Adaptive deep brain stimulation for sleep stage targeting to treat sleep dysfunction

Abstract

Apparatus, systems, software, and methods are provided for treating sleep dysfunction in subjects using nocturnal deep brain stimulation (DBS). DBS is performed using a neural recording device that records electroencephalogram (EEG) signal data during the subject's sleep. Machine learning computational models are used to detect and classify patterns of neural activity associated with different sleep characteristics or sleep stages. An adaptive DBS algorithm is provided that uses intracranial classification of sleep characteristics or sleep stages to adjust stimulation parameters to target sleep dysfunction. Methods and systems for performing closed-loop therapy using a DBS device that records EEG signals from subcortical or cortical neural activity associated with selected sleep characteristics or sleep stages, and automatically adjusts DBS device settings and / or delivers deep EEG stimulation when a pre-defined pattern of neural activity associated with the selected sleep characteristic or sleep stage is detected.

Description

[0001] Cross-reference to related applications

[0002] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 472,191, filed June 9, 2023, and U.S. Provisional Patent Application No. 63 / 522,284, filed June 21, 2023, both of which are incorporated herein by reference in their entirety.

[0003] Statement regarding federally funded research or development

[0004] This invention was made with government support under grant number HR0011-20-0028 from the U.S. Defense Advanced Research Projects Agency (DARPA). The government has certain rights in this invention. Background Technology

[0005] Sleep dysfunction is a consistent feature across many neurological and psychiatric disorders (Bassetti, CL et al. Eur. J Neurol. 22, 1337-1354 (2015)). In people with Parkinson's disease (PD), nonmotor symptoms significantly reduce quality of life, with up to 90% reporting significant sleep dysfunction across both the rapid eye movement (REM) and non-REM (NREM) phases (Videnovic, A. and Golombek, D. Exp. Neurol. 243, 45-56 (2013), Barone, P. et al. Mov. Disord. 24, 1641-1649 (2009), Diederich et al. Sleep Med. 6, 313-318 (2005), Martinez-Martin et al. Mov. Disord. 26, 399-406 (2011)). In healthy individuals, NREM sleep is associated with increased cortical brain activity at low frequencies (0.5–4 Hz), known as slow waves, which are thought to have multiple functions related to metabolism, cognition, and synaptic homeostasis (Léger, D. et al. Sleep Med. Rev. 41, 113–132 (2018)). In PD, reduced slow waves are associated with faster disease progression (Schreiner, SJ et al. Annals of Neurology. Vol. 85, 765–770 (2019)). Conventional high-frequency deep brain stimulation (DBS) delivered to the subthalamic nucleus (STN) has been shown to partially improve sleep structure and NREM slow-wave activity (1–4 Hz) in PD (Baumann-Vogel, H. et al. Sleep 40 (5), (2017), Arnulf, I. et al. Neurology 55, 1732–1734 (2000), Monaca, C. et al. J. Neurol. 251, 214–218 (2004), Iranzo, A. et al. Neurol. Neurosurg. Psychiatry 72, 661–664 (2002)).However, the variability of DBS on overall sleep, the effects of DBS on the medial globus pallidus (GPi) at night, the mechanisms by which DBS improves NREM sleep, and why REM sleep disorders are refractory to DBS are not well understood (Zuzuárregui, JRP and Ostrem, JLJ Parkinsons. Dis. 10, 393-404 (2020), Tolleson et al. Neuromodulation 19, 724-730 (2016)).

[0006] There is still a need for improved methods of using DBS to treat sleep disorders, especially for subjects with neurological or psychiatric conditions. Summary of the Invention

[0007] Apparatus, systems, software, and methods are provided for treating sleep dysfunction in subjects using nocturnal deep brain stimulation (DBS). Specifically, DBS is performed using a neural recording device that records subcortical or cortical electroencephalogram (EEG) signals while the subject is sleeping. Machine learning computational models are used to detect and classify patterns of neural activity associated with different sleep stages or sleep characteristics. An adaptive DBS algorithm is provided that adjusts stimulation parameters using intracranial classification of sleep stages or sleep characteristics to target sleep dysfunction in selected sleep stages or sleep characteristics of interest. The methods and systems can be used to perform open-loop therapy to provide clinical guidance to clinicians or technicians for tuning DBS programming. Methods and systems are also provided for performing closed-loop therapy using a DBS stimulator that records EEG signals from subcortical or cortical neural activity associated with one or more sleep characteristics or sleep stages of interest, and automatically adjusts DBS settings and / or delivers electrical stimulation to the subject's brain when a pre-defined pattern of neural activity associated with a selected sleep characteristic or sleep stage is detected.

[0008] In one aspect, a method for treating sleep dysfunction in a subject is provided, the method comprising: placing a first electrode at a first location in a basal ganglia region or cortical region of the subject's brain to deliver electrical stimulation to the basal ganglia region or cortical region; placing a second electrode at a second location in a subcortical region or cortical region of the subject's brain to record electroencephalogram (EEG) signal data while the subject is sleeping; using the second electrode to detect EEG signals associated with a sleep feature or sleep stage of interest; and when EEG signals associated with a sleep feature or sleep stage of interest are detected using the second electrode, applying electrical stimulation to the basal ganglia region or cortical region of the subject's brain using the first electrode in a manner that effectively treats the subject's sleep dysfunction.

[0009] In some implementations, the EEG signal data includes field potential data.

[0010] In some implementations, the basal ganglia region is the subthalamic nucleus region, the globus pallidus region, or the thalamic region.

[0011] In some implementations, the cortical region is the precentral gyrus or the postcentral gyrus.

[0012] In some implementations, the sleep stages of interest are N2, N3, or REM.

[0013] In some implementations, sleep features of interest include slow waves, sleep spindles, K complexes, beta bursts, pre-awakening periods, awakening periods, post-awakening periods, or sleep stage transitions.

[0014] In some implementations, the method further includes using an accelerometer in conjunction with electroencephalogram (EEG) signals to identify sleep features or sleep stages of interest.

[0015] In some implementations, the method further includes using autonomic nervous data in combination with electroencephalogram (EEG) signals to identify sleep features or sleep stages of interest.

[0016] In some implementations, the method further includes using electroencephalography (EEG) or polysomnography (PSG) to identify sleep features or sleep stages of interest.

[0017] In some implementations, the method further uses a non-invasive sleep monitoring device, a wearable sleep monitoring device, a photoplethysmography (PPG)-based sleep monitoring device, or a radar-based sleep monitoring device to identify sleep features or sleep stages of interest.

[0018] In some implementations, the method further includes generating a sleep structure map.

[0019] In some implementations, the method further includes using a control algorithm to automate the application of electrical stimulation when an electroencephalogram (EEG) signal associated with a sleep feature or sleep stage of interest is detected.

[0020] In some implementations, the control algorithm uses machine learning algorithms to classify sleep characteristics and sleep stages. In some implementations, the machine learning algorithm is a supervised machine learning algorithm.

[0021] In some implementations, the control algorithm further adjusts one or more programmed stimulation parameters to maximize slow-wave activity. In some implementations, the slow-wave activity is in the frequency range of 0.5 Hz to 4 Hz. In some implementations, the stimulation amplitude is optimized during the N3 sleep stage to maximize slow-wave activity.

[0022] In some implementations, the control algorithm further uses linear discriminant analysis (LDA) or other embedded classifiers to adjust the current amplitude and / or frequency of the electrical stimulation.

[0023] In some implementations, electrical stimulation is applied unilaterally or bilaterally.

[0024] In some implementations, the EEG signal includes neural oscillations at beta, gamma, delta, or theta frequencies, or patient-defined spectral features alone.

[0025] In some implementations, the N3 sleep stage is identified by the increase in delta power during the N3 sleep stage.

[0026] In some implementations, the N2 sleep stage or the N3 sleep stage is identified by the attenuation of β power in the frequency range of 12 Hz to 30 Hz, the attenuation of γ power in the frequency range of 30 Hz to 60 Hz, the increase of low-frequency θ power in the frequency range of 5 Hz to 10 Hz, and / or the increase of δ power in the frequency range of 0.5 Hz to 4.5 Hz.

[0027] In some implementations, the second electrode is placed on the surface of the cortical sensorimotor region.

[0028] In some implementations, the first electrode is a non-brain-penetrating surface electrode array or a brain-penetrating electrode array.

[0029] In some implementations, the second electrode is a non-brain-penetrating surface electrode array or a brain-penetrating electrode array.

[0030] In some embodiments, the second electrode is an electroencephalogram (EEG) electrode array, a neurostimulator electrode installed subgaleal or through a burr hole in the skull, or a cranially mounted electrode, or an electrocorticography (ECoG) electrode array. In some embodiments, the ECoG electrode array spans the precentral and postcentral gyri or other areas of the cortex.

[0031] In some implementations, sleep dysfunction is caused by motor disorders or neurological conditions, in which the application of electrical stimulation improves sleep.

[0032] In some implementations, the movement disorder is Parkinson's disease.

[0033] In some implementations, sleep dysfunction is caused by stroke.

[0034] In some implementations, the subject is further administered daytime neural stimulation.

[0035] In some implementations, the subject is further administered a dopaminergic drug.

[0036] In some embodiments, the method further includes assessing the efficacy of treatment for the subject's sleep dysfunction. In some embodiments, assessing the efficacy of treatment for the subject's sleep dysfunction includes using the Visual Analogue Scale (VAS), Likert scale, Stanford Sleepiness Scale (SSS), Maintain Wakefulness Test (MWT), Epworth Sleepiness Scale (ESS), Multiple Sleep Latency Test (MSLT), or Athens Insomnia Scale. In some embodiments, assessing the efficacy of treatment for the subject's sleep dysfunction includes monitoring the subject using a motion recorder, electroencephalogram (EEG), or polysomnography.

[0037] In some implementations, the method further includes mapping the subject's brain to identify optimal locations in subcortical or cortical regions to detect electroencephalogram (EEG) signals associated with sleep characteristics or sleep stages.

[0038] In some implementations, the cortical region is the precentral gyrus or the postcentral gyrus.

[0039] In some embodiments, the method further includes segmenting the recorded EEG signal data into consecutive time epochs. In some embodiments, the method further includes assigning sleep characteristics or sleep stage labels to each time epoch. In some embodiments, each time epoch comprises 0.5 seconds to 1 minute of the recorded EEG signal data.

[0040] In some implementations, the method is performed while the subject is sleeping at home, in a sleep laboratory, or in a hospital.

[0041] In some implementations, the N2 sleep stage or the N3 sleep stage is identified by one or more spectral power changes selected from: a decrease in β power in the frequency range of 12 Hz to 30 Hz compared to the β power when the subject is awake; a decrease in γ power in the frequency range of 30 Hz to 60 Hz compared to the γ power when the subject is awake; an increase in θ power in the frequency range of 5 Hz to 10 Hz compared to the θ power when the subject is awake; and an increase in δ power in the frequency range of 0.5 Hz to 4.5 Hz compared to the δ power when the subject is awake.

[0042] In some implementations, the N2 sleep stage or the N3 sleep stage is identified by detecting one or more changes in spectral power in conjunction with one or more changes in cortico-subcortical spectral coherence, wherein the one or more changes in cortico-subcortical spectral coherence are selected from: an increase in δ-cortico-subcortical spectral coherence compared to δ-cortico-subcortical spectral coherence when the subject is awake; and a decrease in β-cortico-subcortical spectral coherence compared to β-cortico-subcortical spectral coherence when the subject is awake.

[0043] In some embodiments, the pre-wake period or wakefulness period is identified by one or more spectral power variations selected from: a decrease in cortical delta power in the frequency range of 1 Hz to 4 Hz compared to the average cortical delta power during deep non-rapid eye movement (NREM) sleep; an increase in cortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power during deep NREM sleep; an increase in subcortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power during deep NREM sleep; and an increase in subcortical beta power in the frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power during deep NREM sleep. In some embodiments, the increase in subcortical beta power precedes the decrease in cortical delta power.

[0044] In some embodiments, the post-awakening time period is identified by one or more spectral power changes selected from: a decrease in cortical delta power in the frequency range of 1 Hz to 4 Hz compared to the average cortical delta power in the pre-awakening time period; an increase in cortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power in the pre-awakening time period; an increase in subcortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power in the pre-awakening time period; and an increase in subcortical beta power in the frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power in the pre-awakening time period.

[0045] In some implementations, electrical stimulation increases cortical delta power, decreases cortical alpha power, decreases cortical beta power, and decreases cortical σ power.

[0046] In some implementations, electrical stimulation reduces cortical-subcortical σ-spectral coherence.

[0047] On the other hand, a computer-implemented method is provided for programming a deep brain stimulation (DBS) device to treat a subject's sleep dysfunction, the computer performing the following steps: a) receiving recorded electroencephalogram (EEG) signal data from a subcortical or cortical region of the subject's brain while the subject is sleeping; b) analyzing the recorded EEG signal data using a classification model that identifies patterns of electrical signals in the recorded EEG signal data associated with a sleep feature or sleep stage of interest; c) adjusting one or more programmed stimulation parameters based on the recorded EEG signal data according to an algorithmic control rule; and d) when a sleep feature or sleep stage of interest is detected, instructing the DBS device to apply electrical stimulation to a basal ganglia region or cortical region of the subject's brain to treat the subject's sleep dysfunction.

[0048] In some implementations, the EEG signal data includes field potential data.

[0049] In some implementations, machine learning algorithms are used to generate classification models.

[0050] In some implementations, the machine learning algorithm is a supervised machine learning algorithm.

[0051] In some implementations, the computer-implemented method further includes receiving accelerometer data from the subject while the subject is sleeping; and using a classification model to analyze the accelerometer data in combination with the recorded electroencephalogram (EEG) signal data to identify sleep characteristics or sleep stages.

[0052] In some implementations, the computer-implemented method further includes receiving autonomic nervous data of the subject while the subject is sleeping; and using a classification model to analyze the autonomic nervous data in combination with the recorded electroencephalogram (EEG) signal data to identify sleep characteristics or sleep stages.

[0053] In some implementations, the computer-implemented method further includes receiving an electroencephalogram (EEG) or polysomnography (PSG) of the subject while the subject is sleeping; and using a classification model to analyze the EEG or PSG to identify sleep characteristics or sleep stages.

[0054] In some implementations, the computer-implemented method further includes receiving data from a non-invasive sleep monitoring device, a wearable sleep monitoring device, a photoplethysmography (PPG)-based sleep monitoring device, or a radar-based sleep monitoring device; and using a classification model to analyze the data to identify sleep characteristics or sleep stages.

[0055] In some implementations, the computer-implemented method further includes generating a sleep structure map.

[0056] In some implementations, the sleep characteristics or sleep stage classification model is trained by analyzing electroencephalogram (EEG) signal data recorded over multiple nights while the subject is sleeping.

[0057] In some implementations, the computer-implemented method further includes: a) ranking the predicted stimulus efficacy of available settings of the DBS device based on a classifier score of stimulus efficacy for each setting using a linear classification model; b) selecting the stimulus setting predicted to have the highest stimulus efficacy based on the linear classification model; c) receiving recorded electroencephalogram (EEG) signal data from the subcortical or cortical regions of the subject's brain after applying electrical stimulation to the basal ganglia or cortical regions of the subject's brain using the DBS device with the setting predicted to have the highest stimulus efficacy; d) analyzing the recorded EEG signal data to assess the subject's neural response to the electrical stimulation; e) based on the subject's response to the electrical stimulation. The neural response to electrical stimulation is used to update the linear classification model to generate an updated linear classification model; f) the updated linear classification model is used to update the ranking of predicted stimulation efficacy of the available settings of the DBS device; g) the stimulation setting predicted to have the highest stimulation efficacy is selected based on the updated linear classification model; h) after applying electrical stimulation to the basal ganglia region or cortical region of the subject's brain using the DBS device based on the updated linear classification model and the setting predicted to have the highest stimulation efficacy, the recorded electroencephalogram (EEG) signal data is received from the subcortical or cortical regions of the subject's brain; and i) e)-h) are repeated to adjust the available settings of the DBS device to optimize stimulation efficacy.

[0058] In some implementations, the linear classification model uses linear discriminant analysis (LDA) to adjust the current amplitude and frequency of the electrical stimulation.

[0059] In some implementations, the stimulation amplitude is optimized during the N3 sleep stage to maximize slow-wave activity. In some implementations, the slow-wave activity is in the frequency range of 0.5 Hz to 4 Hz. In some implementations, different elements of sleep, such as, but not limited to, N1, N2, N3, phased REM and tonic REM, and REM-related physiology, including slow waves, sleep spindles, K-complexes, and beta bursts, are targeted.

[0060] In some implementations, the classification model identifies the N2 sleep stage or the N3 sleep stage by one or more spectral power variations selected from: a decrease in β power in the frequency range of 12 Hz to 30 Hz compared to the β power when the subject is awake; a decrease in γ power in the frequency range of 30 Hz to 60 Hz compared to the γ power when the subject is awake; an increase in θ power in the frequency range of 5 Hz to 10 Hz compared to the θ power when the subject is awake; and an increase in δ power in the frequency range of 0.5 Hz to 4.5 Hz compared to the δ power when the subject is awake.

[0061] In some implementations, the classification model identifies the N2 sleep stage or the N3 sleep stage by combining one or more spectral power variations with the detection of one or more variations in cortical-subcortical spectral coherence, wherein the one or more variations in cortical-subcortical spectral coherence are selected from: an increase in δ-cortical-subcortical spectral coherence compared to δ-cortical-subcortical spectral coherence when the subject is awake; and a decrease in β-cortical-subcortical spectral coherence compared to β-cortical-subcortical spectral coherence when the subject is awake.

[0062] In some embodiments, the classification model identifies pre-wake or wakefulness periods by one or more spectral power variations selected from: a decrease in cortical delta power in the frequency range of 1 Hz to 4 Hz compared to the average cortical delta power during deep non-rapid eye movement (NREM) sleep; an increase in cortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power during deep NREM sleep; an increase in subcortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power during deep NREM sleep; and an increase in subcortical beta power in the frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power during deep NREM sleep. In some embodiments, the increase in subcortical beta power precedes the decrease in cortical delta power.

[0063] In some implementations, the classification model identifies the post-awakening time period by one or more spectral power changes selected from: a decrease in cortical delta power in the frequency range of 1 Hz to 4 Hz compared to the average cortical delta power in the pre-awakening time period; an increase in cortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power in the pre-awakening time period; an increase in subcortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power in the pre-awakening time period; and an increase in subcortical beta power in the frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power in the pre-awakening time period.

[0064] In some embodiments, the computer-implemented method further includes segmenting the recorded EEG signal data into consecutive time periods. In some embodiments, the computer-implemented method further includes assigning sleep characteristics or sleep stage tags to each time period. In some embodiments, each time period includes 0.5 seconds to 1 minute of the recorded EEG signal data.

[0065] In some embodiments, the computer-implemented method further includes training a linear model by analyzing the recorded EEG signal data using a nonlinear model during all sleep stages while the subject is sleeping, to classify each time period as an N3 sleep stage period or a non-N3 sleep stage period. In some embodiments, canonical delta and beta power bands are used as feature inputs to train a linear classification model using linear discriminant analysis to classify each time period as an N3 sleep stage period or a non-N3 sleep stage period. In some embodiments, subcortical field potentials are used as feature inputs to train a linear classification model using linear discriminant analysis to classify each time period as an N3 sleep stage period or a non-N3 sleep stage period.

[0066] In some implementations, the EEG signal data includes field potential data.

[0067] In some implementations, the computer-implemented method further includes storing a user profile of the subject, which includes information about recorded electroencephalogram (EEG) signal data associated with sleep characteristics or stages.

[0068] In some embodiments, the computer-implemented method further includes storing a user profile of the subject, which includes information about programmed stimulation parameters for applying electrical stimulation to the basal ganglia or cortical regions of the subject's brain based on recorded electroencephalogram (EEG) signal data to treat the subject's sleep dysfunction.

[0069] In another aspect, a non-transitory computer-readable medium is provided, comprising program instructions that, when executed by a processor in a computer, cause the processor to perform the computer-implemented methods described herein.

[0070] On the other hand, a kit is provided that includes a non-transitory computer-readable medium and instructions for use of a deep brain stimulation device to treat a subject’s sleep dysfunction.

[0071] On the other hand, a system for treating sleep dysfunction in a subject is provided, the system comprising: a first electrode adapted to be placed at a location in a basal ganglia region or cortical region of the subject's brain to deliver electrical stimulation to the basal ganglia region or cortical region; a second electrode adapted to be placed at a subcortical region or cortical region of the subject's brain to record electroencephalogram (EEG) signal data while the subject is sleeping; and a processor programmed according to a computer-implemented method described herein to instruct the first electrode to apply electrical stimulation to the basal ganglia region or cortical region of the subject's brain in a manner that effectively treats the subject's sleep dysfunction when an EEG signal associated with a sleep feature or sleep stage of interest is detected using the second electrode.

[0072] In some implementations, the system further includes an accelerometer for recording the subject's movements while the subject is sleeping.

[0073] In some implementations, the system further includes a non-invasive sleep monitoring device, a wearable sleep monitoring device, a photoplethysmography (PPG)-based sleep monitoring device, or a radar-based sleep monitoring device.

[0074] In some implementations, the first electrode is a non-brain-penetrating surface electrode array or a brain-penetrating electrode array.

[0075] In some embodiments, the second electrode is a non-brain-penetrating surface electrode array or a brain-penetrating electrode array. In some embodiments, the second electrode is an electroencephalogram (EEG) electrode array, a neurostimulator electrode installed subgastrically or through a burr hole in the skull, or a craniotomy electrode array, or an electrocorticography (ECoG) electrode array. In some embodiments, the ECoG electrode array spans the precentral and postcentral gyri or other areas of the cortex.

[0076] In some implementations, the sleep dysfunction is caused by a motor disorder or neurological condition, in which electrical stimulation is applied to improve sleep. In some implementations, the motor disorder is Parkinson's disease.

