines whether the user is in a sleeping state based on the acceleration, and can make a sleep judgment every 30 minutes. Specifically, when the user is initially determined to be sleeping, a time window of 7 seconds can be set according to the three-axis acceleration data (X axis, Y axis and Z axis) of the three-axis accelerometer, and multiple three-axis accelerations are collected in each window. For the three-axis acceleration in each 7-second window, the mean and standard deviation of the three-axis acceleration are calculated respectively, and the mean of the three-axis acceleration in the three windows and the maximum standard deviation in the three windows are used as acceleration features, which are respectively input into two LDA models, and the posterior probability of the current sleeping state is output. The two posterior probabilities are multiplied to obtain the final sleep probability of acceleration predicting sleep. When the sleep probability is greater than or equal to 0.5, the processor determines that it is a suspected sleep state, and the processor executes step S2. When the sleep probability is less than 0.5, the processor determines that the user is not in a sleeping state. The two LDA model pre-training models mentioned above are two LDA models trained based on the mean and maximum standard deviation calculated from the historical acceleration data of the user when they are asleep and awake. After determining the sleep state, timing can also be performed, and sleep stages can be performed after a certain period of time.

[0079] In another embodiment, when the processor preliminarily determines whether the user is in a sleeping state in step S1, it may also be:

[0080] At every preset time interval, the user status information (physiological information) and the clock indication are used to determine whether the user is in a sleeping state, and the user status information (physiological information) includes acceleration, biological rhythm, electrocardiogram, and electroencephalogram. The steps are the same as the steps after determining that the user is not in a sleeping state only by using the clock indication, and will not be repeated here.

[0081] In this embodiment, the processor continuously analyzes the user's status information at certain intervals to accurately determine the user's status, so that the controller can accurately adjust the stimulation parameters.

[0082] like Figure 4As shown, in step S3, the processor determines again whether the user is in a sleeping state according to the local field potential LFP and the acceleration, including:

[0083] S31, determining whether there is a frequency division artifact in the local field potential LFP. Preferably, determining whether there is a frequency division artifact below 38 Hz in the local field potential LFP. When the processor determines that there is no frequency division artifact in the local field potential LFP, step S32 is executed;

[0084] The specific method for judging whether there is a frequency division artifact is as follows: the energy ratio of the oscillating signal at the 1 / 4 frequency division of the current local field potential LFP obtained by the processor through the electrode to the fractal signal at the same frequency band is used as the artifact judgment feature; the local field potential LFP without frequency division artifacts collected by daytime open-loop stimulation (stimulation output when the preset stimulation amplitude S0 is used during the daytime) is used, and the 95th percentile of the daytime data feature is used as the artifact judgment threshold. The daytime data feature refers to the local field potential LFP without frequency division artifacts during the daytime. When executing the judgment, the artifact judgment feature is compared with the artifact judgment threshold. If it exceeds the threshold, it is judged that there is a 1 / 4 frequency division artifact. If it does not exceed the threshold, it is judged that there is no 1 / 4 frequency division artifact.

[0085] In some other embodiments, the LFP collected according to the stimulation output when the preset stimulation amplitude S0 is used during the day can be transmitted to the controller, and the controller calculates the artifact judgment threshold and transmits it to the processor.

[0086] In some other embodiments of the present invention, the frequency division artifact judgment can also be carried out by: obtaining the energy of the oscillation signal of the current local field potential LFP at the 1 / 4 frequency division, calculating the energy average of all signals of the local field potential LFP collected by daytime open-loop stimulation (stimulation output when the preset stimulation amplitude S0 is used during the daytime) in a specific frequency band, expanding the energy average by 10 times, and then taking the median of the energy averages expanded by 10 times in all frequency bands, and comparing the energy of the oscillation signal of the current local field potential LFP at the 1 / 4 frequency division with the median. If it exceeds the median, it is determined that a 1 / 4 frequency division artifact exists; if it does not exceed the median, it is determined that no 1 / 4 frequency division artifact exists.

[0087] S32, extracting time domain features of the three-axis acceleration to obtain the acceleration mean and acceleration standard deviation;

[0088] S33, inputting the acceleration mean and the acceleration standard deviation into two pre-trained sleep state prediction models to perform sleep state probability prediction, and multiplying the two prediction results to obtain a sleep probability value;

[0089] S34, judging again whether the user is in a sleeping state according to the sleep probability value, and judging whether the sleep probability value is greater than or equal to a preset probability value; when the sleep probability value is greater than or equal to the preset probability value, the processor executes step S35; when the sleep probability value is less than the preset probability value, the processor executes step S36;

[0090] S35, determining that the user is in a sleeping state;

[0091] S36, determining that the user is not in a sleeping state.