[0077] In some embodiments, the system further includes a user interface comprising an input electrically coupled to the processor for instructing the first electrode to apply electrical stimulation to the basal ganglia region or cortical region to treat the subject's sleep dysfunction. In some embodiments, the user interface is password protected and operable by a healthcare practitioner. Attached Figure Description

[0078] Figures 1A-1E Part. (abbreviated as Part.) recording setup and intracranial cortical field potentials: ( Figure 1A Clinical information of participants. UPDRS - Unified Parkinson's Disease Rating Scale. Figure 1B A schematic diagram of the RC+S system. The illustration provides close-ups of the cortical and subcortical leads. Adapted from Gilron et al. 2021 (Nat. Biotechnol. 39, 1078–1085). Figure 1C The average time spent in each sleep stage per night, categorized from Dreem2 headband polysomnography data (TST = total sleep time). Error bars indicate the standard deviation for each night. Figure 1D With stimulation on, representative traces of field potential time series from the left device of Participant 1 across all sleep stages: left column - precentral gyrus; right column - subthalamic nucleus. Columns share a color legend and scale bar. Figure 1E Power spectral density maps of intracranial field potentials (FPs) divided by sleep stages during routine continuous stimulation. Left quadrant – precentral gyrus; right quadrant – subcortical region (STN for participant 1; GPs (Globus Pallidus) for participant 2). Shaded error bars indicate the standard error across nights; with… Figure 1D Shared color legend.

[0079] Figures 2A-2G Performance of the adaptive DBS classifier during sleep phase: ( Figure 2A The classifier performance of the participants (abbreviated as Part.) of the embedded N3 classifier during the validation (cDBS: no adaptive stimulus variation) and testing (adaptive DBS-aDBS: stimulus variation, participant 1 only) nighttime phases. Error bars indicate the standard deviation. Figure 2B During the validation and test nights, the classifier outputs of Participant 1 and Participant 2 were composed of proportions of the baseline real sleep stages. Utilizing... Figure 1D , Figure 1EThe same color legend is used. The left graph depicts the sleep stage composition determined by the Dreem headband across all nights and participants, based on the "N3" embedded classifier output. Solid and dashed lines correspond to the left and right devices, respectively. For example, 35%–40% of the embedded N3 predictions in the left hemisphere device occur during N2 sleep. The right column depicts the corresponding composition of the embedded "non-N3" classification. Figure 2C A stacked histogram depicting the number of 30-second sleep segments with corresponding δ and β powers is shown, color-coded by sleep stage (and...). Figure 2B (Shared color legend). The dashed lines represent the cumulative density functions of embedded left and right N3 classifications as a function of band power, illustrating the proportion of N3 predictions occurring when the band power during sleep is less than or equal to the x-axis position. Participant 1 - left column; Participant 2 - right column. This indicates that classification sensitivity improves for progressively deeper N3 sleep. Figure 2D Sleep measurements of Participant 1 during nights of cDBS (validation) and aDBS (test). Error bars indicate standard deviation. Specifically, the mean N3 in CDB was 42 min, while the mean N3 in aDBS was 41 min. Figure 2E (Top) For Participant 1's adaptive DBS test night, the Dreem2 headband sleep structure map is superimposed on the spectrogram of the precentral gyrus cortex (FP). (Bottom) Stimulus amplitude as a function of time shares the same x-axis as the sleep structure map. During embedded classification in N3, the stimulation amplitude decreased (50%) (a 16.7-minute span depicts the transition to embedded classification in N3 sleep). Figure 2F ) Figure 2D A magnified depiction of the highlighted portion. The black line indicates the baseline real sleep stage. Below are the traces of the raw delta power (light blue) and raw beta power (dark blue) corresponding to the highlighted portion, as calculated by the embedded INS device. Below is the corresponding linear discriminant embedded classifier output (blue) compared to a user-defined threshold (dashed line). The gray line indicates the obtained stimulus amplitude. Figure 2F All subplots share the same x-axis. Figure 2G As described in Materials and Methods Section C, box plots of the band power calculated by the apparatus for all N3 time periods (cDBS → left: n=169; right: n=168 || aDBS → left: n=162; right: n=134) across the night for participant 1's CDBS and aDBS. An asterisk indicates the significance of the independent samples t-test (δ → left: t=-3.5, p < 1e-3; right: t=-5.8, p << 1e-3 || β → left: t=2.7, p < 1e-2; right: t=0.2, p=0.8).

[0080] Figure 3Sleep Adaptive DBS Process.

[0081] Figures 4A-4G Methods, data collection and analysis procedures: ( Figure 4A A schematic diagram of the RC+S system setup used to record intracranial cortical field potentials (FP) in participants (adapted from Gilron et al. (2021) Nat. Biotechnol. 39(9):1078-1085). Figure 4B Illustration of the placement of the RC+S sensing depth electrodes at the subcortical STN and GPi (right) and the cortical ECoG location (left). Exemplary data from PD2 and PD3 participants. Figure 4C A schematic diagram of the Dreem2 portable headband used for recording nighttime polysomnography at home (adapted from Debellemaniere et al. (2018) Front. Hum. Neurosci. 12:88). Figure 4D A diagram illustrating single-night sleep patterns in PD patients (DBS enabled), where the sleep structure diagram (purple) shows sleep stages (AW: wakefulness; RM: REM; [N1, N2, N3]: NREM), and spectrograms from the cortex (top 2 plots) and subcortex (bottom 2 plots) of both hemispheres show multi-frequency variations across sleep stages, with the x-axis representing time in hours and the y-axis representing frequency (Hz). FP was recorded bilaterally from both cortical and subcortical regions. Figure 4E The flowchart illustrates the data analysis and preprocessing procedures for a 10-day sleep dataset (n=5) and an on / off dataset (n=4). Figure 4F Representative traces of RC+S FP time series from all sleep stages in the cortex (left column) and subcortex (right column; subthalamic nucleus). Columns share a scale bar, and rows share a color legend (awake, REM, N1, N2, and N3). Data were obtained from a PD participant with left-hemispheric activation. Figure 4G Comparison of intracranial spectral power (FP) during cortical (left) and subcortical (right) sleep stages in a single subject with DBS enabled. Shaded error bars indicate standard error. Figure 4F Shared color legend.

[0082] Figures 5A-5G Spectral variations in NREM. Dynamic changes in power spectrum and functional connectivity between cortical and subcortical regions during NREM sleep: ( Figure 5AThe power spectral variation (mean ± SEM) during NREM (N2 and N3) sleep with the waking phase as the baseline was observed in all PD participants (n=4) during onset stimulation in the cortical (top) and subcortical (bottom) regions, within the low-frequency range (1–50 Hz). The y-axis shows the power spectral difference in decibels (dB) between NREM and the waking phase. The thick line represents the mean, and the shaded area represents the standard error (SEM). Figure 5B In both the cortical (top) and subcortical (bottom) regions, power increased in the δ (1–4 Hz) band and decreased in the β (13–31 Hz) band during NREM sleep compared to wakefulness during onset stimulation. Each bar shows the average difference in spectral power across multiple nights for a single participant, and each data point shows the average difference in spectral power across a single night, with data pooled from both hemispheres. Figure 5C In both the cortical (top) and subcortical (bottom) regions, during the stimulation-off condition, δ power increased while β power decreased in NREM compared to the awake phase in PD participants (n=4). The thick line shows the mean, and the shaded area represents the standard error. Figure 5D Differences in cortical spectral power between on and off stimulation conditions in four participants with PD during NREM sleep stages (top), showing increased δ (1–4 Hz) and decreased low α and low β activity (8–15 Hz) with on stimulation. Each colored line shows the spectral variation for one participant, with the thick line showing the mean across participants, and the shaded area representing the SEM. No statistically significant differences in spectral power in the subcortical regions were observed (bottom). The x-axis is frequency (Hz), and the y-axis is the power difference (on-off). Figure 5E During stimulation, changes in cortical-subcortical spectral coherence (mean ± SEM) during NREM (N2 and N3) sleep were measured for all participants (n=5), with the waking phase as the baseline. The y-axis illustrates the difference in spectral coherence between NREM and the waking phase. The horizontal baseline at 0 represents the waking phase baseline. Figure 5F The total difference in spectral coherence of δ (1–4 Hz, left) and β (13–31 Hz, right) during NREM sleep compared to wakefulness during stimulation-on period. Each bar shows the average difference in spectral coherence across multiple nights for one participant, and each point shows the average difference in spectral coherence across one night, where data are pooled from both hemispheres. Figure 5G During the stimulus-off condition, compared with the waking phase in PD participants (n=4), δ coherence increased while β coherence decreased in NREM. For all plots, data from both hemispheres were pooled.

[0083] Figures 6A-6E The inverse relationship between subcortical β and cortical δ activity. Figure 6A An example of single-night subcortical β (purple) and cortical δ (green) power from a PD participant (PD3) during stimulus onset depicts the inverse relationship in the time domain. δ and β power are smoothed using a 20-point Gaussian kernel. Figure 6B The mean Spearman rho correlation between subcortical β power and cortical δ power across multiple nights for all 4 PD participants in both on (left) and off (right) stimulation. Each bar shows the mean correlation for one participant, and each point shows the correlation across one night with data pooled from both hemispheres. Figure 6C Scatter plots depicting the correlation between subcortical FPβ (13–31 Hz) power and cortical FPδ (1–4 Hz) power during NREM sleep in four PD participants during stimulation onset; STN (brown and red) and GPi (blue and light blue). Each point represents data from a 5-s NREM sleep period. Each plot is based on data from one night's data pooled from both hemispheres of one participant. Figure 6D The normalized cross-correlation between subcortical β power and cortical δ power shows that, during NREM with on-stimulus, subcortical β activity precedes cortical δ activity in PD participants. The bar plot (left) shows the lag of subcortical β with cortical δ reference. Each bar shows the mean lag for one participant, and each point shows the lag across one night, where data are pooled from both hemispheres. An example of the cross-correlation shows the lag of subcortical β as a function of time from PD2 over one night during on-stimulus (right). The vertical dashed line indicates zero lag. Figure 6E The interaction between cortical delta (δ) and cortical beta (β) activity was examined as a control for cortical delta versus subcortical β. The bar plot (left) shows the mean Spearman's rho correlation between cortical delta and β power across multiple nights for all four PD participants under on-stimulation. Each bar shows the mean correlation for one participant, and each point shows the correlation across one night with data pooled from both hemispheres. The scatter plot shows cortical delta and β power in the four PD participants for two representative PD participants during on-stimulation. Each point represents data from a 5-s NREM sleep period.

[0084] Figures 7A-7D Changes in spectral power prior to spontaneous arousal. Subcortical β increases and cortical δ decreases prior to spontaneous arousal. Figure 7ACortical delta (1–4 Hz) power during the NREM-to-wake transition episode (left) for all PD participants (n=4; mean ± SEM) during the on-stimulation period. Each data point is the mean for a 5-second data period, and shading represents the SEM of a participant's NREM-to-wake transition across the recording night. Data were pooled from both hemispheres. The vertical purple dashed line shows the wake time. The x-axis (on the left) shows the time in seconds since the onset of NREM sleep and the time since wakefulness (in the middle, around the vertical dashed line). The top black line shows the normalized values ​​(norm) (mean ± SEM) of the RC+S accelerometer data for all NREM-to-wake transitions across all nights for all participants, highlighting the wake time of the episode. The bar graph shows the change in cortical delta power during the pre-awakening (5 s before the awakening event, top) and post-awakening (15 s after the awakening event, bottom) periods, compared to the mean delta power in deep NREM (mean of NREM data 40 s after the start of NREM and 40 s before wakefulness; SWS). Each bar shows the mean power change for one participant, and each point shows the power change across all NREMs to wakefulness transitions in one night, with data pooled from both hemispheres. Cortical delta power gradually increases with deepening sleep and decreases steadily before wakefulness. The mean delta power after wakefulness (15 s) and before wakefulness (-5 s) is lower than the delta power during SWS. The mean cortical delta power after wakefulness (15 s) is lower than the delta power before wakefulness (-5 s). Figure 7B Similar to A, no significant trend was observed across participants or recording sites for subcortical delta power. Figure 7C Similar to A, for cortical β power, no significant trend was observed across participants or recording sites regarding pre- and post-awake states. Figure 7D Similar to A, except for subcortical β power, it illustrates the spontaneous increase in β power before arousal. The average β power in the STN after arousal (15 s) and before arousal (-5 s) is higher than that during SWS.

[0085] Figures 8A-8C Arousal prediction using spectral power variations. Subject-specific machine learning models utilizing cortical and subcortical spectral variations can predict arousal. Figure 8AAwakeness predictions were made before and after spontaneous awakeness using subject-specific QDA models. The vertical black dashed line represents the awakeness time. The x-axis shows the time in seconds since the start of awakeness, and the y-axis shows the awakeness predictions made using individual QDA models based on the posterior probability (mean ± SEM) to the awake event for each subject (n=5). The horizontal green line (y=0.5) represents the classification threshold. Figure 8B ) Receiver operating characteristic (ROC) performance for each subject in binary classification between deep NREM and pre-awake (-5 s) NREM data (blue) and deep NREM and post-awake (+15 s) data (green). Figure 8C Box plots of the distribution of the sobriety predictions by the QDA model for each participant in deep NREM (magenta), pre-sobriety (-5 s) NREM (blue), and post-sobriety (+15 s) data (green). In all cases, the Wilcoxon rank-sum test p-values ​​were < 0.001***, < 0.01**, and < 0.05*.

[0086] Figures 9A-9D Sleep statistics during stimulation on. Sleep statistics (mean ± SEM) during nighttime recordings for all participants (n=5) during stimulation on. Figure 9A The time when sleep begins, Figure 9B The total duration and frequency of wakefulness (awakeness during sleep) after the onset of sleep throughout the night. Figure 9C The duration of all sleep stages, in minutes. Figure 9D The total proportion of sleep stages.

[0087] Figures 10A-10D Sleep statistics with stimulation on and off. Sleep statistics during nighttime recordings for all PD participants (n=4) during a continuous night of on-stimulus conditions and a night of off-stimulus conditions. Figure 10A The time when sleep begins, Figure 10B The total duration and frequency of wakefulness (awakeness during sleep) after the onset of sleep throughout the night. Figure 10C The duration of all sleep stages, in minutes. Figure 10D The total proportion of sleep stages. The x-axis represents the stimulus condition (on / off) and the gray dashed line shows the average across all PD participants.

[0088] Figures 11A-11D Data processing procedure. Removing ECG artifacts from RC+S field potential data ( Figure 11A ) and motion-related spike artifacts ( Figure 11B ). Figure 11CUsing accelerometer data as a reference, the polysomnography from DREEM2 was time-synchronized with the intracranial data stream from the RC+S device. Figure 11D ) Use accelerometer data and correct wake time by detecting motion events.

[0089] Figures 12A-12E Cortical and subcortical spectral power (mean ± SEM) for all sleep stages in each participant (n=5), including those with dystonia ( Figure 12A ), PD3 ( Figure 12B ), PD9 ( Figure 12C ), PD2 ( Figure 12D ) and PD7 ( Figure 12E Patients with [unspecified condition]. Data from activated stimulation. Data averaged over 10 nights and bilateral hemispheres. Detailed Implementation

[0090] Apparatus, systems, software, and methods are provided for treating sleep dysfunction in subjects using nocturnal deep brain stimulation (DBS). DBS is performed while the subject is asleep using a neural recording device that records subcortical or cortical electroencephalogram (EEG) signals. Machine learning computational models are used to detect patterns and classifications of neural activity associated with different sleep characteristics and sleep stages. An adaptive DBS algorithm is provided that adjusts stimulation parameters using intracranial classifications of sleep characteristics and sleep stages to target sleep dysfunction in selected sleep stages of interest. Methods and systems are also provided for performing closed-loop therapy using a DBS stimulator that records EEG signals from subcortical or cortical neural activity associated with selected sleep characteristics or stages of interest, and automatically adjusts DBS settings and / or delivers DBS stimulation when a pre-defined pattern of neural activity associated with the selected sleep characteristic or sleep stage is detected.

[0091] Before describing the apparatus, system, software, and method of the present invention, it should be understood that the invention is not limited to the specific apparatus, system, software, and method described, and therefore variations are naturally possible. It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting, as the scope of the invention will be defined only by the appended claims.

[0092] Where a range of values ​​is provided, it should be understood that, unless the context explicitly dictates otherwise, each intermediate value between the upper and lower limits of the range, up to one-tenth of the lower limit unit, is also specifically disclosed. Every smaller range between any stated value or intermediate value within the stated range and any other stated value or intermediate value within the stated range is covered within this invention. The upper and lower limits of these smaller ranges may be independently included within or excluded from the range, and each range in which any limit, no limit, or both limits are included is also covered within this invention, but is subject to any explicitly excluded limit in the stated range. Where the stated range includes one or both of the limits, ranges excluding any or both of those included limits are also included in this invention.

[0093] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. While any methods and materials similar to or equivalent to those described herein may be used in the practice or testing of this invention, some potential and preferred methods and materials are described hereafter. All publications mentioned herein are incorporated by reference to disclose and describe methods and / or materials associated with the cited publications. It should be understood that, to the extent contradictory, this disclosure supersedes any disclosure in the incorporated publications.

[0094] As will be readily apparent to those skilled in the art upon reading this disclosure, each individual embodiment described and illustrated herein has discrete components and features that can be readily separated from or combined with features of any of the other several embodiments without departing from the scope or spirit of the invention. Any of the described methods may be performed in the order of the described events or in any other logically possible order.

[0095] It should be noted that, as used herein and in the appended claims, the singular forms “a,” “an,” and “the” include plural indicators unless the context clearly specifies otherwise. Thus, for example, reference to “an electrode” or “the electrode” includes a plurality of such electrodes, and reference to “an electrical signal” or “the electrical signal” includes reference to one or more electrical signals, and so on.

[0096] It should also be noted that claims can be drafted to exclude any element that may be optional. Therefore, this statement is intended to serve as a precondition for using exclusive terms such as "uniquely" or "only" in relation to the description of a claim element, or for using a "negative" limitation.

[0097] The publications discussed herein are provided solely for their prior disclosure prior to the filing date of this application. Nothing herein should be construed as an admission that the invention is not entitled to any prior invention prior to such publications. Furthermore, the publication dates provided may differ from the actual publication dates, which may require independent verification.

[0098] Define

[0099] The term “about,” especially when referring to a given quantity, means covering a deviation of plus or minus 5%.

[0100] The term "movement disorder" refers to any type of neurological condition that results in increased, decreased, or slowed movement. Movement disorders include, but are not limited to, Parkinson's disease, Parkinsonian syndrome, progressive supranuclear palsy, ataxia, cervical dystonia, chorea, dystonia, functional movement disorders, Huntington's disease, multiple system atrophy, myoclonus, tardive dyskinesia, Tourette syndrome, tremor, restless legs syndrome, and Wilson's disease. Symptoms may include, but are not limited to, tremor, involuntary movements, bradykinesia (slowness of movement), rigidity, postural instability, twisting movements, poor balance, irregular movement, tripping, and difficulty walking. In some cases, movement disorders are caused by genetic and / or environmental factors, head trauma, infection, inflammation, metabolic disorders, toxins, adverse drug reactions, or stressful life events.

[0101] The term "sleep dysfunction" is used herein to refer to sleep-wake disorders that result in insomnia, sleep disorder, or poor sleep quality. Sleep dysfunction can include difficulty falling asleep or maintaining sleep, frequent nighttime awakenings, early morning awakenings, sleep fragmentation, reduced total sleep time, reduced deep sleep time, reduced non-REM or REM sleep time, and / or inability to achieve deep sleep (N3 stage or delta sleep). Furthermore, sleep dysfunction may be associated with nocturnal motor symptoms, pain, nocturia, sleep-disordered breathing, periodic limb dyskinesia (PLMD), sleep paradoxes, sleep apnea, REM sleep behavior disorder (RBD), circadian rhythm dysfunction, and / or excessive daytime sleepiness. Tremors, rigidity, and motor disturbances associated with motor disorders such as those in Parkinson's disease may occur during nighttime awakenings and contribute to sleep dysfunction by prolonging wakefulness and preventing recovery from sleep.

[0102] In some implementations, sleep dysfunction is associated with stroke. Insomnia may occur after a stroke, particularly in patients with a right hemisphere stroke or a stroke in the thalamus or brainstem, which includes the pontine tegmentum and thalamus-midbrain region. Somnolence may occur after a stroke in patients with subcortical (caudate nucleus, putamen), superior pons, medial pontomedometral, or cortical strokes affecting the reticular activating system (RAS). Paramedian or bilateral thalamic strokes may initially cause coma, followed by somnolence upon awakening. Supratentorial strokes may reduce non-REM sleep, total sleep time, and ipsilateral or bilateral sleep spindles. Sawtooth waves may be reduced after a hemispheric stroke. REM sleep may be reduced after an occipital lobe stroke. Strokes in the pontomedometral junction and raphe nuclei may reduce the amount of non-REM sleep. Strokes in the inferior pons may selectively reduce REM sleep. Strokes in the paramedian thalamus and lower pons may reduce slow-wave sleep.

[0103] The terms “individual,” “subject,” “recipient,” and “patient” are used interchangeably herein and refer to any mammalian subject, especially a human, who requires diagnosis, treatment, or therapy. For therapeutic purposes, “mammal” means any animal classified as a mammal, including both human and non-human mammals such as non-human primates, including chimpanzees and other ape and monkey species; laboratory animals such as mice, rats, rabbits, hamsters, guinea pigs, and chinchillas; domesticated animals such as dogs and cats; and farm animals such as sheep, goats, pigs, horses, and cattle.

[0104] As used herein, the term "user" means a person who interacts with the apparatus and / or system disclosed herein to perform one or more steps of the methods disclosed herein. A user may be a patient being diagnosed with or receiving treatment for a sleep disorder. A user may be a healthcare professional, such as the patient's physician.

[0105] "Treatment" or "treating" means achieving at least the improvement of one or more symptoms associated with a symptom affecting a subject, resulting in a desired or beneficial clinical outcome for the patient, where improvement is defined as at least a reduction in the magnitude of a parameter, such as a symptom, associated with the treated symptom. Therefore, treatment encompasses a wide range of situations, from alleviating the intensity, duration, or extent of damage caused by and / or associated with the symptom, to and including the complete elimination of the symptom and any associated symptoms. Thus, treatment includes situations where the symptom or at least associated symptoms are completely suppressed, for example, prevented from occurring, or stopped, for example, terminated, so that the subject no longer suffers from the symptom, or at least no longer suffers from the symptoms characterizing the symptom. Treatment also includes situations where the progression of the symptom or at least the progression of associated symptoms is slowed, delayed, or stopped. In such cases, the subject may still have residual symptoms associated with the symptom, but any increase in the severity or magnitude of the symptoms is slowed, delayed, or prevented.

[0106] When used in the context of sleep dysfunction, the term "symptoms" may include, but is not limited to, problems with poor sleep quality, sleep duration, and / or amount of sleep experienced by the subject. Symptoms of sleep dysfunction may include insomnia, difficulty falling asleep, difficulty maintaining sleep, reduced total sleep time, reduced deep sleep time, reduced non-REM or REM sleep time, and / or inability to achieve deep sleep (N3 stage or delta sleep).

[0107] method

[0108] This disclosure provides a method for treating sleep dysfunction in subjects using nocturnal deep brain stimulation (DBS). Specifically, DBS is performed while the subject is sleeping using a neural recording device that records subcortical or cortical electroencephalogram (EEG) signal data. A machine learning computational model is used to detect and classify patterns of neural activity associated with different sleep characteristics and / or sleep stages. An adaptive DBS algorithm is provided that adjusts stimulation parameters using intracranial classified sleep characteristics and / or sleep stages to target sleep dysfunction in selected sleep stages of interest. This method and system can be used to perform open-loop therapy to provide clinical guidance to clinicians or technicians for tuning DBS programming. A method and system are also provided for performing closed-loop therapy using a DBS stimulator that records EEG signals from subcortical or cortical neural activity associated with one or more sleep characteristics and / or sleep stages of interest, and automatically adjusts DBS settings and / or delivers electrical stimulation to the basal ganglia or cortical regions of the subject's brain when a pre-defined pattern of neural activity associated with a selected sleep characteristic or sleep stage is detected. The various steps and aspects of this method will be described in more detail below.

[0109] In some implementations, the subject-specific approach is used to treat sleep dysfunction associated with movement disorders. Movement disorders include, but are not limited to, Parkinson's disease, Parkinson's syndrome, progressive supranuclear palsy, ataxia, cervical dystonia, chorea, dystonia, functional movement disorders, Huntington's disease, multiple system atrophy, myoclonus, tardive dyskinesia, Tourette syndrome, tremor, restless legs syndrome, and Wilson's disease.

[0110] In some implementations, the subject approach is used to treat sleep disorders associated with neurological conditions. Neurological conditions include, but are not limited to, neurodegenerative diseases, including Alzheimer's disease, Parkinson's disease, Huntington's disease, multiple system atrophy, or Lewy body dementia; and multiple system atrophy, epilepsy, stroke, bipolar disorder; neuromuscular disorders, including amyotrophic lateral sclerosis (ALS), Charcot-Marie-Tooth disease (CMT), chronic inflammatory demyelinating polyneuropathy (CIDP), Guillain-Barré syndrome (GBS), Lambert-Eaton syndrome, muscular dystrophy, myasthenia gravis, myopathy, and peripheral neuropathy.

[0111] In some implementations, the thematic approach is used to treat stroke-related sleep dysfunction. In some cases, the sleep dysfunction is insomnia, which may occur after a stroke, particularly in patients with a right hemisphere stroke or a stroke in the thalamus or brainstem, including the pontine tegmentum and thalamus-midbrain region. In some cases, the sleep dysfunction is somnolence, which may occur after a stroke in patients with subcortical (caudate nucleus, putamen), superior pons, medial pontomedometral stroke, or cortical stroke affecting the reticular activating system (RAS). Paramedian or bilateral thalamic strokes may initially cause coma, followed by somnolence upon awakening. Supratentorial strokes may reduce non-REM sleep, total sleep time, and ipsilateral or bilateral sleep spindle waves. Sawtooth waves may be reduced after hemispheric strokes. REM sleep may be reduced after occipital lobe strokes. Strokes in the pontomedometral junction and raphe nuclei may reduce the amount of non-REM sleep. Strokes in the lower pons can selectively reduce REM sleep. Strokes in the paramedian thalamus and lower pons may reduce slow-wave sleep.

[0112] The method involves placing a first electrode in the basal ganglia or cortical region of a subject's brain to deliver electrical stimulation (i.e., a DBS electrode) and a second electrode in the subcortical or cortical region of the subject's brain to detect electroencephalogram (EEG) signals from neural activity associated with a sleep feature or sleep stage of interest when the subject is asleep (i.e., a detection electrode). In some embodiments, one or more DBS electrodes are placed in the basal ganglia or cortical region, and one or more detection electrodes are placed in the subcortical or cortical region. The DBS electrode and detection electrode can be non-brain-penetrating surface electrodes; extracranial electrodes, for example, mounted under the galea aponeurotica or on the skull (in the case of a neurostimulator mounted in a skull drill cap or on the side of the skull); or brain-penetrating depth electrodes. When an EEG signal associated with a sleep feature or sleep stage of interest is detected from the subcortical or cortical region of the brain using the detection electrode, the DBS electrode can be used to apply electrical stimulation to the basal ganglia in a manner that is effective in treating sleep dysfunction.

[0113] In some embodiments, one or more detection electrodes are used to record electroencephalogram (EEG) signals of neural activity associated with a sleep feature or sleep stage of interest in one or more brain regions. The detection electrodes may be placed, for example, in the precentral gyrus and / or postcentral gyrus regions of the cortex to detect neural activity associated with the sleep feature or sleep stage of interest; or they may be placed in other brain regions suitable for detection. In some embodiments, the EEG signal data includes field potential data. The site selected for detection may vary for different subjects and may depend on the mapping of the individual subject's brain to identify the optimal location for placing the electrodes to detect EEG signals from neural activity associated with the sleep feature or sleep stage of interest, as discussed further below.

[0114] As used herein, the phrase "an electrode" or "the electrode" refers to a single electrode or multiple electrodes, such as an electrode array. As used herein, the term "contact," as used in the context of contact between an electrode and a region of the brain, refers to the physical association between the electrode and that region. In other words, the detection electrode in contact with a region of the brain physically contacts a region of the brain. DBS electrodes can conduct current to a specific target in the brain. The electrodes used in the methods disclosed herein can be unipolar (cathode or anode) or bipolar (e.g., having both an anode and a cathode).

[0115] Placing detection electrodes for recording neural activity at designated regions of the brain can be performed using standard surgical procedures for placing intracranial electrodes. In some cases, placing detection electrodes may involve placing the electrodes on the surface of a designated region of the brain. For example, the electrodes may be placed on the surface of the brain at the precentral gyrus or postcentral gyrus, or a combination thereof. The electrodes may contact at least a portion of the surface of the brain at the precentral gyrus or postcentral gyrus. In some embodiments, the electrodes may contact substantially the entire surface area at the precentral gyrus and postcentral gyrus. In some embodiments, the electrodes may additionally contact areas adjacent to the precentral gyrus or postcentral gyrus. In some embodiments, sensing electrodes may contact any area of ​​the cortex that allows for the detection of sleep characteristics or sleep stages. In some embodiments, the electrodes may be placed extracranially, such as in the subgastrional space. In some embodiments, sensing electrodes may be contained within a skull drill cap or on the case of an implantable neurostimulator device mounted on the cranial side. In some embodiments, an array of electrodes arranged on a planar support base may be used to detect electroencephalogram (EEG) signals of neural activity from one or more brain regions designated herein. The surface area of ​​the electrode array can be determined by the desired contact area between the electrode array and the brain. Electrodes used for implantation onto the brain surface, such as surface electrodes or surface electrode arrays, are available from commercial suppliers. Commercially available electrodes / electrode arrays can be modified to achieve the desired contact area. Non-brain-penetrating electrodes (also referred to as surface electrodes) that can be used in the methods disclosed herein can be electrocorticography (ECoG) electrodes, subgastroplasty electrodes, or electroencephalography (EEG) electrodes. In some embodiments, multiple electrodes are placed at one or more brain regions described herein for detecting EEG signals via stereotactic electroencephalography (sEEG).

[0116] In some cases, placing detection electrodes in a target region or site (e.g., a subcortical or cortical region of the brain) may involve placing brain-penetrating electrodes (also known as depth electrodes) in the specified region of the brain. For example, detection electrodes may be placed in a subcortical or cortical region of the brain. In some embodiments, the detection electrodes may additionally contact regions adjacent to the subcortical or cortical regions of the brain. In some embodiments, electrode arrays may be used to detect neural activity from cortical regions, such as the precentral gyrus or postcentral gyrus, or combinations thereof, as specified herein.

[0117] The depth to which the detection electrodes are inserted into the brain can be determined by the desired level of contact between the electrode array and the brain. Brain-penetrating electrode arrays are available from commercial suppliers. Commercially available electrode arrays can be tailored to achieve the desired depth of insertion into brain tissue.

[0118] Placing electrodes in the basal ganglia or cortical regions of the brain to deliver electrical stimulation can be performed using standard surgical procedures for placing electrodes for deep brain stimulation. For example, electrodes may be placed in the subthalamic nucleus region, globus pallidus region, or thalamic region, or other intracranial regions. Medical imaging, such as magnetic resonance imaging (MRI) or computed tomography (CT), can be used to guide the placement of DBS electrodes and verify their correct placement in the brain. Alternatively, a neurostimulator that generates electrical impulses is placed under the skin of the chest, typically below the clavicle, or in the abdomen. In some embodiments, the neurostimulator is cranially implanted. The surgical procedure may involve placing the DBS electrode in the brain through a small opening in the skull. Electrode leads are tunneled under the skin of the neck and chest to connect to the neurostimulator implanted in the chest.

[0119] An electrical current is supplied to the DBS electrodes by a neurostimulator. Parameters such as pulse width, shape, frequency, amplitude, pattern, and temporal distribution can be adjusted in response to changes in neural activity in subcortical or cortical regions of the brain, or alternatively in response to accelerometer readings, pulse oximeter readings, temperature, or heart rate, to treat sleep dysfunction. In some embodiments, a closed-loop system is used to automatically adjust DBS settings in response to changes in neural activity in subcortical or cortical regions of the brain. In other embodiments, an open-loop system is used, wherein the DBS settings are adjusted by the user or medical practitioner based on neural activity in subcortical or cortical regions of the brain.

[0120] Electrical stimulation can be applied using a single electrode, an electrode pair, or an array of electrodes. In some embodiments, the number of electrodes used to deliver electrical stimulation to the brain ranges from 8 to 32, including any number of electrodes within this range, such as 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, or 32 electrodes. In some embodiments, electrical stimulation is applied to more than one site in the basal ganglia or cortex. The sites of electrical stimulation can be alternated or otherwise patterned spatially or temporally. Electrical stimulation can be applied to the site simultaneously or sequentially. In some embodiments, the region of the basal ganglia to which electrical stimulation is applied is the subthalamic nucleus region or the globus pallidus region, or other suitable regions of the basal ganglia for stimulation. The site selected for stimulation may vary for different subjects and will depend on the mapping of the basal ganglia region or cortical region of the individual subject's brain to identify the optimal location for placing the electrodes used to deliver electrical stimulation to treat sleep dysfunction.

[0121] In some embodiments, an electrode array disposed on a planar support substrate can be used to electrically stimulate the basal ganglia. The surface area of ​​the electrode array can be determined by the desired contact area between the electrode array and the basal ganglia. In some cases, cylindrical electrode arrays, paddle electrode arrays, or plate electrode arrays can be used in the methods for deep brain stimulation disclosed herein. Such DBS electrode arrays for implantation in the brain are available from commercial suppliers. Commercially available electrode / electrode arrays can be adjusted to achieve the desired contact area.

[0122] The exact number of DBS electrodes or detection electrodes included in an electrode array (e.g., for electrical stimulation or detection of neural activity) can vary. In some aspects, the electrode array may include two or more electrodes, such as three or more, including four or more, for example, about 3 to 6 electrodes, about 6 to 12 electrodes, about 12 to 18 electrodes, about 18 to 24 electrodes, about 24 to 30 electrodes, about 30 to 48 electrodes, about 48 to 72 electrodes, about 72 to 96 electrodes, or about 96 or more electrodes. The electrodes may be arranged in a regular repeating pattern (e.g., a grid, such as a grid with about 1 cm spacing between electrodes), or without a pattern. Electrodes that conform to the target site can be used for optimal delivery of electrical stimulation. One such example is a single multi-contact electrode with eight contacts spaced 2.5 mm apart. Each contact will have a span of about 2 mm. Another example is an electrode with two 1 cm contacts with a 2 mm intermediate gap. Furthermore, another example of an electrode that can be used in this method is a two- or three-branched electrode covering the target area. Each of these tri-branched electrodes has four 1-2 mm contacts, with a center-to-center spacing of 2 mm or 2.5 mm and a span of 1.5 mm.

[0123] The size of each electrode can also vary depending on factors such as the number of electrodes in the array, the location of the electrodes, the material, the patient's age, and other factors. In some respects, electrode arrays have a size (e.g., diameter) of about 5 mm or smaller, such as about 4 mm or smaller, including 4 mm-0.25 mm, 3 mm-0.25 mm, 2 mm-0.25 mm, 1 mm-0.25 mm, or about 3 mm, about 2 mm, about 1 mm, about 0.5 mm, or about 0.25 mm.

[0124] In some embodiments, the method further includes mapping the subject's brain to optimize the placement of electrodes for applying electrical stimulation. The placement of the DBS electrodes is optimized to maximize the clinical response to electrical stimulation for treating sleep disorders, which may include insomnia, difficulty falling asleep, difficulty maintaining sleep, insufficient total sleep time, insufficient deep sleep time, insufficient non-REM or REM sleep time, and / or inability to achieve deep sleep (N3 stage or delta sleep). In some embodiments, the DBS is optimized to achieve neurophysiologically defined changes, such as an increase or decrease in slow waves or sleep spindles. In some embodiments, the subthalamic nucleus region, globus pallidus region, or other regions of the thalamus or brain are mapped to determine the optimal placement of the DBS electrodes.

[0125] The efficacy of electrical stimulation at specific sites for treating sleep disorders can be assessed using any standard method. In some implementations, subjects are monitored during sleep using a motion recorder, electroencephalogram (EEG), or polysomnography. Autonomic data can also be collected while the subject is sleeping. In some cases, the observer may monitor the subject at night to determine whether the subject remains asleep or experiences nocturnal awakenings. Furthermore, the Visual Analogue Scale (VAS), Likert scale, Stanford Sleepiness Scale (SSS), Maintenance of Wakefulness Test (MWT), Epworth Sleepiness Scale (ESS), Multiple Sleep Latency Test (MSLT), or Athens Insomnia Scale can be used to evaluate the efficacy of electrical stimulation in treating sleep disorders.

[0126] In some embodiments, the method further includes mapping the subject's brain to optimize the placement of detection electrodes. The placement of the detection electrodes in subcortical or cortical regions is optimized to detect brain activity characteristics associated with sleep characteristics or sleep stages to be treated with electrical stimulation. For example, the level of total power or power in a specific frequency range (e.g., α, β, γ, δ, and / or θ) can be associated with different sleep stages. In some embodiments, the N3 sleep stage is identified by an increase in δ power during the N3 sleep stage. In some embodiments, the N2 or N3 sleep stage is identified by a decrease in β power in the frequency range of 12 Hz to 30 Hz, a decrease in γ power in the frequency range of 30 Hz to 60 Hz, an increase in low-frequency θ power in the frequency range of 5 Hz to 10 Hz, and / or an increase in δ power in the frequency range of 0.5 Hz to 4.5 Hz. Therefore, detection electrodes can be placed to optimize the detection of brain activity within a specific frequency range that is associated with sleep characteristics and / or sleep stages to be treated with electrical stimulation.

[0127] Detection of brain activity can be performed using any method known in the art. For example, functional brain imaging of neural activity can be performed using electrical methods such as electroencephalography (EEG), stereotactic electroencephalography (sEEG), electrocorticography (ECoG), magnetoencephalography (MEG), single-photon emission computed tomography (SPECT), and metabolic and blood flow studies such as functional magnetic resonance imaging (fMRI) and positron emission tomography (PET). In some embodiments, subcortical regions, cortical regions, or other regions are mapped to determine the optimal placement for detection electrodes. One or more of these regions may be implanted with detection electrodes to measure electrical signals from neural activity associated with sleep characteristics or sleep stages to be treated with electrical stimulation.

[0128] In some implementations, accelerometers, non-invasive sleep monitoring devices, wearable sleep monitoring devices (e.g., smart rings, smartwatches, wristbands, or headband sleep trackers), photoplethysmography (PPG)-based sleep monitoring devices, or radar-based sleep monitoring devices are used to assist in the classification of sleep characteristics or sleep stages. For descriptions of such sleep monitoring devices, see, for example, Toften et al. (2020) Sleep Med. 75: 54-61, Kwon et al. (2021) IEEE J Biomed Health Inform. 25(10): 3844-3853, Lauteslager et al. (2020) Annu Int ConfIEEE Eng Med Biol Soc. 2020: 5150-5153, Chung et al. (2017) Sensors (Basel) 17(7): 1685, An et al. (2022) Sci Rep. 12(1): 21052, Chinoy et al. (2021) Sleep 44(5): zsaa291, Chinoy et al. (2022) Nat Sci Sleep 14: 493-516, Shelgikar et al. (2016) Chest 150(3): 732-43, Zhao et al. (2021) Entropy (Basel) 23(1):116; incorporated herein by reference in its entirety.

[0129] As illustrated herein, the subject approach involves applying electrical stimulation to basal ganglia regions (e.g., the subthalamic nucleus region and / or the globus pallidus region and / or the thalamus region) or cortical regions in a manner that effectively treats a subject's sleep dysfunction when neural activity associated with a sleep characteristic or sleep stage requiring treatment is detected. In some embodiments, electrical stimulation is applied to basal ganglia regions (e.g., the subthalamic nucleus region and / or the globus pallidus region and / or the thalamus region) when N2 sleep stages and / or N3 sleep stages are detected.

[0130] Closed-loop therapy can be performed using a neurostimulator used in conjunction with a neural recording device that records brain electrical activity while the subject is sleeping. Electrical stimulation is delivered to the basal ganglia of the subject's brain when a pattern of neural activity associated with a selected sleep characteristic or sleep stage to be treated is detected. Parameters for applying electrical stimulation to the brain can be determined empirically during treatment or can be predefined, for example, from trial studies on subjects. For instance, subcortical or cortical EEG signal data (e.g., from the precentral gyrus and / or postcentral gyrus) can be recorded while the subject is sleeping. Varying stimulation settings, including baseline (stimulus off), optimal therapeutic stimulation, adjusted stimulation and ineffective stimulation, and maximally tolerated stimulation, can be applied during sleep stages or when certain sleep characteristics are detected to identify individual neural characteristics of a patient's "sleep dysfunction" and "remission of sleep dysfunction," which are used to assist in programming the DBS device to determine the optimal therapeutic stimulation parameters for treating sleep dysfunction during individual sleep stages or when certain sleep characteristics are detected. The parameters of electrical stimulation may include one or more of the following: frequency, pulse width / duration, duty cycle, intensity / amplitude, pulse mode, programmed duration, programmed frequency, etc.

[0131] Frequency refers to the number of pulses generated per second during stimulation and is expressed in Hertz (Hz, e.g., 60 Hz = 60 pulses per second). The frequency of electrical stimulation used in this method can vary widely depending on many factors and can be determined empirically during treatment of the subject or can be predetermined. In some embodiments, the method may involve applying electrical stimulation to the brain at frequencies ranging from 2 Hz to 250 Hz, such as 25 Hz–200 Hz, 50 Hz–250 Hz, 50 Hz–185 Hz, 50 Hz–150 Hz, 75 Hz–200 Hz, 100 Hz–200 Hz, 100 Hz–180 Hz, 100 Hz–160 Hz, or 130 Hz–150 Hz. In some embodiments, electrical stimulation of the brain is applied at a frequency of about 120 Hz to about 160 Hz, including any pulse frequency within this range, such as 120 Hz, 122 Hz, 124 Hz, 126 Hz, 128 Hz, 130 Hz, 132 Hz, 134 Hz, 136 Hz, 138 Hz, 140 Hz, 142 Hz, 144 Hz, 146 Hz, 148 Hz, 150 Hz, 152 Hz, 154 Hz, 156 Hz, 158 Hz, or 160 Hz. In some embodiments, non-integer pulse frequencies (e.g., 130.2 Hz, 130.4 Hz, etc.) are used.

[0132] Electrical stimulation can be applied in pulses such as monophasic or biphasic pulses. The time span of a single pulse is called the pulse width or pulse duration. The pulse width used in this method can vary widely depending on many factors (e.g., the severity of the disease, the patient's condition, etc.) and can be determined empirically or predefined. In some embodiments, the method may involve applying electrical stimulation with pulse widths of approximately 10 microseconds, –500 microseconds, such as 20 microseconds, –450 microseconds, 40 microseconds, –450 microseconds, 60 microseconds, –450 microseconds, 60 microseconds, –220 microseconds, 60 microseconds, –120 microseconds, or 60 microseconds, –90 microseconds. In some implementations, electrical stimulation of the brain is applied with a pulse width of about 60 microseconds μ to about 210 microseconds μ, including any pulse width within this range, such as 60 microseconds, 65 microseconds, 70 microseconds, 75 microseconds, 80 microseconds, 85 microseconds, 90 microseconds, 95 microseconds, 100 microseconds, 105 microseconds, 110 microseconds, 115 microseconds, 120 microseconds, 125 microseconds, 130 microseconds, 135 microseconds, 140 microseconds, 145 microseconds, 150 microseconds, 155 microseconds, 160 microseconds, 165 microseconds, 170 microseconds, 175 microseconds, 180 microseconds, 185 microseconds, 190 microseconds, 195 microseconds, 200 microseconds, 205 microseconds, 210 microseconds, 215 microseconds, or 220 microseconds.