[0092] The processor performs frequency division artifact detection on the local field potential LFP collected by the electrodes to ensure the accuracy of the local field potential LFP, and then determines the user's sleep stage.

[0093] like Figure 5 As shown, in step S4, the processor determines the sleep stage according to the local field potential LFP, specifically including:

[0094] S41, power spectrum estimation and denoising of the local field potential LFP.

[0095] Specifically, the current local field potential LFP is subjected to the fast Fourier transform method NFFT with the sampling frequency of the nearest 2 to the power of N data points, and then the Pwelch method power spectrum estimation with a window length of 1 second and a 50% overlap rate (50% overlap between adjacent signal segments) is performed to obtain the psd of 0-fs / 2Hz. Then the fractal signal psd of the current local field potential LFP is estimated using the IRASA method, and the 38Hz and above parts of the psd of 0-fs / 2Hz are replaced with the fractal signal psd part to achieve local field potential LFP denoising and make it more accurate.

[0096] Processing non-uniformly sampled signals with 2N-power data points allows the use of efficient algorithms of Fast Fourier Transform (FFT). In addition, 2N-power data points provide sufficient frequency resolution to analyze the signal in the frequency domain. "0-fs / 2Hz" means the frequency range is from zero to half the sampling frequency (fs / 2), because according to the sampling theorem, a discrete signal cannot contain frequency components exceeding half its sampling frequency.

[0097] S42, merge the denoised power spectrum into multiple frequency band energies and normalize them. The default frequency band range is 2-3Hz, 4-7Hz, 8-14Hz, 15-25Hz, 26-38Hz, 39-50Hz, and merge the denoised PSD into the energy of the above six frequency bands and normalize them.

[0098] S43, performing dimension reduction on the normalized multiple frequency band energies to obtain multiple dimension reduction features, where the multiple dimension reduction features include a first dimension reduction feature w1, a second dimension reduction feature w2, a third dimension reduction feature w3, and a fourth dimension reduction feature w4.

[0099] Specifically, the LDA model is used to perform feature dimensionality reduction on the local field potential LFP according to the normalized energy of multiple frequency bands, which is suitable for estimating different sleep stages and sleep depth states, and outputs the first dimensionality reduction feature w1, the second dimensionality reduction feature w2, the third dimensionality reduction feature w3, and the fourth dimensionality reduction feature w4.

[0100] S44, multiple dimension reduction features are input into the pre-built ARIMA model to obtain the predicted value group W ′ .

[0101] The ARIMA model is a prediction model with an input step of 10 and a prediction step of 10, which is constructed based on historical dimensionality reduction features. The model assumes that the information of the time series at the input step is correlated, while the information between the input steps is uncorrelated. The model segments the input dimensionality reduction features according to the input step (10) to obtain 10 subsequences. After obtaining the predicted value of a subsequence data, the predicted value is used as the input of the next subsequence to predict the next subsequence, and so on, until the entire sequence is predicted. In the prediction of each subsequence, the ARIMA model will make predictions based on the autoregressive term, the difference term, and the moving average term of the sequence. The autoregressive term reflects the past information of the sequence, the difference term reflects the non-stationarity of the sequence, and the moving average term reflects the noise component of the sequence.

[0102] S45, according to the predicted value group W ′ Determine sleep stages, which include wakefulness, rapid eye movement, and non-rapid eye movement.

[0103] Among them, the predicted value group W ′ Including the first dimension reduction feature prediction value w1 ′ , the second dimension reduction feature prediction value w2 ′ , the third dimension reduction feature prediction value w3 ′ And the fourth dimension reduction feature prediction value w4 ′ ; The prediction step size of the ARIMA model is 10, and the predicted values ​​of each dimension reduction feature obtained include 10 predicted values. For example, the first dimension reduction feature w1 is input into the ARIMA model, and it will predict 10 first dimension reduction feature predicted values ​​w1 ′ ;

[0104] Therefore, the processor groups W according to the predicted value ′ Determine sleep stages, including:

[0105] Calculate the mean of the predicted values ​​of each dimension reduction feature; calculate the mean of the 10 values ​​in each dimension reduction feature prediction value.