[0133] Electrical stimulation can be applied for durations ranging from 0.1 seconds to 1 month, with rest periods (i.e., no electrical stimulation). In some cases, the duration of electrical stimulation can be 0.1 seconds to 1 week, 1 second to 1 day, 10 seconds to 12 hours, 1 minute to 6 hours, 10 minutes to 1 hour, etc. In other cases, the duration of electrical stimulation can be 1 second to 1 minute, 1 second to 30 seconds, 1 second to 15 seconds, 1 second to 10 seconds, 1 second to 6 seconds, 1 second to 3 seconds, 1 second to 2 seconds, or 6 seconds to 10 seconds. The rest periods between each stimulation period can be 60 seconds or less, 30 seconds or less, 20 seconds or less, or 10 seconds. In some implementations, electrical stimulation can be applied for a duration of one year or more, two years or more, three years or more, five years or more, or ten years or more. In some implementations, as part of a long-term DBS therapy regimen, electrical stimulation can be continued indefinitely.

[0134] Electrical stimulation can be applied at a current amplitude of 0.1 mA-30 mA, such as 0.1 mA-25 mA, such as 0.1 mA-20 mA, 0.1 mA-15 mA, 0.1 mA-10 mA, 0.1 mA-2 mA, 0.1 mA-1 mA, 1 mA-20 mA, 1 mA-10 mA, 2mA-30 mA, 2 mA-15 mA, 2 mA-10 mA or 1 mA-3 mA. In some embodiments, the magnitude of the current is 0.1 mA - 3.5 mA, or any current magnitude within this range, such as 0.1 mA, 0.2 mA, 0.3 mA, 0.4 mA, 0.5 mA, 0.6 mA, 0.7 mA, 0.8 mA, 0.9 mA, 1.0 mA, 1.1 mA, 1.2 mA, 1.3 mA, 1.4 mA, 1.5 mA, 1.6 mA, 1.7 mA, 1.8 mA, 1.9 mA, 2.0 mA, 2.1 mA, 2.2 mA, 2.3 mA, 2.4 mA, 2.5 mA, 2.6 mA, 2.7 mA, 2.8 mA, 2.9 mA, 3.0 mA, 3.1 mA, 3.2 mA, 3.3 mA, 3.4 mA or 3.5 mA.

[0135] Electrical stimulation can be applied in voltage amplitudes ranging from 0.1 V to 15 V, such as 0.1 V to 10 V, 0.1 V to 5 V, 1 V to 10 V, 1 V to 5 V, or 1 V to 3.5 V. In some implementations, the voltage amplitude is 1 V to 3.5 V, or any voltage amplitude within that range, such as 1 V, 1.1 V, 1.2 V, 1.3 V, 1.4 V, 1.5 V, 1.6 V, 1.7 V, 1.8 V, 1.9 V, 2.0 V, 2.1 V, 2.2 V, 2.3 V, 2.4 V, 2.5 V, 2.6 V, 2.7 V, 2.8 V, 2.9 V, 3.0 V, 3.1 V, 3.2 V, 3.3 V, 3.4 V, or 3.5 V.

[0136] Electrical stimulation with the parameters described above can be applied for a programmed duration of approximately one day or less, such as 18 hours, 6 hours, 3 hours, 2 hours, 1 hour, 45 minutes, 30 minutes, 20 minutes, 10 minutes, or 5 minutes or less, for example, 1 minute to 5 minutes, 2 minutes to 10 minutes, 2 minutes to 20 minutes, 2 minutes to 30 minutes, 5 minutes to 10 minutes, 5 minutes to 30 minutes, or 5 minutes to 15 minutes, 10 minutes to 400 minutes, 25 minutes to 300 minutes, 50 minutes to 200 minutes, or 75 minutes to 150 minutes, the time period including the application of pulses and the intermediate rest periods. This programmed procedure can be repeated at a desired programmed frequency to alleviate the subject's sleep dysfunction. Therefore, the treatment protocol can include a procedure for applying electrical stimulation at a desired programmed frequency and duration. In some embodiments, the treatment protocol is controlled by a control unit that communicates with a pulse generator connected to one or more DBS electrodes in the closed-loop therapy protocol.

[0137] In some implementations, an upper limit can be set on the maximum number of electrical stimulation sessions per night. For example, the maximum number of electrical stimulation sessions per night can range from 50 sessions to 500 sessions per night, including any number of sessions per night within this range, such as 50, 75, 100, 125, 150, 175, 200, 225, 250, 275, 300, 325, 350, 375, 400, 425, 450, 475, or 500 sessions per night. Alternatively or additionally, an upper limit can be set on the total duration of electrical stimulation per night. For example, the total duration of electrical stimulation per night can range from 10 minutes to 100 minutes per night, including any duration within this range, such as 10, 20, 30, 40, 50, 60, 70, 80, 90, or 100 minutes per night.

[0138] As described above, this treatment can improve sleep dysfunction suffered by the subject. Improvement in sleep dysfunction may include increasing non-REM or REM sleep time, increasing N2 stage sleep time and / or increasing N3 stage sleep time. The efficacy of the treatment can be assessed using any known method for assessing sleep dysfunction. In some implementations, subjects are monitored during sleep using a motion recorder, electroencephalogram (EEG), or polysomnography. Autonomic data may also be collected while the subject is sleeping. In some cases, the observer may monitor the subject at night to determine whether the subject remains asleep or has nocturnal awakenings. In addition, the Visual Analogue Scale (VAS), Likert scale, Stanford Sleepiness Scale (SSS), Maintenance of Wakefulness Test (MWT), Epworth Sleepiness Scale (ESS), Multiple Sleep Latency Test (MSLT), or Athens Insomnia Scale can be used to assess the efficacy of the treatment for the subject's sleep dysfunction.

[0139] In some cases, the efficacy of treatment can be assessed by detecting activity (e.g., electrical signals) associated with sleep characteristics or sleep stages, which can occur in subcortical, cortical, or other regions. For example, brain regions could be the precentral gyrus and / or postcentral gyrus, and may include readouts of physiologically important variables associated with sleep, such as delta / slow waves, spindle waves, K-complexes, beta bursts and beta oscillations, and spectral coherence. Detection of brain activity can be performed using functional brain imaging. Functional brain imaging can be performed using electrical methods such as electroencephalography (EEG), chronic subgastroplasty, burr hole or cranial lateral neurostimulator electrode recording, electrocorticography (ECoG), magnetoencephalography (MEG), single-photon emission computed tomography (SPECT), and metabolic and blood flow studies such as functional magnetic resonance imaging (fMRI) and positron emission tomography (PET). In some implementations, the electrical method for assessing the efficacy of treatment may involve using detection electrodes as described herein or placing additional electrodes to measure electrical signals at secondary regions of the brain or within or outside the skull. Electrodes may be implanted in one or more regions of the brain, and electrical signals may be measured to assess the efficacy of treatment. Any suitable electrode may be used for measurement and may include one or more surface electrodes (non-brain-penetrating electrodes) or one or more depth electrodes (brain-penetrating electrodes) as described herein.

[0140] Assessments of the efficacy of treatment and improvements in sleep dysfunction can be performed at any appropriate time point after the start of the treatment program, such as during open-loop or closed-loop therapy or after the completion of the treatment protocol. Implementation of the subject-specific approach includes assessing the efficacy of treatment or improvement in sleep dysfunction within seconds, minutes, hours, or days after the initial treatment protocol has been completed. In some cases, assessments can be performed at multiple time points. In some cases, more than one type of assessment can be performed at different time points. In some implementations, subcortical or cortical brain activity in the subject can be measured before the application of electrical stimulation (e.g., in the precentral gyrus and / or postcentral gyrus), and the assessment may include comparing the subject's brain activity after treatment with brain activity before treatment, and changes in brain activity after treatment may indicate successful treatment.

[0141] After the treatment regimen is completed, the effectiveness of the treatment can be assessed, and the treatment regimen can be repeated if necessary. In some cases, the treatment regimen can be modified before repetition. For example, one or more of the following can be changed before starting a second treatment regimen: frequency, pulse width, current amplitude, duration of electrical stimulation, program duration, program frequency, and / or placement of DBS electrodes or detection electrodes.

[0142] The application of this method may include a prior step of selecting patients for treatment based on needs determined by clinical assessment, which may include assessing the severity of chronic sleep dysfunction (e.g., sleep dysfunction lasting at least 3 months), physical condition, medication regimen, cognitive assessment, anatomical assessment, behavioral assessment, and / or neurophysiological assessment. In some cases, subjects may be further assessed to determine whether deep brain stimulation (DBS) will completely or partially (e.g., at least 50%) alleviate sleep dysfunction. Such patients may undergo DBS on a trial basis to determine whether DBS reduces the severity of the sleep dysfunction experienced by the patient. Such patients may also have implanted detection electrodes to identify personalized neurofeatures of “sleep dysfunction” and “sleep dysfunction relief” at selected sleep characteristics or sleep stages to aid in deep brain stimulation programming to determine the patient’s therapeutic stimulation parameters and / or assess whether DBS therapy will effectively improve the patient’s sleep.

[0143] In some aspects, the methods and systems of this disclosure may include measuring brain activity, such as electrical activity in subcortical or cortical regions, wherein the level of β and / or δ frequency power may be measured. In some cases, electrical activity from multiple locations in the subcortical or cortical regions may be measured and averaged. In some embodiments, electrical activity in a β frequency range (such as 12 Hz to 30 Hz) and / or a δ frequency range (such as 0.5 Hz to 4.5 Hz) may be measured from subcortical or cortical regions of a subject's brain. In some cases, data-driven methods are used to identify spectral characteristics that are personalized and differ from normative power bands. In some cases, electrical activity in one or more locations in the brain can be measured and the power of the activity or other signals can be monitored during a period of time, extending from before stimulation to the application of stimulation to basal ganglia regions (e.g., subthalamic nucleus, globus pallidus, or thalamic regions) or cortical regions, or extending to the period after stimulation of the basal ganglia has been applied. This includes monitoring for increases or decreases in the power of delta frequency (e.g., 0.5 Hz to 4.5 Hz) and / or beta frequency (e.g., 12 Hz to 30 Hz) activity. In some cases, when the power of beta frequency (e.g., 12 Hz to 30 Hz) and / or delta frequency (e.g., 0.5 Hz to 4.5 Hz) activity is within normal ranges (e.g., ranges associated with normal sleep), the method and system do not apply further stimulation to the brain. Alternatively, when the power of beta frequency (e.g., 12 Hz to 30 Hz) and / or delta frequency (e.g., 0.5 Hz to 4.5 Hz) activity is outside the normal range (e.g., the range associated with severe sleep dysfunction), the method and system can apply further stimulation to the brain. In some cases, applying electrical stimulation to the brain can suppress beta frequency (e.g., 12 Hz to 30 Hz) and / or increase or decrease gamma frequency (e.g., 30 Hz to 60 Hz) activity detected in subcortical or cortical regions. This decrease can be compared to the power before stimulation was applied. In some cases, applying electrical stimulation to the brain may alter other neural characteristics from one or more regions of the brain. The changes can be compared to the state of these characteristics before stimulation was applied.

[0144] The closed-loop method allows for the determination of electrical stimulation parameters based on real-time feedback signals from the subject's brain. The closed-loop method and system allow for the automation of treatment of the subject, including real-time, demand-based adjustment of the treatment protocol. Exemplary closed-loop methods and associated systems for treating sleep dysfunction are further discussed in the Examples section and... Figure 1B The following describes a closed-loop method and system for the automated delivery of electrical stimulation.

[0145] Closed-loop method for automated delivery of electrical stimulation

[0146] In some embodiments, a control algorithm is used to automate the delivery of electrical stimulation to the brain in response to the detection of neural activity associated with sleep characteristics or sleep stages selected for treatment. According to some embodiments, the method may include receiving electrical signals from subcortical or cortical regions of a subject's brain (e.g., the precentral or postcentral gyrus) via detection electrodes; applying electrical signal measurements to a control algorithm modulated for clinically relevant goals (e.g., the range of signals indicating effective treatment); and, if the electrical signal measurements indicate that the patient requires treatment, automatically delivering electrical stimulation via DBS electrodes to basal ganglia regions of the brain (e.g., the subthalamic nucleus, globus pallidus, or thalamic region) or cortical regions in a manner effective for treating sleep dysfunction. For example, electrical activity in a beta frequency range (such as 12 Hz to 30 Hz) and / or delta frequency range (such as 0.5 Hz to 4.5 Hz) from subcortical or cortical regions (e.g., the precentral or postcentral gyrus) can be measured using detection electrodes. When the power levels of the beta frequencies (such as 12 Hz to 30 Hz) and / or delta frequencies (such as 0.5 Hz to 4.5 Hz) indicate that the patient is in sleep stage N2 or N3, a control algorithm receives the electrical activity data from the detection electrodes and automates the delivery of electrical stimulation to the brain via DBS electrodes. In some embodiments, one or more programmed stimulation parameters are adjusted based on the recorded electrical activity data according to the algorithm's control rules; and the adjusted electrical stimulation is delivered to the brain via DBS electrodes in a manner that effectively improves the subject's sleep (e.g., deep sleep).

[0147] In some implementations, the N2 sleep stage or the N3 sleep stage is identified by one or more spectral power changes selected from: a decrease in β power in the frequency range of 12 Hz to 30 Hz compared to the β power when the subject is awake; a decrease in γ power in the frequency range of 30 Hz to 60 Hz compared to the γ power when the subject is awake; an increase in θ power in the frequency range of 5 Hz to 10 Hz compared to the θ power when the subject is awake; and an increase in δ power in the frequency range of 0.5 Hz to 4.5 Hz compared to the δ power when the subject is awake.

[0148] In some implementations, the N2 sleep stage or the N3 sleep stage is identified by detecting one or more changes in spectral power in conjunction with one or more changes in cortico-subcortical spectral coherence, wherein the one or more changes in cortico-subcortical spectral coherence are selected from: an increase in δ-cortico-subcortical spectral coherence compared to δ-cortico-subcortical spectral coherence when the subject is awake; and a decrease in β-cortico-subcortical spectral coherence compared to β-cortico-subcortical spectral coherence when the subject is awake.

[0149] In some embodiments, the pre-wake period or wakefulness period is identified by one or more spectral power variations selected from: a decrease in cortical delta power in the frequency range of 1 Hz to 4 Hz compared to the average cortical delta power during deep non-rapid eye movement (NREM) sleep; an increase in cortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power during deep NREM sleep; an increase in subcortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power during deep NREM sleep; and an increase in subcortical beta power in the frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power during deep NREM sleep. In some embodiments, the increase in subcortical beta power precedes the decrease in cortical delta power.

[0150] In some embodiments, the post-awakening time period is identified by one or more spectral power changes selected from: a decrease in cortical delta power in the frequency range of 1 Hz to 4 Hz compared to the average cortical delta power in the pre-awakening time period; an increase in cortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power in the pre-awakening time period; an increase in subcortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power in the pre-awakening time period; and an increase in subcortical beta power in the frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power in the pre-awakening time period.

[0151] In some implementations, electrical stimulation increases cortical delta power, decreases cortical alpha power, decreases cortical beta power, and decreases cortical σ power.

[0152] As described in the preceding sections, the efficacy of treatment for sleep dysfunction can be assessed by using detection electrodes to detect brain electrical activity associated with selected sleep characteristics or sleep stages. In an open-loop system, stimulation is delivered in a pre-programmed manner or manually by the user, but is not automatically controlled by real-time neural feedback from the patient's brain. The electrical activity can be analyzed by a computing device that can output recommendations based on a comparison of the electrical activity with predetermined ranges. The user can then implement the recommendations, such as changing the parameters of the electrical stimulation program before initiating another treatment regimen. In contrast, in a closed-loop system, the computing device can automatically update stimulation parameters and / or automatically deliver stimulation to the brain according to the electrical stimulation program based on analysis of recorded electrical signals. In some implementations, the open-loop or closed-loop system can be integrated with mechanisms for user intervention, such as allowing the user to override the open-loop or closed-loop stimulation program to formulate or prevent stimulation that would normally occur, or to manually change the parameters of such stimulation.

[0153] In some embodiments, the computing device for guiding closed-loop stimulation may be a combination of hardware and software that can be wirelessly or wiredly connected to the measuring electrodes. The computing device may communicate with a control unit (also called a control module) that controls a neurostimulator pulse generator connected to the DBS electrodes. In some embodiments, the computing device may be connected to a recorder (e.g., a neurophysiological recorder or neural recording device) that records brain activity measured by the detection electrodes. The computing device may include a control algorithm that determines adjustments to stimulation parameters based on the real-time output of the neurophysiological recorder. This algorithm may operate by simple on / off control of stimulation at set parameters, adjusting only the on / off parameters in each evaluation cycle, or by determining complex adjustments to the range of stimulation parameters in each cycle. In some cases, the algorithm may be based on information about which sleep stages are associated with sleep dysfunction, such as indicating the sleep characteristics to be treated by electrical stimulation or the range of electrical activity in the sleep stage. The algorithm may also include additional information, such as a profile of brain activity in a normal subject (without sleep dysfunction). Regardless of the specific control algorithm structure, the computing device can be tuned for clinically relevant goals (e.g., the range of signals indicating effective treatment and / or the range of signals indicating sleep dysfunction and treatment needs). It guides the adjustment of one or more programmed stimulation parameters according to the control laws of the algorithm, and applies the tuned electrical stimulation to the basal ganglia region or cortical region of the brain via DBS electrodes.

[0154] In some cases, the computing device, via a control algorithm, can determine whether the received electrical signal is within or outside a predetermined range of neural signals indicating a selected sleep characteristic or sleep stage for targeted treatment using electrical stimulation. When the received electrical signal is outside the predetermined range, the computing device determines that the subject is in a non-target sleep characteristic or sleep stage. The computing device can then communicate with the control unit to guide stimulation shutdown via the neurostimulator pulse generator. When the received electrical signal is within a predetermined range of neural signals indicating a target sleep characteristic or sleep stage, the computing device determines that the subject should be treated with deep brain stimulation. The control algorithm within the computing device can then determine whether the initial step of applying electrical stimulation to the brain should be repeated, and / or whether the parameters of the electrical stimulation should be adjusted before the step of applying electrical stimulation in the selected sleep stage or when the selected sleep characteristic is detected. The computing device can then communicate with the control unit via the control unit to provide appropriate instructions to the neurostimulator pulse generator.

[0155] In some implementations, the computing device can determine whether the received electrical signal is within or outside a second predetermined range, where the second predetermined range indicates a second sleep characteristic or sleep stage targeted for treatment with electrical stimulation. When the received electrical signal is within the second predetermined range, the computing device determines that the subject should be treated with deep brain stimulation. Then, when the received electrical signal is outside the second predetermined range, the computing device can communicate with a control unit to guide stimulation cutoff via a pulse generator. A control algorithm within the computing device can then determine whether the initial step of applying electrical stimulation should be repeated and / or whether the parameters of the electrical stimulation should be adjusted before the step of applying electrical stimulation. The processor can then communicate with the control unit to provide appropriate instructions to the pulse generator.

[0156] Therefore, in some aspects, the subject method operates as a closed-loop control system that can automatically adjust one or more parameters and / or automatically deliver stimulation to the brain according to an electrical stimulation program in response to electrical activity from regions of the subject's brain. In some embodiments, the closed-loop control system automatically delivers stimulation according to set parameters when the received electrical signals are within a predetermined range of sleep characteristics or sleep stages that indicate targeted treatment with electrical stimulation. Exemplary closed-loop methods and associated systems are described in the Embodiments section of this application and Figure 1B Example in.

[0157] In some aspects, closed-loop systems can be used to sense a subject's need for treatment using the methods disclosed herein. For example, a closed-loop system can be programmed to monitor brain activity from one or more subcortical or cortical regions of the brain and compare brain activity corresponding to one or more sleep characteristics or sleep stages with a range indicating a sleep disorder. Upon detection of electrical activity indicating a sleep disorder, the closed-loop system can automatically initiate a treatment regimen targeting the sleep disorder by applying electrical stimulation to the brain at one or more sleep stages or when one or more sleep characteristics indicating sleep impairment are detected. In some embodiments, a control algorithm adjusts one or more programmed stimulation parameters to maximize slow-wave activity to improve deep sleep. In some embodiments, the closed-loop system is programmed to monitor brain activity from one or more subcortical or cortical regions of the brain to determine when the subject is in N2, N3, or REM sleep stages, and automatically initiates a treatment regimen by applying electrical stimulation to the brain when a pattern of neural activity associated with N2, N3, or REM sleep stages is detected.

[0158] In an additional aspect, closed-loop systems can be used to monitor brain activity at specific sleep stages or when specific sleep features are detected and to correlate brain activity with sleep dysfunction. For example, because a closed-loop system is configured to record electrical signals from a subject's brain, sleep features or sleep stages can be monitored in real time while the subject is sleeping and correlated with the measured electrical signals to provide biomarkers associated with sleep dysfunction in the subject at individual sleep stages or when certain sleep features are detected. For instance, electrical activity measured when a subject is experiencing sleep dysfunction can be used to develop biomarkers, such as those indicating the range of electrical activity associated with sleep dysfunction. Therefore, closed-loop systems are useful for detecting sleep dysfunction.

[0159] It should be understood that electrical signals indicating a subject's sleep dysfunction or its relief can be recorded from the subject's brain and can be used for aspects outside the closed-loop system. For example, electrical signals indicating a subject's sleep dysfunction or its relief can be recorded from subcortical or cortical regions (e.g., the precentral or postcentral gyrus) or other brain regions using electrodes or another device operatively coupled to the patient's brain, which may or may not be part of the closed-loop system. The patient can be treated as disclosed herein (e.g., by applying electrical stimulation to the brain), and electrical signals can be recorded in real time from subcortical or cortical regions or other regions during or after treatment. The electrical signals recorded after the start of electrical stimulation can then be compared with those recorded before treatment to determine characteristics in the recorded electrical signals that have changed after treatment. These characteristics provide feedback signals indicating whether the treatment has an effect on the patient's sleep dysfunction at a particular sleep stage. These characteristics can also be used as feedback signals for the closed-loop system. These characteristics may include total power or power within a specific frequency range (e.g., α, β, γ, δ, and / or θ). In some cases, these features may be patient-specific, specific to a particular sleep stage, or both. For example, some features may be those found in multiple patients with sleep dysfunction at a specific sleep stage; others may be features specific to a particular patient that may not be found in a significant number of other patients with sleep dysfunction. In some implementations, a combination of patient-specific features and sleep dysfunction-specific features may be monitored to assess the efficacy of treatment.

[0160] In a specific aspect, the closed-loop system and method provided herein may involve recording electrical signals from one or more subcortical or cortical regions of a patient's brain (e.g., the precentral gyrus or postcentral gyrus, or other regions) where the patient suffers from sleep dysfunction associated with motor disorders or neurological conditions. The patient can then be treated by applying electrical stimulation to regions of the basal ganglia or cortical areas of the brain, and electrical signals can be recorded from these subcortical or cortical regions (e.g., the precentral gyrus or postcentral gyrus, or other regions) and compared with pre-treatment recordings. Characteristics in the recorded signals that change after treatment will correspond to biomarkers indicating whether the treatment is effective. Changes in the recorded signals may also optionally be correlated with the level of sleep dysfunction reported by the patient after treatment. This change can be used to modulate treatment within the closed-loop system. For example, when changes in the recorded signals are correlated with improvements in sleep, these characteristics will indicate to the computational apparatus of the closed-loop system that further treatment is not required.

[0161] In some embodiments, one or more pattern recognition methods may be used to analyze recorded EEG activity data to automate the detection of brain activity features that distinguish sleep stages (e.g., wakefulness, N1, N2, N3, and REM) and / or sleep characteristics (e.g., slow waves, sleep spindles, K-complexes, beta bursts, pre-wake periods, wake periods, post-wake periods, or sleep stage transitions) from one another. Models and / or algorithms may be provided in a machine-readable format and may be used to correlate the level of total power or power in a specific frequency range (e.g., α, β, γ, δ, and / or θ) with the sleep characteristics or sleep stage for which deep brain stimulation therapy is to be administered. In some embodiments, the level of power at beta frequencies (e.g., 12 Hz to 30 Hz) and / or delta frequencies (e.g., 0.5 Hz to 4.5 Hz) is correlated with sleep stage N2 or N3 to determine whether the patient is eligible for electrical stimulation therapy. In some implementations, the N2 or N3 sleep stage is identified by a decrease in β power in the frequency range of 12 Hz to 30 Hz, a decrease in γ power in the frequency range of 30 Hz to 60 Hz, an increase in low-frequency θ power in the frequency range of 5 Hz to 10 Hz, and an increase in δ power in the frequency range of 0.5 Hz to 4.5 Hz. Alternatively or additionally, other characteristics of coherence or network connectivity within certain spectral bands may be associated with the sleep characteristics or sleep stage to be treated with electrical stimulation.

[0162] In some implementations, the N2 sleep stage or the N3 sleep stage is identified by one or more spectral power changes selected from: a decrease in β power in the frequency range of 12 Hz to 30 Hz compared to the β power when the subject is awake; a decrease in γ power in the frequency range of 30 Hz to 60 Hz compared to the γ power when the subject is awake; an increase in θ power in the frequency range of 5 Hz to 10 Hz compared to the θ power when the subject is awake; and an increase in δ power in the frequency range of 0.5 Hz to 4.5 Hz compared to the δ power when the subject is awake.

[0163] In some embodiments, the N2 sleep stage or N3 sleep stage is identified by detecting one or more changes in cortical-subcortical spectral coherence, wherein the changes are selected from: an increase in delta-cortical-subcortical spectral coherence compared to delta-cortical-subcortical spectral coherence when the subject is awake; and a decrease in β-cortical-subcortical spectral coherence compared to β-cortical-subcortical spectral coherence when the subject is awake. In some embodiments, the N2 sleep stage or N3 sleep stage is identified by combining one or more spectral power changes with the detection of one or more changes in cortical-subcortical spectral coherence.

[0164] In some embodiments, the pre-wake period or wakefulness period is identified by one or more spectral power variations selected from: a decrease in cortical delta power in the frequency range of 1 Hz to 4 Hz compared to the average cortical delta power during deep non-rapid eye movement (NREM) sleep; an increase in cortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power during deep NREM sleep; an increase in subcortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power during deep NREM sleep; and an increase in subcortical beta power in the frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power during deep NREM sleep. In some embodiments, the increase in subcortical beta power precedes the decrease in cortical delta power.

[0165] In some embodiments, the post-awakening time period is identified by one or more spectral power changes selected from: a decrease in cortical delta power in the frequency range of 1 Hz to 4 Hz compared to the average cortical delta power in the pre-awakening time period; an increase in cortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power in the pre-awakening time period; an increase in subcortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power in the pre-awakening time period; and an increase in subcortical beta power in the frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power in the pre-awakening time period.

[0166] In some embodiments, a computer-implemented method is provided for programming a DBS device to treat a subject's sleep dysfunction, the computer performing the following steps: a) receiving recorded electroencephalogram (EEG) signal data from a subcortical or cortical region of the subject's brain while the subject is sleeping; b) analyzing the recorded EEG signal data using a classification model that identifies patterns of electrical signals in the recorded EEG signal data associated with a sleep feature or sleep stage of interest; c) adjusting one or more programmed stimulation parameters based on the recorded EEG signal data according to an algorithmic control rule; and d) instructing the DBS device to apply electrical stimulation to a basal ganglia region or cortical region of the subject's brain during the sleep feature or sleep stage of interest to treat the subject's sleep dysfunction. See, for example, Example 1 and... Figure 3 .

[0167] Analyzing recorded EEG activity can include using algorithms or classifiers. In some implementations, machine learning algorithms are used to generate sleep feature or sleep stage classification models. Machine learning algorithms can include supervised learning algorithms. Examples of supervised learning algorithms can include Average Single Dependency Estimator (AODE), artificial neural networks (e.g., backpropagation), Bayesian statistics (e.g., I-Bayes classifier, Bayesian network, Bayesian knowledge base), case-based reasoning, decision trees, inductive logic programming, Gaussian process regression, grouping methods for data processing (GMDH), learning automata, learning vector quantization, minimum message length (decision tree, decision graph, etc.), lazy learning, instance-based learning nearest neighbor algorithms, analogy modeling, possibly approximating correct learning (PAC) learning, ripple pull-down rules, knowledge acquisition methods, symbolic machine learning algorithms, sub-symbolic machine learning methods, support vector machines (SVM), random forests, classifier ensembles, bootstrap aggregating (bagging), and boosting methods. Supervised learning can include ordered classification, such as regression analysis and information fuzzy networks (IFN). Alternatively, supervised learning methods can include statistical classification, such as AODE, linear classifiers (e.g., Fisher linear discriminant, logistic regression, Naive Bayes classifier, perceptron, and support vector machine), quadratic classifiers, k-nearest neighbors, boosting methods, decision trees (e.g., C4.5, random forest), Bayesian networks, and hidden Markov models.

[0168] Machine learning algorithms can also include unsupervised learning algorithms. Examples of unsupervised learning algorithms can include artificial neural networks (recursive or convolutional), data clustering, expectation-maximization algorithms, self-organizing maps, radial basis function networks, vector quantization, generative topology maps, information bottleneck methods, and IBSEAD. Unsupervised learning can also include association rule learning algorithms, such as the Apriori algorithm, the Eclat algorithm, and the FP-growth algorithm. Hierarchical clustering, such as single-link clustering and concept clustering, can also be used. Alternatively, unsupervised learning can include partitioned clustering, such as the K-means algorithm and fuzzy clustering.

[0169] In some cases, machine learning algorithms include reinforcement learning algorithms. Examples of reinforcement learning algorithms include, but are not limited to, temporal difference learning, Q-learning, and learning automata. Alternatively, machine learning algorithms may include data preprocessing. In some implementations, sleep feature or sleep stage classification models are trained by analyzing electroencephalogram (EEG) signal data recorded over multiple nights while the subject is sleeping.

[0170] In some implementations, the classification model identifies the N2 sleep stage or the N3 sleep stage by one or more spectral power variations selected from: a decrease in β power in the frequency range of 12 Hz to 30 Hz compared to the β power when the subject is awake; a decrease in γ power in the frequency range of 30 Hz to 60 Hz compared to the γ power when the subject is awake; an increase in θ power in the frequency range of 5 Hz to 10 Hz compared to the θ power when the subject is awake; and an increase in δ power in the frequency range of 0.5 Hz to 4.5 Hz compared to the δ power when the subject is awake.

[0171] In some implementations, the classification model identifies the N2 or N3 sleep stage by combining one or more spectral power variations with the detection of one or more variations in cortical-subcortical spectral coherence, wherein the one or more variations in cortical-subcortical spectral coherence are selected from: an increase in δ-cortical-subcortical spectral coherence compared to δ-cortical-subcortical spectral coherence when the subject is awake; and a decrease in β-cortical-subcortical spectral coherence compared to β-cortical-subcortical spectral coherence when the subject is awake.

[0172] In some embodiments, the classification model identifies pre-wake or wakefulness periods by one or more spectral power variations selected from: a decrease in cortical delta power in the frequency range of 1 Hz to 4 Hz compared to the average cortical delta power during deep non-rapid eye movement (NREM) sleep; an increase in cortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power during deep NREM sleep; an increase in subcortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power during deep NREM sleep; and an increase in subcortical beta power in the frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power during deep NREM sleep. In some embodiments, the increase in subcortical beta power precedes the decrease in cortical delta power.

[0173] In some implementations, the classification model identifies the post-awakening time period by one or more spectral power changes selected from: a decrease in cortical delta power in the frequency range of 1 Hz to 4 Hz compared to the average cortical delta power in the pre-awakening time period; an increase in cortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power in the pre-awakening time period; an increase in subcortical gamma power in the frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power in the pre-awakening time period; and an increase in subcortical beta power in the frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power in the pre-awakening time period.

[0174] In some implementations, the computer-implemented method further includes receiving accelerometer data of a subject during sleep; and using a classification model to analyze the accelerometer data in combination with the recorded electroencephalogram (EEG) signal data to identify sleep characteristics or sleep stages.

[0175] In some implementations, the computer-implemented method further includes receiving data from a non-invasive sleep monitoring device, a wearable sleep monitoring device, a photoplethysmography (PPG)-based sleep monitoring device, or a radar-based sleep monitoring device; and using a classification model to analyze the data to identify sleep characteristics or sleep stages.

[0176] In some implementations, the computer-implemented method further includes receiving autonomic nervous data of a subject during sleep; and using a classification model to analyze the autonomic nervous data in combination with the recorded electroencephalogram (EEG) signal data to identify sleep characteristics or sleep stages.

[0177] In some embodiments, the computer-implemented method further includes receiving an electroencephalogram (EEG), a recording of a neurostimulator electrode implanted under the galea aponeurotica or through a burr hole / lateral cranial implant, or a polysomnogram of a subject during sleep; and using a classification model to analyze the EEG, the recording of the neurostimulator electrode implanted through a burr hole / lateral cranial implant, or the polysomnogram to identify sleep characteristics or sleep stages.

[0178] In some implementations, the computer-implemented method further includes generating a sleep structure map.

[0179] In some implementations, the computer-implemented method further includes: a) ranking the predicted stimulation efficacy of available settings of the DBS device based on a classifier score of stimulation efficacy for each setting using a linear classification model; b) selecting the stimulation setting predicted to have the highest stimulation efficacy based on the linear classification model; c) receiving recorded electroencephalogram (EEG) signal data from the subcortical or cortical regions of the subject's brain after applying electrical stimulation to the basal ganglia or cortical regions of the subject's brain using the DBS device with the setting predicted to have the highest stimulation efficacy; d) analyzing the recorded EEG signal data to assess the subject's neural response to the electrical stimulation; e) based on the subject's response to the electrical stimulation... The neural response to the stimulus is used to update the linear classification model to generate an updated linear classification model; f) the updated linear classification model is used to update the ranking of predicted stimulus efficacy for available settings of the DBS device; g) the stimulus setting predicted to have the highest stimulus efficacy is selected based on the updated linear classification model; h) after applying electrical stimulation to the basal ganglia region or cortical region of the subject's brain using the DBS device based on the updated linear classification model and the setting predicted to have the highest stimulus efficacy, recorded EEG signal data are received from the subcortical or cortical regions of the subject's brain; and i) e)-h) are repeated to adjust the available settings of the DBS device to optimize stimulus efficacy. In some embodiments, the linear classification model uses linear discriminant analysis (LDA) to adjust the current amplitude and frequency of the electrical stimulation.

[0180] In some embodiments, the computer-implemented method further includes segmenting the recorded EEG signal data into consecutive time intervals. In some embodiments, each time interval comprises 0.5 seconds to 1 minute of recorded EEG signal data, including any amount of time within that range, such as 0.5, 0.75, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 15, 20, 25, 30, 35, 40, 45, 50, 55, or 60 seconds. In some embodiments, the computer-implemented method further includes assigning sleep characteristics or sleep stage labels (e.g., wakefulness, N1, N2, N3, and REM) to each time interval.

[0181] In some embodiments, the computer-implemented method further includes training a linear model by analyzing the recorded EEG signal data using a nonlinear model during all sleep stages while the subject is sleeping, to classify each time period as an N3 sleep stage or a non-N3 sleep stage. In some embodiments, canonical delta and beta power bands are used as feature inputs to train a linear classification model using linear discriminant analysis to classify each time period as an N3 sleep stage or a non-N3 sleep stage. In some embodiments, subcortical field potentials are used as feature inputs to train a linear classification model using linear discriminant analysis to classify each time period as an N3 sleep stage or a non-N3 sleep stage. In some embodiments, the stimulation amplitude is optimized during the N3 sleep stage to maximize slow-wave activity. In some embodiments, the slow-wave activity is in the frequency range of 0.5 Hz to 4 Hz.

[0182] In some implementations, the computer-implemented method further includes storing a user profile of the subject, which includes information about recorded electroencephalogram (EEG) signal data associated with sleep characteristics or sleep stages.

[0183] In some embodiments, the computer-implemented method further includes storing a user profile of the subject, which includes information about programmed stimulation parameters for applying electrical stimulation to the basal ganglia or cortical regions of the subject's brain based on recorded electroencephalogram (EEG) signal data to treat the subject's sleep dysfunction.

[0184] The methods described herein can be implemented in digital electronic circuits or in computer software, firmware, or hardware. The disclosed embodiments and other embodiments can be implemented as one or more computer program products, i.e., one or more modules of computer program instructions encoded on a computer-readable medium for execution by a data processing device or for controlling the operation of a data processing device. The computer-readable medium can be a machine-readable storage device, a machine-readable storage substrate, a memory device, a combination of materials that realize machine-readable propagated signals, or any combination thereof.

[0185] Computer programs (also known as programs, software, software applications, scripts, or code) can be written in any programming language, including compiled or interpreted languages; and can be deployed in any form, including as standalone programs or as modules, components, subroutines, or other units suited to a computing environment. A computer program does not necessarily correspond to a file in a file system. A program can be stored as part of a file containing other programs or data (e.g., one or more scripts stored in a markup language document), in a single file dedicated to the program in question, or in multiple collaborating files (e.g., a file storing one or more modules, subroutines, or code sections). A computer program can be deployed to execute on a single computer or on multiple computers located in one location or distributed across multiple locations and interconnected via a communication network.

[0186] In another aspect, as described, a system for performing the computer-implemented method may include a computer comprising a processor, storage components (i.e., memory), a display component, and other components typically found in general-purpose computers. The storage components store information accessible to the processor, including instructions executable by the processor and data that can be retrieved, manipulated, or stored by the processor.

[0187] Storage components can be any type capable of storing information accessible to the processor, such as hard drives, memory cards, ROM, RAM, DVDs, CD-ROMs, USB flash drives, writable memory, and read-only memory. The processor can be any well-known processor, such as those from Intel Corporation. Alternatively, the processor can be a dedicated controller, such as an ASIC.

[0188] Instructions can be any set of instructions to be executed directly by the processor (such as machine code) or indirectly (such as scripts). In this regard, the terms "instruction," "step," and "program" are used interchangeably. Instructions can be stored as object code for direct processing by the processor; or stored in any other computer language, including scripts or sets of independent source code modules that are interpreted or pre-compiled as needed.

[0189] Data can be retrieved, stored, or modified by the processor according to instructions. For example, although the system is not limited to any particular data structure, data can be stored in computer registers, as a table with multiple different fields and records, an XML document, or a flat file in a relational database. Data can also be formatted in any computer-readable format, such as, but not limited to, binary values, ASCII, or Unicode. Furthermore, data can include any information sufficient to identify the relevant information, such as numbers, descriptive text, proprietary code, pointers, references to data stored in other memory (including other network locations), or information used by functions to calculate the relevant data.

[0190] In some implementations, the processor and storage components may include multiple processors and storage components, which may or may not be stored in the same physical housing. For example, some instructions and data may be stored on a removable CD-ROM, and other instructions and data may be stored in a read-only computer chip. Some or all of the instructions and data may be stored in a location physically remote from the processor but still accessible to the processor. Similarly, the processor may include a collection of processors that can operate in parallel or not. In some implementations, a hardware accelerator is used.

[0191] In some implementations, cloud computing systems are used to perform the method. In these implementations, data files and programming can be exported to a cloud computer, which runs the program and returns the output to the user.

[0192] The components of a system for implementing the methods disclosed herein are further described in the following embodiments.

[0193] system

[0194] This disclosure also provides, for example, a system for practicing the subject methods. The system may be an open-loop or closed-loop system configured to perform the methods provided herein. In some embodiments, the system may include a DBS electrode adapted to be placed in a location in the basal ganglia region (e.g., the subthalamic nucleus region, the globus pallidus region, or the thalamus region) or cortical region of the subject's brain to deliver electrical stimulation to the basal ganglia region or cortical region; and a detection electrode adapted to be placed in a subcortical or cortical region (e.g., the precentral gyrus region or the postcentral gyrus region) of the subject's brain to record electroencephalogram (EEG) signal data while the subject is sleeping, before, during, or after the application of electrical stimulation to the brain. In a closed-loop system, the system may further include a computing device and a control unit programmed to, when an EEG signal associated with a selected sleep feature or sleep stage is detected using a second electrode, instruct the DBS electrode to apply electrical stimulation to the basal ganglia or cortical region of the subject's brain in a manner that effectively treats the subject's sleep dysfunction; to analyze the recorded EEG signal data using a sleep feature or sleep stage classification model that identifies patterns of electrical signals in the recorded EEG signal data associated with the selected sleep feature or sleep stage; to adjust one or more programmed stimulation parameters based on the recorded EEG signal data according to an algorithmic control rule; and to automatically deliver electrical stimulation to the basal ganglia or cortical region of the subject's brain via the control unit, the neurostimulator pulse generator, and the DBS electrode in a manner that effectively treats the sleep dysfunction if the electrical signal measurement indicates that the patient requires treatment. In some embodiments, frequency variations are introduced within one hemisphere to create a frequency difference between the two sides, and endogenous brain rhythms are induced / enhanced at the frequency difference between the two stimulation frequencies. In some implementations, the stimulation intervention may take the form of auditory stimulation or non-invasive stimulation, including transcranial electrical stimulation or transcranial magnetic stimulation. In some implementations, the N2 sleep stage or N3 sleep stage is identified by attenuation of β power in the frequency range of 12 Hz to 30 Hz, attenuation of γ power in the frequency range of 30 Hz to 60 Hz, increase of low-frequency θ power in the frequency range of 5 Hz to 10 Hz, and increase of δ power in the frequency range of 0.5 Hz to 4.5 Hz. In some implementations, one or more programmed stimulation parameters are adjusted based on recorded electrical activity data according to the control law of the algorithm, and the adjusted electrical stimulation is delivered to the brain via a control unit, a pulse generator, and DBS electrodes in a manner that effectively treats sleep dysfunction in selected sleep characteristics or sleep stages.The closed-loop system may include an on-body pulse generator connected to an implanted DBS electrode, which can therefore automatically apply electrical stimulation to the brain upon receiving communication from a control unit; or a cranially mounted neurostimulator that can also sense cortical nerve signals via electrodes mounted on the housing of the device.

[0195] The processor of the closed-loop system can run programming for evaluating the efficacy of treatment and adjust treatment parameters as needed without user intervention. Therefore, the closed-loop system may not necessarily include a user interface for the user to instruct DBS electrodes to apply electrical stimulation to the brain to treat the subject's sleep dysfunction. However, in some embodiments, a user interface may be included in the closed-loop system, which can be used to confirm, override, or modify the recommendations of the closed-loop system.

[0196] In some aspects, the control algorithm for the methods and systems of this disclosure may include a step of comparing electrical signals from a region of a subject's brain with normal electrical signals or reference electrical signals (e.g., normal sleep, generally without sleep dysfunction), wherein when the electrical signals are significantly different from the normal electrical signals or reference electrical signals, the control algorithm includes a step of guiding a device to apply electrical stimulation to the subject's brain, subsequently measuring the electrical signals from the brain region and comparing them with normal electrical signals or reference electrical signals, wherein when the measured signals are significantly different from the normal electrical signals or reference electrical signals, the algorithm includes a step of applying another electrical stimulation to the brain.

[0197] In some implementations, the control algorithm utilizes machine learning algorithms to analyze the input EEG activity data to automate the detection of brain activity features that distinguish sleep stages. If the brain activity features indicate that the subject is in a sleep stage that should be treated with electrical stimulation, the control algorithm then instructs the device to apply electrical stimulation to the subject's brain. For example, machine learning algorithms can be used to correlate the level of total power or power in a specific frequency range (e.g., α, δ, β, γ, and / or θ) with sleep features or sleep stages that should be treated with deep brain stimulation. In some implementations, N2 sleep stages or N3 sleep stages are identified by attenuation of β power in the frequency range of 12 Hz to 30 Hz, attenuation of γ power in the frequency range of 30 Hz to 60 Hz, increase of low-frequency θ power in the frequency range of 5 Hz to 10 Hz, and increase of δ power in the frequency range of 0.5 Hz to 4.5 Hz. In some implementations, field potential data is fitted to a sleep trait or sleep stage classification model to determine how to adjust one or more programmed stimulation parameters, including physiologically relevant events such as sleep stages, slow waves, spindle waves, wakefulness, and sleep stage transitions. In some implementations, the algorithm provides clinicians with updated optimal stimulation settings recommendations to guide programming and decision-making.