[0106] The first optimal classification threshold and the first dimension reduction feature prediction value w1 are preset ′ The mean values ​​were compared to determine the rapid eye movement or wakefulness period and the non-rapid eye movement period.

[0107] First, use the training set data to train the classification model, and then use the model to predict the samples in the training set. Next, calculate the performance indicators of the model, such as AUC, based on the prediction results and the actual sleep stage category labels of the samples. By trying different classification thresholds, different AUC values ​​can be obtained. Finally, the classification threshold corresponding to the maximum AUC is selected as the optimal classification threshold. Determine the feature mean and threshold traversal interval of the two category data of rapid eye movement or wakefulness (REM / Wake) and non-rapid eye movement (NREM) from the training set data, select the corresponding classification threshold based on the feature mean and threshold traversal interval, and take the maximum AUC as the goal to determine the preset first optimal classification threshold. Compare the preset first optimal classification threshold with the first dimensionality reduction feature prediction value w1 ′ If w1 ′ If it is less than the preset first optimal classification threshold, it is determined to be non-rapid eye movement (NREM), otherwise it is rapid eye movement or wakefulness (REM / Wake).

[0108] When it is determined to be a rapid eye movement period or a wakeful period, the second optimal classification threshold and the second dimension reduction feature prediction value w2 are preset ′ The mean values ​​were compared to distinguish between REM and wakefulness.

[0109] Since we can only determine the REM period or the wake period (REM / Wake), but cannot specifically distinguish between the REM period and the wake period, we need to distinguish again when we determine it to be the REM period or the wake period (REM / Wake). Specifically, we need to determine the feature mean and threshold traversal interval of the two categories of data, REM and Wake, from the training set data, select the corresponding classification threshold based on the feature mean and threshold traversal interval, and take the maximum AUC as the goal to determine the preset second best classification threshold. The preset second best classification threshold is compared with the second dimensionality reduction feature prediction value w2 ′ If w2 ′ If the threshold is less than the preset second optimal classification threshold, it is determined as the awake period (Wake), otherwise it is the rapid eye movement (REM). It should be noted that the preset optimal classification threshold can also be determined in other ways. The above is mainly for determining the sleep stages.

[0110] After the processor in the sleep state recognition device provided by the present invention recognizes the sleep state of the user, the controller in the implantable closed-loop stimulation system of the present invention will output a stimulation signal according to the sleep state. Specifically, as follows:

[0111] The first is that when the processor preliminarily determines whether the user is in a sleeping state according to the sleep probability value in step S12b, when the sleep probability is less than the preset probability value, it is determined that the user is not in a sleeping state, and the processor outputs the result that the user is not in a sleeping state. At this time, the controller in the implantable closed-loop stimulation system will determine whether the user is at night according to the clock indication. When it is determined that the user is at night, the controller uses the preset stimulation amplitude S0 to output the stimulation signal. The preset stimulation amplitude S0 is an open-loop stimulation parameter and is a fixed value. Because when the controller executes a certain stimulation mode, the processor needs to judge the human body state again according to the preset interval duration, so that the controller can adjust the stimulation mode in time after the state changes. Therefore, when the stimulation amplitude S0 is used to output the stimulation signal, the preset interval duration is used, and the processor executes S1 again to adapt to individual needs and state changes.

[0112] Second, when the processor determines in step S31 that a frequency division artifact exists in the local field potential LFP, the processor outputs a result that a frequency division artifact exists in the local field potential LFP. At this time, the controller in the implantable closed-loop stimulation system will use a preset stimulation amplitude S0 to output a stimulation signal.

[0113] The third method is to use the processor in step S45 to calculate the predicted value group W. ′ After determining the sleep stage, the controller outputs the stimulation signal with the following stimulation amplitude:

[0114] During the awake period, the controller outputs a stimulation signal using a preset stimulation amplitude S0;

[0115] During the non-rapid eye movement period, the corresponding stimulation amplitude S1 of the stimulation signal output by the controller is:

[0116] S1=-a3×(N1-N3)+N1

[0117] Among them, a3 is the sleep state coefficient of the non-rapid eye movement period, N1 is the upper limit of the stimulation of the non-rapid eye movement period, N3 is the lower limit of the stimulation of the non-rapid eye movement period, and N1 and N3 are pre-set stimulation parameters.

[0118] During the rapid eye movement period, the corresponding stimulation amplitude S2 of the stimulation signal output by the controller is:

[0119] S2=a4×(PT)+T

[0120] Among them, a4 is the sleep state coefficient of the rapid eye movement period, P is the upper limit of the rapid eye movement stimulation, T is the lower limit of the rapid eye movement stimulation, and P and T are pre-set stimulation parameters.