[0198] In some embodiments, the system further includes a user interface comprising an input electrically coupled to the processor for instructing the DBS electrodes to apply electrical stimulation to the basal ganglia region or cortical region to treat a subject's sleep dysfunction. In some embodiments, the user interface is password protected and operable by a healthcare practitioner.

[0199] In some implementations, the system further includes an accelerometer for recording the subject's movements while the subject is sleeping. Accelerometer data can be combined with electroencephalogram (EEG) data to aid in the classification of sleep characteristics or sleep stages.

[0200] In some implementations, the system further includes a non-invasive sleep monitoring device, a wearable sleep monitoring device (e.g., a smart ring, smartwatch, wristband, or headband sleep tracker), a photoplethysmography (PPG)-based sleep monitoring device, or a radar-based sleep monitoring device. Data from such devices can be used to assist in the classification of sleep characteristics or sleep stages. For descriptions of such sleep monitoring devices, see, for example, Toften et al. (2020) SleepMed.75:54-61, Kwon et al. (2021) IEEE J Biomed Health Inform.25(10):3844-3853, Lauteslager et al. (2020) Annu Int Conf IEEE Eng Med Biol Soc.2020:5150-5153, Chung et al. (2017) Sensors (Basel) 17(7):1685, An et al. (2022) Sci Rep. 12(1):21052, Chinoy et al. (2021) Sleep 44(5):zsaa291, Chinoy et al. (2022) Nat Sci Sleep14:493-516, Shelgikar et al. (2016) Chest 150(3):732-43, Zhao et al. (2021) Entropy (Basel) 23(1):116; incorporated herein by reference in its entirety.

[0201] The components of a system for implementing the methods disclosed herein are further described in the following embodiments.

[0202] Pharmacological drug administration

[0203] Implementations of the methods and systems provided in this disclosure may also include administering an effective amount of at least one pharmacological agent. "Effective amount" means a dose sufficient to treat the sleep dysfunction in a subject as needed. In some implementations, the sleep dysfunction is caused by a motor disorder or neurological condition. The effective amount will vary between subjects and may depend on factors such as the subject's age and physical condition, the type of motor disorder or neurological condition causing the sleep dysfunction, the severity of the sleep dysfunction being treated, the duration of treatment, the nature of any concurrent treatments, the form of the agent, the pharmaceutically acceptable carrier used, if any, the route and method of delivery, and similar factors within the knowledge and skill of those skilled in the art. An appropriate dose may be determined according to conventional pharmacological procedures known to those skilled in the art, as described in more detail below.

[0204] If a pharmacological approach is used in the treatment of movement disorders or neurological conditions, the specific properties of the drug and the administration regimen will vary depending on the specific nature of the condition being treated. Representative pharmacological agents that can be used to treat Parkinson's disease include, but are not limited to, L-DOPA (l-3,4-dihydroxyphenylalanine, also known as levodopa); carbidopa (N-amino-α-methyl-3-hydroxy-L-tyrosine monohydrate); carbidopa-levodopa (Rytary, Sinemet, Duopa); dopamine agonists, including but not limited to pramipexole (Mirapex ER), rotigotine, apomorphine (Apokyn), and amantadine (Gocovri); monoamine oxidase B (MAO-B) inhibitors, including but not limited to selegiline (Zelapar), rasagiline (Azilect), and safamide (Xadago); catechol O-methyltransferase (COMT) inhibitors, including but not limited to entacapone (Comtan), octopone (Ongentys), and tocapone (Tasmar); anticholinergic agents, including but not limited to benzalkonium chloride (Cogentin) and trihexyphenidyl; and adenosine receptor antagonists, including but not limited to A2A. Receptor antagonists such as itratheline (Nourianz); or antipsychotic drugs, including but not limited to nuplazid (Pimavanserin), or any combination thereof.

[0205] In some respects, the administration of a pharmacological agent involves the use of a pharmacological delivery device, such as, but not limited to, a pump (implantable or external), an epidural syringe, a syringe or other injection device, a catheter, and / or a reservoir operatively associated with a catheter. For example, in some embodiments, a delivery device for delivering at least one pharmacological agent to a subject may be a pump, syringe, catheter, or a reservoir operatively associated with a connection device such as a catheter, tubing, etc. Containers suitable for delivering at least one pharmacological agent to a pharmacological agent administration device include containers that can be used to deliver, place, attach, and / or insert at least one pharmacological agent into a delivery device for administering the pharmacological agent to a subject, and include, but are not limited to, vials, ampoules, tubes, capsules, bottles, syringes, and bags. The administration of a pharmacological agent may be performed by a user or by a closed-loop system.

[0206] practicality

[0207] The methods and systems disclosed herein can be used to treat sleep dysfunction using nocturnal deep brain stimulation (DBS). Compared to conventional daytime DBS techniques, closed-loop stimulation can be finely targeted and modulated in a personalized manner to achieve more reliable and / or more effective relief of sleep dysfunction at selected sleep stages.

[0208] In some cases, sleep disorders are caused by movement disorders, such as, but not limited to, Parkinson's disease, Parkinsonian syndrome, progressive supranuclear palsy, ataxia, cervical dystonia, chorea, dystonia, functional movement disorders, Huntington's disease, multiple system atrophy, myoclonus, tardive dyskinesia, Tourette syndrome, tremor, restless legs syndrome, and Wilson's disease. Symptoms may include, but are not limited to, tremor, involuntary movements, bradykinesia, rigidity, postural instability, twisting movements, poor balance, irregular movements, tripping, and difficulty walking. In some cases, movement disorders are caused by genetic and / or environmental factors, head trauma, infection, inflammation, metabolic disorders, toxins, adverse drug reactions, or stressful life events.

[0209] In some cases, sleep disorders are caused by neurological conditions such as, but not limited to, neurodegenerative diseases including Alzheimer's disease, Parkinson's disease, Huntington's disease, multiple system atrophy, or Lewy body dementia; and multiple system atrophy, epilepsy, stroke, bipolar disorder; neuromuscular disorders including amyotrophic lateral sclerosis (ALS), peroneal muscular atrophy (CMT), chronic inflammatory demyelinating polyneuropathy (CIDP), Guillain-Barré syndrome (GBS), Lambert-Eton syndrome, muscular dystrophy, myasthenia gravis, myopathy, and peripheral neuropathy.

[0210] In some implementations, sleep dysfunction is caused by stroke. Insomnia may occur after a stroke, particularly in patients with a right hemisphere stroke or a stroke in the thalamus or brainstem, which includes the pontine tegmentum and thalamus-midbrain region. Somnolence may occur after a stroke in patients with subcortical (caudate nucleus, putamen), superior pons, medial pontomedium-medulla oblongata, or cortical strokes affecting the reticular activating system (RAS). Paramedian or bilateral thalamic strokes may initially cause coma, followed by somnolence upon awakening. Supratentorial strokes may reduce non-REM sleep, total sleep time, and ipsilateral or bilateral sleep spindles. Sawtooth waves may be reduced after a hemispherical stroke. REM sleep may be reduced after an occipital lobe stroke. Strokes in the pontomedinostomy junction and raphe nuclei may reduce the amount of non-REM sleep. Strokes in the inferior pons may selectively reduce REM sleep. Strokes in the paramedian thalamus and lower pons may reduce slow-wave sleep.

[0211] The efficacy of treatment for patients with sleep disorders can be measured in a manner acceptable in the art, such as by using the Visual Analogue Scale (VAS), Likert Scale, Stanford Sleepiness Scale (SSS), Maintain Wakefulness Test (MWT), Epworth Sleepiness Scale (ESS), Multiple Sleep Latency Test (MSLT), or Athens Insomnia Scale. In some implementations, assessing the efficacy of treatment for a subject's sleep disorder includes monitoring the subject using a motion recorder, electroencephalogram (EEG), or polysomnography.

[0212] Examples of non-limiting aspects of this disclosure

[0213] The various aspects of the subject matter of the invention described above, including embodiments, may be advantageous individually or in combination with one or more other aspects or embodiments. Without limiting the foregoing description, certain non-limiting aspects of this disclosure, numbered 1-96, are provided below. As will be apparent to those skilled in the art upon reading this disclosure, each of the individually numbered aspects may be used or combined with any of the preceding or following individually numbered aspects. This is intended to support all such combinations of aspects and is not limited to the combinations of aspects explicitly provided below.

[0214] 1. A method for treating sleep dysfunction in a subject, the method comprising:

[0215] The first electrode is placed at a first location in the basal ganglia region or cortical region of the subject's brain to deliver electrical stimulation to the basal ganglia region or the cortical region;

[0216] The second electrode is placed at a second location in the subcortical or cortical region of the subject's brain to record electroencephalogram (EEG) signal data while the subject is sleeping.

[0217] The second electrode is used to detect electroencephalogram (EEG) signals associated with sleep features or sleep stages of interest; and

[0218] When the second electrode detects an EEG signal associated with the sleep feature or sleep stage of interest, electrical stimulation is applied to the basal ganglia region or cortical region of the subject's brain using the first electrode in a manner that effectively treats the subject's sleep dysfunction.

[0219] 2. The method according to aspect 1, wherein the electroencephalogram (EEG) signal data includes field potential data.

[0220] 3. The method according to aspect 1 or 2, wherein the basal ganglia region is the subthalamic nucleus region, the globus pallidus region, or the thalamic region.

[0221] 4. The method according to any one of aspects 1-3, wherein the cortical region is the precentral gyrus region or the postcentral gyrus region.

[0222] 5. The method according to any one of aspects 1-4, wherein the sleep stage of interest is N2, N3 or REM.

[0223] 6. The method according to any one of aspects 1-5, the method further comprising using an accelerometer in conjunction with the electroencephalogram signals to identify the sleep feature or the sleep stage of interest.

[0224] 7. The method according to any one of aspects 1-6, the method further comprising using autonomic nervous data in conjunction with the electroencephalogram (EEG) signals to identify the sleep feature or sleep stage of interest.

[0225] 8. The method according to any one of aspects 1-7, the method further comprising using electroencephalography, polysomnography, noninvasive sleep monitoring device, wearable sleep monitoring device, photoplethysmography (PPG) based sleep monitoring device or radar-based sleep monitoring device to identify the sleep feature or sleep stage of interest.

[0226] 9. The method according to any one of aspects 1-8, the method further comprising generating a sleep structure map.

[0227] 10. The method according to any one of aspects 1-9, the method further comprising, when the electroencephalogram (EEG) signal associated with the sleep feature of interest or the sleep stage is detected, using a control algorithm to automate the application of electrical stimulation.

[0228] 11. The method according to aspect 10, wherein the control algorithm uses a machine learning algorithm for sleep feature or sleep stage classification.

[0229] 12. The method according to aspect 11, wherein the machine learning algorithm is a supervised machine learning algorithm.

[0230] 13. The method according to any one of aspects 10-12, wherein the control algorithm further adjusts one or more programmed stimulation parameters to maximize slow-wave activity.

[0231] 14. The method according to aspect 13, wherein the slow wave activity is in the frequency range of 0.5 Hz to 4 Hz.

[0232] 15. The method according to any one of aspects 10-14, wherein the control algorithm further uses linear discriminant analysis (LDA) to adjust the stimulation amplitude or frequency of the electrical stimulation.

[0233] 16. The method according to aspect 15, wherein the stimulation amplitude is optimized during the N3 sleep stage to maximize slow-wave activity.

[0234] 17. The method according to any one of aspects 1-16, wherein the electrical stimulation is applied unilaterally or bilaterally.

[0235] 18. The method according to any one of aspects 1-17, wherein the electroencephalogram (EEG) signal comprises neural oscillations of beta frequency, gamma frequency, delta frequency, or theta frequency.

[0236] 19. The method according to any one of aspects 5-18, wherein the N3 sleep stage is identified by an increase in δ power during the N3 sleep stage compared to when the subject is awake.

[0237] 20. The method according to any one of aspects 5-19, wherein the N2 sleep stage or the N3 sleep stage is identified by one or more spectral power variations selected from: a decrease in β power in a frequency range of 12 Hz to 30 Hz compared to β power when the subject is awake; a decrease in γ power in a frequency range of 30 Hz to 60 Hz compared to γ ​​power when the subject is awake; an increase in θ power in a frequency range of 5 Hz to 10 Hz compared to θ power when the subject is awake; and an increase in δ power in a frequency range of 0.5 Hz to 4.5 Hz compared to δ power when the subject is awake.

[0238] 21. The method according to aspect 20, wherein the N2 sleep stage or the N3 sleep stage is identified by means of the one or more spectral power changes combined with the detection of one or more changes in cortical-subcortical spectral coherence, the one or more changes in cortical-subcortical spectral coherence being selected from: an increase in the δ-cortical-subcortical spectral coherence compared to the δ-cortical-subcortical spectral coherence when the subject is awake; and a decrease in the β-cortical-subcortical spectral coherence compared to the β-cortical-subcortical spectral coherence when the subject is awake.

[0239] 22. The method according to any one of aspects 1-21, wherein the second electrode is placed on the surface of the subcortical region or the cortical region.

[0240] 23. The method according to any one of aspects 1-22, wherein the first electrode is a non-brain-penetrating surface electrode array or a brain-penetrating electrode array.

[0241] 24. The method according to any one of aspects 1-23, wherein the second electrode is a non-brain-penetrating surface electrode array or a brain-penetrating electrode array.

[0242] 25. The method according to any one of aspects 1-24, wherein the second electrode is an electroencephalogram (EEG) electrode array, a neurostimulator electrode installed under the galea aponeurotica or through a burr hole in the skull or installed on the cranial side, or an electrocorticogram (ECoG) electrode array.

[0243] 26. The method according to aspect 25, wherein the ECoG electrode array spans the anterior central gyrus and the posterior central gyrus.

[0244] 27. The method according to any one of aspects 1-26, wherein the sleep dysfunction is caused by a motor disorder or a neurological condition, wherein the application of the electrical stimulation improves sleep.

[0245] 28. The method according to aspect 27, wherein the movement disorder is Parkinson's disease.

[0246] 29. The method according to aspect 27 or 28, wherein the subject is further subjected to daytime neural stimulation.

[0247] 30. The method according to aspect 28 or 29, wherein the subject is further administered a dopaminergic drug.

[0248] 31. The method according to any one of aspects 1-30, the method further comprising evaluating the efficacy of treatment for the sleep dysfunction of the subject.

[0249] 32. The method according to aspect 31, wherein the assessment includes using the Visual Analogue Scale (VAS), Likert Scale, Stanford Somnolence Scale (SSS), Maintain Wakefulness Test (MWT), Appleworth Somnolence Scale (ESS), Multiple Sleep Latency Test (MSLT) or Athens Insomnia Scale.

[0250] 33. The method according to aspect 31 or 32, wherein the assessment includes monitoring the subject using a motion recorder, electroencephalogram, or polysomnography.

[0251] 34. The method according to any one of aspects 1-33, the method further comprising mapping the subject's brain to identify optimal locations in the subcortical region or the cortical region to detect the electroencephalogram (EEG) signals associated with the sleep characteristics or the sleep stage.

[0252] 35. The method according to aspect 34, wherein the cortical region is the precentral gyrus region or the postcentral gyrus region.

[0253] 36. The method according to any one of aspects 1-35, the method further comprising segmenting the recorded electroencephalogram (EEG) signal data into consecutive time periods.

[0254] 37. The method according to aspect 36, the method further comprising assigning sleep characteristics or sleep stage labels to each time period.

[0255] 38. The method according to aspect 36 or 37, wherein each time interval includes 0.5 seconds to 1 minute of the recorded EEG signal data.

[0256] 39. The method according to any one of aspects 1-38, wherein the method is performed while the subject is sleeping at home, in a sleep laboratory, or in a hospital.

[0257] 40. The method according to any one of aspects 1-39, wherein the sleep stage is N1, N2, N3 or phase- or tonic-motor rapid eye movement (REM).

[0258] 41. The method according to any one of aspects 1-40, wherein the sleep characteristics are slow waves, sleep spindle waves, K complex waves, beta bursts, pre-wake period, wake period, post-wake period, or sleep stage transitions.

[0259] 42. The method according to aspect 41, wherein the pre-wake period or the wake period is identified by one or more spectral power variations selected from: a decrease in cortical delta power in a frequency range of 1 Hz to 4 Hz compared to the average cortical delta power during deep non-rapid eye movement (NREM) sleep; an increase in cortical gamma power in a frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power during deep NREM sleep; an increase in subcortical gamma power in a frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power during deep NREM sleep; and an increase in subcortical beta power in a frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power during deep NREM sleep.

[0260] 43. The method according to aspect 42, wherein the increase in subcortical β power precedes the decrease in cortical δ power.

[0261] 44. The method according to any one of aspects 41-43, wherein the post-awakening time period is identified by one or more spectral power variations selected from: a decrease in cortical delta power in a frequency range of 1 Hz to 4 Hz compared to the average cortical delta power in the pre-awakening time period; an increase in cortical gamma power in a frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power in the pre-awakening time period; an increase in subcortical gamma power in a frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power in the pre-awakening time period; and an increase in subcortical beta power in a frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power in the pre-awakening time period.

[0262] 45. The method according to any one of aspects 1-44, wherein the electrical stimulation increases cortical delta power, decreases cortical alpha power, decreases cortical beta power, and decreases cortical σ power.

[0263] 46. ​​The method according to any one of aspects 1-45, wherein the electrical stimulation reduces cortical-subcortical σ-spectral coherence.

[0264] 47. The method according to any one of aspects 1-46, the method further comprising:

[0265] The second electrode is used to detect electroencephalogram (EEG) signals associated with one or more additional sleep features or sleep stages of interest; and

[0266] When the second electrode detects the EEG signal associated with one or more additional sleep features or sleep stages of interest, the first electrode is used to apply electrical stimulation to the basal ganglia region or the cortical region of the subject's brain in a manner that effectively treats the subject's sleep dysfunction.

[0267] 48. A computer-implemented method for programming a deep brain stimulation (DBS) device to treat sleep dysfunction in a subject, the computer performing the following steps:

[0268] a) While the subject is sleeping, receive recorded electroencephalogram (EEG) signal data from the subcortical or cortical regions of the subject's brain;

[0269] b) Use a classification model to analyze the recorded EEG signal data, the classification model identifying patterns of electrical signals in the recorded EEG signal data that are associated with sleep features or sleep stages of interest;

[0270] c) Adjust one or more programmed stimulation parameters based on the recorded EEG signal data according to the algorithmic control rules; and

[0271] d) When the sleep feature or sleep stage of interest is detected, instruct the DBS device to apply electrical stimulation to the basal ganglia region or cortical region of the subject's brain to treat the subject's sleep dysfunction.

[0272] 49. The computer-implemented method according to aspect 48, wherein the electroencephalogram (EEG) signal data includes field potential data.

[0273] 50. The computer-implemented method according to aspect 48 or 49, wherein a machine learning algorithm is used to generate the classification model.

[0274] 51. The computer-implemented method according to aspect 50, wherein the machine learning algorithm is a supervised machine learning algorithm.

[0275] 52. The computer-implemented method according to any one of aspects 48-51, wherein the basal ganglia region is the subthalamic nucleus (STN) region, the globus pallidus region, or the thalamic region.

[0276] 53. The computer-implemented method according to any one of aspects 48-52, wherein the cortical region is the precentral gyrus or the postcentral gyrus.

[0277] 54. The computer-implemented method according to any one of aspects 48-53, wherein the sleep phase is N2, N3 or REM.

[0278] 55. The computer-implemented method according to any one of aspects 48-54, the computer-implemented method further comprising:

[0279] Accelerometer data of the subject are received while the subject is sleeping; and

[0280] The classification model is used to analyze the accelerometer data combined with the recorded electroencephalogram (EEG) signal data to identify the sleep characteristics or sleep stages.

[0281] 56. The computer-implemented method according to any one of aspects 48-55, the computer-implemented method further comprising:

[0282] The subject's autonomic nervous data is received while the subject is sleeping; and

[0283] The classification model is used to analyze the autonomic nervous data combined with the recorded electroencephalogram (EEG) signal data to identify the sleep characteristics or sleep stages.

[0284] 57. The computer-implemented method according to any one of aspects 48-56, the computer-implemented method further comprising:

[0285] Receive the subject's electroencephalogram (EEG) or polysomnography while the subject is sleeping; and

[0286] The classification model is used to analyze the electroencephalogram (EEG) or polysomnography (PSG) combined with the recorded EEG signal data to identify the sleep characteristics or sleep stages.

[0287] 58. The computer-implemented method according to any one of aspects 48-57, the computer-implemented method further comprising receiving data from a non-invasive sleep monitoring device, a wearable sleep monitoring device, a photoplethysmography (PPG)-based sleep monitoring device, or a radar-based sleep monitoring device; and using the classification model to analyze the data to identify the sleep characteristics or the sleep stages.

[0288] 59. The computer-implemented method according to any one of aspects 48-58, the computer-implemented method further comprising generating a sleep structure diagram.

[0289] 60. The computer-implemented method according to any one of aspects 48-59, wherein the classification model is trained to identify the sleep characteristics or the sleep stages by analyzing electroencephalogram (EEG) signal data recorded over multiple nights while the subject is sleeping.

[0290] 61. The computer-implemented method according to any one of aspects 48-60, the computer-implemented method further comprising:

[0291] a) Using a linear classification model, the predicted stimulus efficacy of the available settings of the DBS device is ranked based on the classifier score of the stimulus efficacy for each setting.

[0292] b) Select the stimulus setting predicted to have the highest stimulus efficacy based on the linear classification model;

[0293] c) After applying electrical stimulation to the basal ganglia region or the cortical region of the subject's brain using the DBS device in the setting predicted to have the highest stimulating efficacy, receiving the recorded electroencephalogram (EEG) signal data from the subcortical region or the cortical region of the subject's brain.