[0121] The present invention obtains two model parameters miuN1 and miuN3 according to the pre-training of the ARIMA model, wherein miuN1 and miuN3 are the characteristic means of the deep states of NREM stage 1 sleep and NREM stage 3 sleep, respectively. The training set data can be obtained by training with follow-up data, and the follow-up data includes physiological data, such as LFP, and miuN1 and miuN3 are obtained offline according to the training set data.

[0122] Among them, miuN1<miuN3;

[0123] If w3 ′ ≤miuN1, then the sleep state coefficient a3 of the non-rapid eye movement period is 0;

[0124] If w3 ′ ≥miuN3, the sleep state coefficient a3 of the non-rapid eye movement period is 1;

[0125] miuN1<w3 ′ <miuN3, then the sleep state coefficient of non-rapid eye movement period a3=(w3 ′ -miuN1) / (miuN3-miuN1).

[0126] The present invention obtains two model parameters miuPhasic and miuTonic according to the pre-training of the ARIMA model. The present invention obtains two model parameters miuPhasic and miuTonic according to the pre-training of the ARIMA model. MiuPhasic and miuTonic are the characteristic means of the deep states of phasic REM sleep and tonic REM sleep, respectively. The training set data can be obtained by training with follow-up data, and the follow-up data includes physiological data, such as LFP, and miuPhasic and miuTonic are obtained offline according to the training set data.

[0127] Among them, miuPhasic>miuTonic;

[0128] If w4 ′ ≥miuPhasic, the sleep state coefficient a4 of the REM period is 1;

[0129] If w4 ′ ≤miuTonic, then the sleep state coefficient a4 of the REM period is 0;

[0130] miuTonic<w4 ′ <miuPhasic, then the sleep state coefficient of the rapid eye movement period a4=(w4 ′-miuTonic) / (miuPhasic-miuTonic).

[0131] In this embodiment, the controller integrates the discrimination and proportional adjustment based on NREM and REM sleep depth states, which helps to achieve precise regulation of sleep, while reducing the total amount of cumulative stimulation and improving the therapeutic effect on the user.

[0132] In summary, the present invention provides an implantable closed-loop stimulation system suitable for sleep state, which can determine the corresponding stimulation mode for the user according to the sleep staging results, output the corresponding stimulation signal according to the stimulation mode, and improve the accuracy of the stimulation parameters. By accurately identifying the user's sleep state and stage according to the sleep state recognition device, the controller can intelligently adjust the stimulation parameters and output the stimulation signal to achieve automatic stimulation of the user's deep brain, realize individualized treatment, and improve the treatment effect.

[0133] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0134] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0135] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0136] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the computer or other programmable device. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0137] Obviously, the above embodiments are merely examples for the purpose of clear explanation, and are not intended to limit the implementation methods. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the implementation methods here. The obvious changes or modifications derived therefrom are still within the scope of protection of the invention.

Claims

1. A sleep state recognition device, characterized in that: Including processors; The processor is used to preliminarily determine whether the user is in a sleeping state according to user status information and / or clock indications, wherein the user status information includes at least one of acceleration, biorhythm, electrocardiogram, and electroencephalogram; when the user is in a suspected sleeping state, obtain the local field potential LFP collected by the electrodes of the implantable closed-loop stimulation system and the acceleration collected by the three-axis accelerometer of the implantable closed-loop stimulation system; and determine again whether the user is in a sleeping state according to the local field potential LFP and the acceleration; When the user is in a sleeping state, the sleep stage is determined according to the local field potential LFP.

2. The device according to claim 1, characterized in that Preliminary determination of whether the user is in a sleeping state, including: Determine whether the clock indication is within a preset time period; When the clock indication is within a preset time period, determining that the user is in a suspected sleeping state; When the clock indication is not within the preset time period, it is determined that the user is not in the sleeping state.

3. The device according to claim 1, characterized in that Preliminary determination of whether the user is in a sleeping state, including: At preset intervals, the acceleration is used to preliminarily determine whether the user is in a sleeping state, wherein: Extract time domain features of the three-axis acceleration, including the acceleration mean and acceleration standard deviation; Inputting the acceleration mean and the acceleration standard deviation into two pre-trained sleep state prediction models to perform sleep state probability prediction, and multiplying the two prediction results to obtain a sleep probability value; Preliminarily determine whether the user is in a sleeping state based on the sleep probability value; When the sleep probability value is greater than or equal to a preset probability value, it is determined that the user is in a suspected sleep state; When the sleep probability value is less than a preset probability value, it is determined that the user is not in a sleep state.