[0294] d) Analyze the recorded electroencephalogram (EEG) data to assess the subject's neural response to the electrical stimulation;

[0295] e) Update the linear classification model based on the subject's neural response to the electrical stimulation to generate an updated linear classification model;

[0296] f) Use the updated linear classification model to update the ranking of the predicted stimulus efficacy for the available settings of the DBS device;

[0297] g) Select the stimulus setting predicted to have the highest stimulus efficacy based on the updated linear classification model;

[0298] h) After applying the electrical stimulation to the basal ganglia region or the cortical region of the subject's brain using the DBS device based on the updated linear classification model and the setting predicted to have the highest stimulation efficacy, receiving the recorded electroencephalogram (EEG) signal data from the subcortical region or the cortical region of the subject's brain; and

[0299] i) Repeat e)-h) to adjust the available settings of the DBS device to optimize stimulation efficacy.

[0300] 62. The computer-implemented method according to aspect 61, wherein the linear classification model uses linear discriminant analysis (LDA) to adjust the current amplitude and frequency of the electrical stimulation.

[0301] 63. The computer-implemented method according to aspect 62, wherein the stimulus amplitude is optimized during the N3 sleep stage to maximize slow-wave activity.

[0302] 64. The computer-implemented method according to aspect 63, wherein the slow-wave activity is in the frequency range of 0.5 Hz to 4 Hz.

[0303] 65. The computer-implemented method according to any one of aspects 48-64, the computer-implemented method further comprising segmenting the recorded electroencephalogram (EEG) signal data into consecutive time periods.

[0304] 66. The computer-implemented method according to aspect 65, the computer-implemented method further comprising assigning sleep characteristics or sleep stage tags to each time period.

[0305] 67. The computer-implemented method according to aspect 65, wherein each time period includes 0.5 seconds to 1 minute of recorded EEG signal data.

[0306] 68. The computer-implemented method according to any one of aspects 61-67, the computer-implemented method further comprising training a linear model by analyzing the recorded electroencephalogram (EEG) signal data using a nonlinear model during all sleep stages while the subject is sleeping, to classify each time period as an N3 sleep stage period or a non-N3 sleep stage period.

[0307] 69. The computer-implemented method according to aspect 68, wherein the standardized δ and β power frequency bands are used as feature inputs to train the linear classification model using linear discriminant analysis to classify each time period as an N3 sleep stage period or a non-N3 sleep stage period.

[0308] 70. The computer-implemented method according to aspect 68, wherein subcortical field potentials are used as feature inputs to train the linear classification model using linear discriminant analysis to classify each time period as an N3 sleep stage period or a non-N3 sleep stage period.

[0309] 71. The computer-implemented method according to any one of aspects 48-70, wherein the electroencephalogram (EEG) signal data includes field potential data.

[0310] 72. The computer-implemented method according to any one of aspects 48-71, the computer-implemented method further comprising storing a user profile of the subject, the user profile including information about recorded electroencephalogram (EEG) signal data associated with the sleep characteristics or the sleep stage.

[0311] 73. A computer-implemented method according to any one of aspects 48-72, the computer-implemented method further comprising storing a user profile of the subject, the user profile including information about the programmed stimulation parameters for applying electrical stimulation to the basal ganglia region or the cortical region of the subject's brain based on recorded electroencephalogram (EEG) signal data to treat the subject's sleep dysfunction.

[0312] 74. A computer-implemented method according to any one of aspects 48-73, wherein the classification model identifies the N2 sleep stage or the N3 sleep stage by one or more spectral power variations, the one or more spectral power variations being selected from: a decrease in β power in a frequency range of 12 Hz to 30 Hz compared to β power when the subject is awake; a decrease in γ power in a frequency range of 30 Hz to 60 Hz compared to γ ​​power when the subject is awake; an increase in θ power in a frequency range of 5 Hz to 10 Hz compared to θ power when the subject is awake; and an increase in δ power in a frequency range of 0.5 Hz to 4.5 Hz compared to δ power when the subject is awake.

[0313] 75. The computer-implemented method according to aspect 74, wherein the classification model identifies the N2 sleep stage or the N3 sleep stage by combining the detection of one or more variations in cortical-subcortical spectral coherence with the one or more variations in cortical-subcortical spectral coherence, wherein the one or more variations in cortical-subcortical spectral coherence are selected from: an increase in the δ-cortical-subcortical spectral coherence compared to the δ-cortical-subcortical spectral coherence when the subject is awake; and a decrease in the β-cortical-subcortical spectral coherence compared to the β-cortical-subcortical spectral coherence when the subject is awake.

[0314] 76. The computer-implemented method according to any one of aspects 48-75, wherein the sleep characteristics are slow waves, sleep spindle waves, K complex waves, beta bursts, pre-wake periods, wake periods, post-wake periods, or sleep stage transitions.

[0315] 77. The computer-implemented method according to aspect 76, wherein the classification model identifies the pre-wake period or the wake period by one or more spectral power variations, the one or more spectral power variations being selected from: a decrease in cortical delta power in a frequency range of 1 Hz to 4 Hz compared to the average cortical delta power during deep non-rapid eye movement (NREM) sleep; an increase in cortical gamma power in a frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power during deep NREM sleep; an increase in subcortical gamma power in a frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power during deep NREM sleep; and an increase in subcortical beta power in a frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power during deep NREM sleep.

[0316] 78. The computer-implemented method according to aspect 77, wherein the increase in subcortical β power precedes the decrease in cortical δ power.

[0317] 79. A computer-implemented method according to any one of aspects 76-78, wherein the classification model identifies the post-awakening time period by one or more spectral power variations, the one or more spectral power variations being selected from: a decrease in cortical delta power in a frequency range of 1 Hz to 4 Hz compared to the average cortical delta power in the pre-awakening time period; an increase in cortical gamma power in a frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power in the pre-awakening time period; an increase in subcortical gamma power in a frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power in the pre-awakening time period; and an increase in subcortical beta power in a frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power in the pre-awakening time period.

[0318] 80. A non-transitory computer-readable medium comprising program instructions that, when executed by a processor in a computer, cause the processor to perform the method according to any one of aspects 48-79.

[0319] 81. A kit comprising a non-transitory computer-readable medium according to aspect 80 and instructions for using a deep brain stimulation device to treat a subject's sleep dysfunction.

[0320] 82. A system for treating sleep dysfunction in a subject, the system comprising:

[0321] A first electrode, adapted to be placed in a location in the basal ganglia region or cortical region of the subject's brain, to deliver electrical stimulation to the basal ganglia region or the cortical region;

[0322] A second electrode, adapted to be placed in a subcortical or cortical region of the subject's brain, to record electroencephalogram (EEG) signal data while the subject is sleeping; and

[0323] A processor, programmed according to a computer-implemented method according to any one of aspects 48-79, is configured to instruct the first electrode to apply electrical stimulation to the basal ganglia region or the cortical region of the subject's brain in a manner that effectively treats the subject's sleep dysfunction when an electroencephalogram (EEG) signal associated with the sleep feature or sleep stage of interest is detected using the second electrode.

[0324] 83. The system according to aspect 82, wherein the electroencephalogram (EEG) signal data includes field potential data.

[0325] 84. The system according to aspect 82 or 83, wherein the basal ganglia region is the subthalamic nucleus region, the globus pallidus region, or the thalamic region.

[0326] 85. The system according to any one of aspects 82-84, wherein the cortical region is the precentral gyrus region or the postcentral gyrus region.

[0327] 86. The system according to any one of aspects 82-85, wherein the sleep stage of interest is N2, N3 or REM.

[0328] 87. The system according to any one of aspects 82-86, the system further comprising an accelerometer for recording the motion of the subject while the subject is sleeping.

[0329] 88. The system according to any one of aspects 82-87, the system further comprising a non-invasive sleep monitoring device, a wearable sleep monitoring device, a sleep monitoring device based on photoplethysmography (PPG), or a radar-based sleep monitoring device.

[0330] 89. The system according to any one of aspects 82-88, wherein the first electrode is a non-brain-penetrating surface electrode array or a brain-penetrating electrode array.

[0331] 90. The system according to any one of aspects 82-89, wherein the second electrode is a non-brain-penetrating surface electrode array or a brain-penetrating electrode array.

[0332] 91. The system according to any one of aspects 82-90, wherein the second electrode is an electroencephalogram (EEG) electrode array, a neurostimulator electrode installed under the galea aponeurotica or through a burr hole in the skull or on the cranial side, or an electrocorticography (ECoG) electrode array.

[0333] 92. The system according to aspect 91, wherein the ECoG electrode array spans the anterior central gyrus and the posterior central gyrus.

[0334] 93. The system according to any one of aspects 82-92, wherein the sleep dysfunction is caused by a motor disorder or a neurological condition, wherein the application of the electrical stimulation improves sleep.

[0335] 94. The system according to aspect 93, wherein the movement disorder is Parkinson's disease.

[0336] 95. The system according to any one of aspects 82-94, wherein the system further includes a user interface, the user interface including an input terminal electrically coupled to the processor for instructing the first electrode to apply electrical stimulation to the basal ganglia region or the cortical region to treat the sleep dysfunction of the subject.

[0337] 96. The system according to aspect 95, wherein the user interface is password protected and can be operated by a healthcare practitioner.

[0338] It will be readily apparent to those skilled in the art that various changes and modifications can be made without departing from the spirit or scope of the invention.

[0339] Experimental

[0340] The following embodiments are provided to provide a complete disclosure and description of how to prepare and use the invention to those skilled in the art, and are not intended to limit the scope of what the inventors consider to be their invention, nor to represent that the following experiments are all or only the experiments performed. Efforts have been made to ensure the accuracy of the figures used (e.g., quantities, temperatures, etc.), but some experimental errors and biases should be taken into account. Unless otherwise specified, parts are parts by weight, molecular weights are weight-average molecular weights, temperatures are in degrees Celsius, and pressures are at or near atmospheric pressure.

[0341] All publications and patent applications referenced in this specification are incorporated herein by reference, as if each individual publication or patent application were specifically and individually indicated to be incorporated by reference.

[0342] The invention has been described according to specific embodiments discovered or proposed by the inventors to include preferred modes for practicing the invention. Those skilled in the art will understand that, based on this disclosure, many modifications and changes can be made to the illustrated specific embodiments without departing from the intended scope of the invention. For example, changes can be made to the underlying DNA sequence without affecting the protein sequence due to codon redundancy. Furthermore, changes can be made to the protein structure without affecting the type or amount of biological action due to considerations of biological functional equivalence. All such modifications are intended to be included within the scope of the appended claims.

[0343] Example 1

[0344] Adaptive deep brain stimulation targeting sleep stages in Parkinson's disease

[0345] introduce

[0346] Sleep dysfunction can be disabling in people with Parkinson's disease and is associated with worse motor and non-motor outcomes. Sleep-specific adaptive deep brain stimulation (DBS) has the potential to target the pathophysiology of sleep. Improvements in sleep structure following DBS initiation are incidental byproducts of stimulation targeting daytime motor symptoms (including tremor, bradykinesia, and rigidity) rather than optimizing nighttime sleep physiology. 7,8,14Furthermore, the adaptive scheme of DBS in PD mainly focuses on beta activity, with minimal attention paid to nighttime slow-wave activity. 15 Modulation of DBS stimulation parameters specifically targeting NREM and REM sleep stages, along with adjustments for neurophysiological and behavioral outcomes (e.g., RBD), will provide a key tool for revealing the interaction between DBS and sleep neurophysiology. Identifying optimal parameters for individual sleep stages has the potential to advance novel neuromodulation therapies targeting sleep dysfunction to improve motor and nonmotor symptoms the following day, and possibly slow disease progression by optimizing slow-wave activity. 7 However, sleep physiology is multifaceted and dynamic across many frequency bands, making it more complex than conventional beta-band-centric adaptive DBS. Therefore, the sensitivity and specificity of stimulation parameters to individual sleep stages benefit from machine learning-based sleep staging discrimination based on intracranial data.

[0347] We report a novel approach to modulating sleep in patients with PD using a fully automated adaptive DBS algorithm that adjusts stimulus amplitude based on sleep stage-specific intracranial cortical biomarkers, demonstrated in two participants with PD. We target N3 sleep as a proof-of-principle for sleep-specific adaptive DBS to investigate preliminary effects on slow-wave activity and propose a procedure that can be implemented entirely remotely in the patient's home to potentially target other sleep stages.

[0348] Materials and Methods

[0349] A. RC+S system

[0350] This study was reviewed by our institutional review board and registered on clinicaltrials.gov (NCT0358289; IDE G180097). We recruited two participants diagnosed with idiopathic PD who provided written informed consent. Bilateral electrodes were implanted in either the STN (Participant 1) or the globus pallidus (GP; Participant 2) nucleus. DBS targets were determined by the participants' treatment clinical teams and included two primary targets to test procedures during stimulation at both the STN and GPi. The implanted electrodes were connected to an investigational sense-enabled Summit RC+S (Medtronic) DBS implantable neurostimulator (INS) as part of a parent study investigating daytime closed-loop DBS for motor symptoms. Figure 1B ) 16 Patients undergo routine DBS programming by movement disorder specialists to optimize stimulation of daytime motor symptoms. Our electrode implementation consists of bilateral sensory and stimulatory quadrupole leads in basal ganglia targets and bilateral quadrupole subdural electroencephalography (ECoG) arrays spanning the precentral and postcentral gyri.16 Field potential (FP) time series recordings were analyzed via time-frequency decomposition using a Fast Fourier Transform (FFT) embedded in the INS. All data recordings and stimulation tests were performed remotely from the patients' homes. For safety reasons, participants could manually switch from adaptive mode via a personal programmer if needed.

[0351] B. Collection of polysomnography and electrocorticography data

[0352] Participants received clinically optimized chronic neurostimulation and dopaminergic drug therapy while simultaneously receiving nocturnal subcortical and cortical (precentral gyrus) FP and data from portable polysomnography (PSG, Dreem2 headband, DreemCo., Paris, France). 17 The Dreem2 headband provides scalp EEG time series and automated sleep stage classification sleep structure maps, aligned with the American Academy of Sleep Medicine sleep scoring method, but using an automated algorithm validated on healthy adult subjects. 17,18 During offline analysis, sleep structure maps of participants' sleep stages for a given night were time-synchronized with intracranial cortical and subcortical FP data via data timestamps with a resolution up to 1 second.

[0353] C. Development of Sleep Stage Classifier Model

[0354] We recorded PSG plus intracranial subcortical and precentral cortical neural data for five and six consecutive nights for Participant 1 and Participant 2, respectively; a total of 30 hours for Participant 1 and a total of 36.3 hours for Participant 2. Figure 1C During the deeper sleep stages of N2 / N3, attenuation of β power (12–30 Hz) and γ power (30–60 Hz) was observed in cortical ECoG data, with increases in low-frequency θ power (5–10 Hz) and δ power (0.5–4.5 Hz), where less pronounced differences were found in the subcortical region. Figures 1D-1E Therefore, we utilize cortical, rather than subcortical, FP data for embedded RC+S sleep stage classification to minimize stimulus-related artifacts and align with the cortical input of an automated sleep structure map.

[0355] The RC+S INS is capable of implementing up to two linear discriminative classifiers, each using up to four spectral power bands as input. The INS' embedded classifier computes the inner product of a researcher-defined weight vector (w) with a vector (x) of up to four feature inputs and compares the result to a user-defined threshold (α).

[0356]

[0357] The above- and below-threshold evaluation of the inner product leads to changes in the control strategy of stimulation parameters, such as a predefined increase or decrease in stimulation amplitude. We implement a single classifier for each INS. The feature input of the classifier is the power data averaged over 60 FFT intervals in a 1-second window (250 samples) with 50% overlap and 100% Hann filter (equivalent to a 30-second sleep stage).

[0358] We used canonical sleep frequency bands (δ and β) as feature inputs to train an offline linear discriminant analysis (LDA) model (scikit-learn; Python) to classify N3 and non-N3 sleep periods. Figure 2A ) 19,20 Participants 1 and 2 spent their respective 5 and 6 nights training and developing the classifier. We also tested including the θ and γ bands as input features for Participant 2; however, this inclusion did not significantly improve classification performance. LDA model weights were determined independently for each hemisphere, programmed into the embedded linear discriminant function for each INS, and validated in vivo over two consecutive nights. Figures 2A-2C During the validation night, the embedded device performed real-time continuous N3 sleep stage classification, where stimulus amplitude remained continuous (cDBS). For Participant 1, two additional test nights were run, where a positive N3 classification resulted in a 50% reduction in stimulus amplitude for the subsequent 30 seconds (aDBS). Figures 2D-2E To ensure safety and tolerability among participants, the stimulation amplitude was reduced during N3 sleep.

[0359] result

[0360] We demonstrated high specificity (0.94 ± 1.4 e⁻²) and significantly higher chance sensitivity (0.62 ± 4 e⁻²) across subjects and hemispheres for classifying N3 sleep using an intracranial cortical embedded neural classifier. Figure 2A Most false positives correspond to misclassifications of N2 sleep, which have spectral profiles overlapping with N3. Figures 2B-2C The N3 timeframe with a “deeper” profile (i.e., elevated intracranial cortical delta power and decreased beta power) further increased the sensitivity of embedded N3 classification. Figure 2C In the two aDBS nights, there was a successful reduction in the amplitude of N3 stimulation in the left and right recordings in 67% and 83% of the time, respectively; among which, there was incorrect stimulation modulation in only 3% and 6% of the left and right non-N3 recordings. Figures 2E-2F The classifier performance was not affected by any potential sensory contamination from stimulus adjustment. Figure 2A ) 21No temporal differences in each sleep stage were observed between nights with cDBS and aDBS. Figure 2D However, when stimulation was reduced, there was an increase in mean δ power on the left (11%) and right (22%) sides during the N3 period of the aDBS night (left: t(331)=-3.5, p < 1e-3; right: t(302)=-5.8, p<< 1e-3; Figure 2G ).

[0361] discuss

[0362] We demonstrated proof-of-principle for intracranial controlled, embedded, adaptive deep brain stimulation (DBS) targeting N3 NREM sleep in two participants with Parkinson's disease (PD). Our method demonstrated high specificity for N3 stage sleep and good tolerability for sleep stage adaptive DBS, where stimulus variations did not elicit detectable adverse reactions. From a clinical perspective, high specificity (low false positives) is advantageous as it reduces unwanted variations from therapeutic stimuli in untargeted sleep stages. Sensitivity could be further improved by using subject-specific characteristic inputs opposite to the canonical power band and by adjusting the LDA threshold to match the desired labeling. Although the 50% reduction in stimulus amplitude during embedded N3 classification was primarily chosen for safety reasons, the adaptive stimulation paradigm also provided evidence of increased slow-wave activity. We propose that slow waves may be suppressed by both intrinsic pathophysiological neural rhythms such as β (13–30 Hz) oscillations and excessively high DBS amplitude. 22 Since β itself is also suppressed by DBS, this may result in a subject-specific inverted U-shaped curve that correlates the stimulus amplitude with the NREM slow-wave amplitude. Furthermore, there may be other complex nonlinear interactions between DBS subharmonics and potential slow-wave entrainment, which could allow for increased endogenous slow-wave activity at optimal DBS amplitude. 23 Slow-wave activity is associated with the progression of Parkinson's disease (PD), and therefore, if confirmed in a large number of nights and subjects, slow-wave optimization via adaptive deep brain stimulation (DBS) could represent a promising new potential treatment for PD. 1]7,22 This is a step toward implementing a stimulation protocol to investigate the optimal overnight stimulation amplitude during N3 to support maximal slow-wave activity, which we anticipate will be subject-specific. Furthermore, customized DBS delivery during alternative, separate sleep stages may help restore normal sleep physiology and indicators in individuals with PD, addressing the major nonmotor symptoms of Parkinson's disease.

[0363] Limitations of this proof-of-principle study include the baseline real sleep stage markers obtained through portable polysomnography and automated sleep scoring algorithms

[12] . However, simultaneous intracranial recordings showed expected, canonical ECoG power band variations in different sleep stages classified by the Dreem2 band, particularly an increase in delta power during N3 sleep, supporting the potential separation of sleep stages in our patient population. Furthermore, our portable remote setup supports multi-night recordings in a natural setting, compared to single-night laboratory PSG, for improving sleep quality and classification model training. We also report a small sample size and no utilization of subcortical data for N3 classification, nor the incorporation of subjective measures of sleep quality. Nevertheless, multi-night home recordings support personalized sleep classification models within subjects, and the proposed approach is flexible to accommodate expanded participant cohorts, different stimulus targets, and the inclusion of auxiliary intracranial data streams. In addition, subjective measures of sleep quality can serve as outcome measures for a more complete assessment of the sleep aDBS paradigm.

[0364] If sleep stages can be categorized using subcortical electrodes, it could accelerate the translation of the proposed process into patient care. Supplemental research has shown that STN and GPi field potentials exhibit distinct NREM and REM physiology in individuals with PD, and that the obtained sleep stages can be discerned without stimulation. 24–26 Therefore, the proposed method can be adapted to include subcortical field potentials as feature inputs to a personalized linear classifier, although local stimulus-related artifacts and signal distortions may reduce classification accuracy.

[0365] Personalized sleep-stage adaptive deep brain stimulation (DBS) provides a technique for studying sleep neurophysiology in Parkinson's disease (PD). Furthermore, this method can be used for adaptive therapies targeting specific sleep symptoms and may influence motor and non-motor function the following day in PD. 27–29 .

[0366] References

[0367] 1. Bassetti, CL et al. Neurology and psychiatry: waking up toopportunities of sleep. State of the art and clinical / research priorities for the next decade. Eur. J. Neurol. 22, 1337–1354 (2015).

[0368] 2. Videnovic, A., and Golombek, D. Circadian and sleep disorders in Parkinson’s disease. Exp. Neurol. 243, 45–56 (2013).