4. The device according to claim 1, characterized in that Determine again whether the user is in sleep state, including: Determining whether the local field potential LFP has a frequency division artifact; When the local field potential LFP does not have a frequency division artifact, time domain feature extraction is performed on the three-axis acceleration to obtain the acceleration mean and acceleration standard deviation; Inputting the acceleration mean and the acceleration standard deviation into two pre-trained sleep state prediction models to perform sleep state probability prediction, and multiplying the two prediction results to obtain a sleep probability value; determining again whether the user is in a sleeping state according to the sleeping probability value; When the sleep probability value is greater than or equal to a preset probability value, it is determined that the user is in a sleep state; When the sleep probability value is less than a preset probability value, it is determined that the user is not in a sleep state.

5. The device according to claim 1, characterized in that Determining the sleep stage according to the local field potential LFP includes: Performing power spectrum estimation and denoising on the local field potential LFP; The denoised power spectrum is combined into multiple frequency band energies and normalized; Performing dimension reduction on the normalized multiple frequency band energies to obtain multiple dimension reduction features, wherein the multiple dimension reduction features include a first dimension reduction feature w1, a second dimension reduction feature w2, a third dimension reduction feature w3, and a fourth dimension reduction feature w4; The multiple dimension reduction features are input into the pre-built ARIMA model to obtain the predicted value set W ′ ; According to the predicted value group W ′ Determine sleep stages, which include wakefulness, rapid eye movement, and non-rapid eye movement.

6. The device according to claim 5, characterized in that The predicted value set W ′ Including the first dimension reduction feature prediction value w1 ′ , the second dimension reduction feature prediction value w2 ′ , the third dimension reduction feature prediction value w3 ′ And the fourth dimension reduction feature prediction value w4 ′ , the first dimension reduction feature prediction value w1 ′ , the second dimension reduction feature prediction value w2 ′ , the third dimension reduction feature prediction value w3 ′ And the fourth dimension reduction feature prediction value w4 ′ Each contains multiple predicted values; According to the predicted value group W ′ Determine sleep stages, including: Calculate the mean of the predicted values ​​of each dimension reduction feature; The first optimal classification threshold and the first dimension reduction feature prediction value w1 are preset ′ The mean values ​​were compared to determine the rapid eye movement period or wakefulness period and the non-rapid eye movement period; When it is determined to be a rapid eye movement period or a wakeful period, the second optimal classification threshold and the second dimension reduction feature prediction value w2 are preset ′ The mean values ​​were compared to distinguish between REM and wakefulness.

7. An implantable closed-loop stimulation system suitable for use in a sleeping state, characterized in that: A sleep state recognition device comprising any one of claims 1 to 6, and a controller; The controller is used to proportionally adjust the output stimulation signal using the stimulation amplitude corresponding to the sleep stage, wherein different sleep stages correspond to different stimulation amplitudes.

8. The system according to claim 7, characterized in that When it is preliminarily determined that the user is not in a sleeping state, it is determined whether it is nighttime according to the clock indication. If it is nighttime, a stimulation signal is outputted using a preset stimulation amplitude S0; When it is determined again that the user is not in the sleeping state, a stimulation signal is outputted using the preset stimulation amplitude S0.

9. The system according to claim 7, characterized in that When it is determined that the local field potential LFP has a frequency division artifact, a stimulation signal is outputted using a preset stimulation amplitude S0.

10. The system according to claim 7, characterized in that For the awake period in the sleep stage, the stimulation signal is output using the preset stimulation amplitude S0; For the non-rapid eye movement period in the sleep stage, the corresponding stimulation amplitude S1 of the output stimulation signal is: S1=-a3×(N1-N3)+N1 Among them, a3 is the sleep state coefficient of the non-rapid eye movement period, N1 is the upper limit of the stimulation of the non-rapid eye movement period, and N3 is the lower limit of the stimulation of the non-rapid eye movement period; For the rapid eye movement stage in sleep, the corresponding stimulation amplitude S2 of the output stimulation signal is: S2=a4×(PT)+T Among them, a4 is the sleep state coefficient of the rapid eye movement period, P is the upper limit of the rapid eye movement stimulation, and T is the lower limit of the rapid eye movement stimulation.

Images