[0369] 3. Barone, P. et al. The PRIAMO study: A multicenter assessment of nonmotor symptoms and their impact on quality of life in Parkinson’s disease. Mov. Disord. 24, 1641–1649 (2009).

[0370] 4. Diederich, N. J., Vaillant, M., Mancuso, G., Lyen, P., and Tiete, J. Progressive sleep ‘destructuring’ in Parkinson’s disease. A polysomnographic study in 46 patients. Sleep Med. 6, 313–318 (2005).

[0371] 5. Martinez-Martin, P., Rodriguez-Blazquez, C., Kurtis, M. M., Chaudhuri, K. R., and NMSS Validation Group. The impact of non-motor symptoms on health-related quality of life of patients with Parkinson’s disease. Mov. Disord. 26, 399–406 (2011).

[0372] 6. Léger, D. et al. Slow-wave sleep: From the cell to the clinic. Sleep Med. Rev. 41, 113–132 (2018).

[0373] 7. Schreiner, S. J. et al. Slow‐wave sleep and motor progression in Parkinson disease. Annals of Neurology vol. 85 765–770

[0374] 8. Baumann-Vogel, H. et al. The Impact of Subthalamic Deep Brain Stimulation on Sleep-Wake Behavior: A Prospective Electrophysiological Study in 50 Parkinson Patients. Sleep 40, (2017).

[0375] 9. Arnulf, I. et al. Improvement of sleep architecture in PD with subthalamic nucleus stimulation. Neurology 55, 1732–1734 (2000).

[0376] 10. Monaca, C. et al. Effects of bilateral subthalamic stimulation on sleep in Parkinson’s disease. J. Neurol. 251, 214–218 (2004).

[0377] 11. Iranzo, A., Valldeoriola, F., Santamaría, J., Tolosa, E. and Rumià, J. Sleep symptoms and polysomnographic architecture in advanced Parkinson’s disease after chronic bilateral subthalamic stimulation. J. Neurol. Neurosurg. Psychiatry 72, 661–664 (2002).

[0378] 12. Zuzuárregui, J. R. P., and Ostrem, J. L. The Impact of Deep Brain Stimulation on Sleep in Parkinson’s Disease: An update. J. Parkinsons. Dis. 10, 393–404 (2020).

[0379] 13. Tolleson, C. M., Bagai, K., Walters, A. S., and Davis, T. L. A Pilot Study Assessing the Effects of Pallidal Deep Brain Stimulation on Sleep Quality and Polysomnography in Parkinson’s Patients. Neuromodulation 19, 724–730 (2016).

[0380] 14. Deane, K. H. O., et al. Priority setting partnership to identify the top 10 research priorities for the management of Parkinson’s disease. BMJ Open 4, e006434 (2014).

[0381] 15. Little, S., et al. Adaptive deep brain stimulation in advanced Parkinson disease. Ann. Neurol. 74, 449–457 (2013).

[0382] 16. Gilron, R., et al. Long-term wireless streaming of neural recordings for circuit discovery and adaptive stimulation in individuals with Parkinson’s disease. Nat. Biotechnol. 39, 1078–1085 (2021).

[0383] 17. Arnal, P. J. et al. The Dreem Headband compared to polysomnography for electroencephalographic signal acquisition and sleep staging. Sleep 43, (2020).

[0384] 18. Berry, R. B. et al. The AASM manual for the scoring of sleep and associated events: rules, terminology and technical specifications. (American Academy of Sleep Medicine, 2018).

[0385] 19. Mika, S., Ratsch, G., Weston, J., Scholkopf, B. and Mullers, K. R. Fisher discriminant analysis with kernels. in Neural Networks for Signal Processing IX: Proceedings of the 1999 IEEE Signal Processing Society Workshop (Cat. No.98TH8468) 41–48 (ieeexplore.ieee.org, 1999).

[0386] 20. Pedregosa, F. et al. Scikit-learn: Machine Learning in Python. arXiv[cs.LG] (2012).

[0387] 21. Ansó, J. et al. Concurrent stimulation and sensing in bi-directional brain interfaces: a multi-site translational experience. J. NeuralEng. 19, (2022).

[0388] 22. Mizrahi-Kliger, A. D., Kaplan, A., Israel, Z., Deffains, M. and Bergman, H. Basal ganglia beta oscillations during sleep underlie Parkinsonian insomnia. Proc. Natl. Acad. Sci. U. S. A. 117, 17359–17368 (2020).

[0389] 23. Duchet, B., Sermon, J. J., Weerasinghe, G., Denison, T. and Bogacz, R. How to entrain a selected neuronal rhythm but not others: open-loop dithered brain stimulation for selective entrainment. J. Neural Eng. 20, (2023).

[0390] 24. Thompson, J. A. et al. Sleep patterns in Parkinson’s disease: direct recordings from the subthalamic nucleus. J. Neurol. Neurosurg. Psychiatry 89, 95–104 (2018).

[0391] 25. Chen, Y. et al. Automatic Sleep Stage Classification Based on Subthalamic Local Field Potentials. IEEE Trans. Neural Syst. Rehabil. Eng. 27, 118–128 (2019).

[0392] 26. Yin, Z. et al. Pallidal activities during sleep and sleep decoding in dystonia, Huntington’s, and Parkinson's disease. Neurobiol. Dis. 106, 143 (2023).

[0393] 27.Zahed, H. et al. The Neurophysiology of Sleep in Parkinson'sDisease.Mov.Disord.36, 1526–1542 (2021).

[0394] 28. Verma, AK et al. Parkinsonian daytime sleep-wake classification using deep brain stimulation lead recordings. Neurobiol. Dis. 176, 105963(2023).

[0395] 29.Amara, A. et al. Spindles and Slow Waves Predict Parkinson'sDisease-Mild Cognitive Impairment.

[0396] Example 2

[0397] Other applications of adaptive deep brain stimulation

[0398] Other conditions

[0399] Our sleep-adaptive DBS method has been validated in Parkinson's disease, but it is applicable to other neurological and psychiatric conditions treated with brain stimulation. This can also be extended to patients without intracranial implanted devices, but for external stimulation, to implement personalized sleep-specific non-invasive stimulation algorithms using transcranial stimulation or auditory / vibration-tactile stimulation.

[0400] Other sleep stages or microsleep stages

[0401] Our proof of principle includes one implementation of sleep-adaptive DBS—based on sleep stage—involving deep brain stimulation amplitude modulation during deep NREM sleep. This process can be generalized to any sleep stage, including N1 or REM, and can also be parameterized to target faster sleep-related dynamics to enhance or suppress sleep spindle, slow-wave, or REM features. REM is often important for mood and memory processing across neuropsychiatric disorders. In Parkinson's disease, there is a specific feature of REM called REM behavioral sleep disorder (when the patient is dreaming intensely), which is also targeted using this approach. Furthermore, biomarkers that indicate or predict upcoming awakening events can be targeted using this method.

[0402] Alternative stimulus packages

[0403] To date, we have demonstrated that we can alter the amplitude of stimulation based on intracranial sleep-specific physiology. However, this approach can be extended to altering any parameters of the stimulation, including but not limited to stimulation frequency, pattern, pulse width, electrode contact (vertical or directional or brain region location), and cathodic or anodic stimulation.

[0404] We have also recently tested and validated a novel method that applies a change in stimulation frequency only on one side of the brain—to induce healthy / beneficial (e.g., slow-wave) oscillations on both sides of the brain. We believe that we can actively amplify specific brainwaves by using therapeutic interference associated with frequency differences across the two stimulation sides. This is achieved through the phenomenon where frequency differences lead to the emergence of new frequencies. If the stimulation is set to 130 Hz (the standard stimulation frequency) on both sides, but the closed-loop algorithm detects a sleep phase that guarantees an increase in a specific stimulation frequency (e.g., delta waves at 2 Hz)—then by changing the stimulation frequency to either 128 Hz or 132 Hz—a new oscillation (2 Hz) will be generated at the difference between the two frequencies, which may then lead to underlying oscillations and potentially be therapeutic.

[0405] Example 3

[0406] Algorithm Training

[0407] In addition to having separate polysomnography stages, surface electrodes can be embedded in the shell of a cranially mounted deep brain pacemaker (DBP), which itself can serve as the EEG electrodes for polysomnography (to provide sleep tagging), eliminating the need for polysomnography or additional intracranial hardware (e.g., chronic cortical EEG). Further extensions to this simplification would include electrodes with subgastrional electrodes (above the skull, below the scalp) or electrodes embedded in an electrode "cap" that secures the DBS electrodes to the skull. All of these methods can provide long-term recording of cortical EEG but do not require additional hardware within the skull (which increases surgical risks).

[0408] We also envision an extension to classify sleep stages from the deep subcortical electrodes used for stimulation. This has already been shown to be possible in the absence of stimulation. With stimulation on, this would require mitigating stimulation artifacts and modeling the effect of stimulation on the rhythm of the underlying recording to avoid self-triggering of neural recordings by stimulus-related artifact changes.

[0409] At the individual level, to achieve the best possible classifier, we plan to incorporate reinforcement learning to further optimize stimulus parameters, specifically slow waves and sleep spindles. Other potential inputs for training this classifier will include: 1) nocturnal neurophysiology and sleep stage metrics; 2) patient self-reports of sleep quality; 3) morning mood, motivation, and cognitive tests (self-reported or computational behavioral paradigms); 4) resting-state intracranial brain activity associated with positive or negative states; 5) resting-state connectivity metrics, including evoked response activity, theta-pattern stimulation, and evoked resonant neural activity; 6) daytime behavioral metrics from devices such as smartwatches, motion quantification wearables, smartphones, voice recording models, posture estimation, and GPS devices, used to measure activity and clinical status; and 7) classification inaccuracies in embedded devices.

[0410] Furthermore, following within-subject normalization, we plan to generalize findings from individual (or multiple) test subjects to a combined model that does not require individualized model training at the outset. This is likely to be more scalable than having individualized models for each subject, and we hope that with appropriate within-subject normalization and a generalizable model, the reduction in classification accuracy will be low and acceptable.

[0411] Our current implementation is limited by the embedded linear discriminant capability of the DBS device; that is, two linear discriminant equations dividing the 4-dimensional feature space into nine partitions. We utilize a single linear discriminant analysis equation on the optimized frequency power band for sleep stage classification. We have extended the embedded sleep stage classification to include support vector machines, 2-step decision tree classifier models, 2-step gradient boosting machines, and 2-step linear discriminant analysis classifiers, and plan to extend the classification to include optimization via convex optimization methods through estimates of parabolic or other nonlinear boundaries via two first-order Taylor expansions along nonlinear boundaries. Furthermore, if not constrained by the capabilities of the embedded device, such as performing real-time classification via a nearby tablet or computer, then our process can be extended to any nonlinear classifier of sleep stages or sleep microstructures. This includes, but is not limited to, artificial neural networks (recurrent neural networks, transformers, long short-term memory networks, feedforward networks), gradient boosting machines, random forests, and quadratic discriminant analysis. These nonlinear classifiers will utilize a larger feature space, including but not limited to the entropy or slope of frequency power, the ratio of various power band combinations, the second or higher order moments of the time domain and / or frequency distribution, and alternative data streams such as accelerometers, temperature, or respiration rate.

[0412] Example 4

[0413] Safety of adaptive deep brain stimulation

[0414] Safety is paramount in our research and in the development of any new stimulation methods. In current implementations, numerous safety methods exist that ensure safe and tolerable adaptive DBS for sleep.

[0415] Currently, stimulation limits are set for pacemakers, and these limits have been tested in a supervised manner by clinical neurologists to be safe and tolerable. These stimulation limits provide a safety barrier against dangerous stimuli. An extension of this current practice would be to iteratively (or using Bayesian optimization or other data-driven methods) implement stimulation parameter limits that start conservatively (without change) and gradually diverge toward increasingly larger changes in the stimulation parameters.

[0416] Currently, patients have their own patient programmers, where they can adjust the stimulation amplitude or change the mode of entering or exiting adaptive DBS. Therefore, patients always have ultimate control over their stimulation. However, flexibility in managing these changes is needed, especially at night when it's dark—when patient programmers may be difficult to find—or in cases of worsening clinical conditions. Extensions to this could include voice or motion activation of the patient programmer, allowing the patient (or caregiver) to easily command a switch back to regular DBS without manual interaction with the physical programmer. Alternatively, an app accessible via a wearable smartwatch would allow participants to have full control over stimulation parameters in a dark environment, from supine or prone positions.

[0417] In addition to manual stimulation control via a patient programmer (by the patient), there are significantly automated methods that can be used to automatically switch a patient from adaptive DBS back to conventional DBS. First, a classifier on the device itself (or from an externally worn sleep-sensing wearable / polysomnography headband) is used. These can be programmed to detect both arousal and movement, and therefore, if stimulation is shown to significantly increase arousal, disrupt sleep physiology, or cause abnormal nocturnal movement (which can be classified from motion sensors), it can be automatically set to switch the stimulation settings back to conventional DBS. Furthermore, it is important for the device to analyze its own stimulation patterns and behavior. The nervous system inherently possesses a degree of randomness, and therefore, if the stimulator control algorithm results in highly regular / stereotypical algorithmic behavior, this will indicate that the algorithm is self-triggering and will also guarantee the termination of adaptive DBS and the switch back to conventional DBS. Other markers that may highlight signals for a return to adaptive DBS can be analyzed using EKG or pulse measurements to analyze changes in heart rate or heart rate variability, which can signal underlying physiological stress. These constraints can be compared to a standard reference distribution, or alternatively, to a personalized distribution of heart rate, heart rate variability, electromyography, skin conductance response, motion on a wearable device (accelerometer or gyroscope) (or to non-invasive measurements such as bed sensors or radar), or other personalized measures.

[0418] Example 5

[0419] Multi-night natural cortical-basal recordings reveal the mechanisms of NREM slow-wave suppression and spontaneous arousal in Parkinson's disease.

[0420] introduce

[0421] Sleep disturbances are one of the most common nonmotor symptoms of Parkinson's disease (PD), with up to 90% of PD patients experiencing sleep dysfunction. 1 Furthermore, 60% of PD patients experience multiple sleep disturbance symptoms. 1,2 Changes in sleep patterns typically precede classic neurological symptoms in PD and are correlated with the rate of progression and disease severity. 3 Sleep dysfunction in Parkinson's disease (PD) has a negative impact on daytime mood, cognition, fatigue, and other comorbidities. 4–8 Among these, non-motor symptoms and sleep symptoms are more indicative of quality of life than classic motor symptoms. 9–11 Therefore, understanding the neurophysiology of sleep disorders in PD may lead to new principled therapies aimed at improving sleep quality, alleviating daytime symptoms, and improving patients' quality of life.

[0422] The sleep architecture in humans is broadly defined by the physiologically distinct stages of rapid eye movement (REM) and non-REM (NREM) sleep. Further characteristics of NREM sleep include rhythmic low-frequency electroencephalogram (EEG) activity in the delta (0–4 Hz) and theta (4–7 Hz) ranges, increased parasympathetic activity, and limited dreaming. NREM currently has three formally defined substages: N1 (light sleep), N2 (characterized by K-complexes and sleep spindles), and deep N3 (characterized by slow delta waves). 13 Sleep dysfunction in PD manifests as parasomnias, fragmented sleep, and interrupted sleep patterns, including a significant reduction in both REM and NREM sleep. 11 In particular, reduced NREM slow-wave sleep activity in the delta range (< 4 Hz) was associated with worsening daytime motor symptoms and accelerated disease progression in PD. 3,14,15 .

[0423] During wakefulness, beta oscillations (13–30 Hz) are a hallmark oscillatory feature of PD and are associated with daytime motor symptoms. 16 Recent studies on non-human primates (NHPs) during sleep have shown that subcortical β activity is also associated with reduced cortical δ activity, and have demonstrated the role of subcortical β in spontaneous arousal in PD. 17 Beta oscillations have been detected during sleep in PD patients. 18–22 However, to date, human studies have not investigated the mechanical interactions between subcortical β and cortical sleep physiology (including slow waves), nor have they studied spontaneous arousal, and previous studies have been conducted in the absence of deep brain stimulation (DBS). Understanding the true contribution of the corticobasal ganglia circuit to sleep dysfunction in PD and its interaction with DBS has been limited until now by the inability to record intracranial activity throughout the night at high resolution over a long period. This challenge has been addressed by the advent of a new generation of sensor-enabled DBS devices that can remotely transmit neural data from the patient's own home. 23 A better understanding of cortical-basal activity during sleep may reveal underlying mechanisms of sleep dysfunction in Parkinson's disease and could help improve sleep therapies, including sleep-targeted adaptive deep brain stimulation (aDBS).

[0424] In this study, we recruited four patients diagnosed with Parkinson's disease (PD) and one control patient with cervical dystonia. All patients had long-term implanted intracranial electrodes capable of sensing sensorimotor cortical and basal ganglia (STN / GPi) field potentials (FP). We performed nocturnal at-home intracranial cortical and subcortical recordings over multiple nights (n=58) with and without DBS stimulation, paired with portable polysomnography. We demonstrated a significant interaction between subcortical β oscillations and cortical slow-wave activity in the delta band during NREM, an effect modulated by DBS, and also showed a significant increase in subcortical β prior to spontaneous arousal.

[0425] result

[0426] Four individuals with PD (Table 1) (x2 individuals received bilateral STN + sensorimotor cortical ECoG, and x2 individuals received bilateral GPi electrodes + sensorimotor cortical ECoG) and one individual with cervical dystonia (bilateral GPi electrodes + sensorimotor cortical ECoG) successfully initiated recordings remotely from intracranial cortical-basal and external portable polysomnography (Dreem2) at their own homes over 58 nights (54 nights with stimulation on and 4 nights with stimulation off). Intracranial and extracranial recordings were synchronized, and artifacts were removed. Figure 4E (Figure 11), resulting in interpretable cortical and subcortical recordings, even in the presence of DBS. A total of 415 hours of sleep was recorded across all participants (Supplementary Table 1; Figure 9). PD subjects slept an average of 7.25 ± 0.18 hours per night during the extended multi-night on-stimulation period (n = 45; duration in minutes: N1 = 34.26 ± 1.34; N2 = 164.72 ± 7.92; N3 = 90.06 ± 10.53; REM = 97.64 ± 6.54; wakefulness after sleep = 48.18 ± 4.71). In separate consecutive nights of on vs. off DBS recordings, all four PD subjects showed increased durations of deep NREM (N3) and REM during on-stimulation compared to the corresponding off-stimulation night (Supplementary Table 1 and Figure 5). Power spectral density maps from intracranial electrodes ( Figure 4G Figure 12 shows the expected classical variations of the canonical frequency bands in NREM and REM sleep stages, supporting the separation of different sleep stages using our portable PSG device.

[0427] Spectral power variation of NREM

[0428] We investigated spectral changes in intracranial femoral activity during NREM (N2 and N3) with wakefulness as baseline, prior to which we focused on cortical-basal delta and β-cell activity. 24Power spectral analysis and a linear mixed-effects (LME) model for mean nighttime frequency band power, with a fixed effect on sleep stage (NREM vs. wakefulness; considering multiple nights in participants) and a random effect on subjects (n=5), showed an increase in mean delta power (1–4 Hz) in both cortical and subcortical regions during NREM sleep compared to wakefulness (cortical: β=0.42, 95% CI=[0.36, 0.47], p=3.7e-33; subcortical: β=0.1, 95% CI=[0.07, 0.13], p=3.3e-12; n=105; CI=confidence interval) and a decrease in β power (13–31 Hz) in both cortical and subcortical regions during NREM sleep. Hz; Cortical: β=-0.4, 95%CI=[-0.44, -0.37], p=1.5e-41; Subcortical: β=-0.2, 95%CI=[-0.22, -0.17], p=1.4e-25) Figures 5A-5B (Multi-night stimulation). These spectral changes in NREM compared to wakefulness were also observed during single-night DBS sleep recordings in both cortical and subcortical regions of all four PD participants. Figure 5C ).

[0429] Our direct comparison of PD (n=4) and dystonia (n=1) analyses revealed that during NREM sleep, subcortical β power was lower in patients with dystonia than in all four PD patients (LME model of fixed effects of PD and dystonia: β=0.18; 95% CI=[0.09, 0.27]; p=0.0001). Furthermore, compared with PD patients, the frequency band power variation between NREM sleep and wakefulness was smaller in dystonia participants ( Figure 5B The LME model showed a statistically significant fixed effect of disease state (PD and dystonia) on band power changes between NREM and awake phases in the cortex (δ:β=0.24, 95%CI=[0.14, 0.35], p=1.3e-5; β:β=-0.21; p=3.4e-7) and subcortical (δ:β=0.08, 95%CI=[0.013, 0.14], p=0.02; β:β=-0.13, 95%CI=[-0.2, -0.07], p=0.0001).

[0430] We also investigated how these FP activities varied with stimulation and compared the power spectra between on- and off stimulation conditions in our PD cohort (n=4). Spectral power comparisons revealed a further relative increase in delta activity in cortical FP, and a further decrease in alpha and low-β activity, during NREM sleep under both on- and off DBS conditions. Figure 5DA random-effects LME model with subjects as subjects revealed that, for NREM and wakefulness, stimulation under on- and off-conditions resulted in a further increase in cortical delta activity (1–4 Hz; β = 0.026, 95% CI = [0.003, 0.05], p = 0.03) and a decrease in cortical alpha activity (8–13 Hz; β = -0.0297, 95% CI = [-0.05, -0.003], p = 0.03), and low β activity (13–15 Hz; β = -0.026, 95% CI = [-0.042, -0.01], p = 0.006). No significant changes in subcortical power spectral density (FP) were observed during NREM in comparisons of on- and off-conditions; however, changes in baseline subcortical power levels under on- and off-conditions of DBS may have masked any underlying changes. Overall, these data reveal that DBS during NREM sleep leads to relatively high cortical delta activity and decreased alpha and low β activity.

[0431] Changes in functional connectivity in NREM

[0432] We then explored NREM-related changes in functional connectivity between cortical and subcortical regions to investigate sleep-related changes in the cortical-basal ganglia circuit in PD. To this end, we compared the spectral coherence of cortical and subcortical FP activity between NREM sleep and wakefulness. In all participants, an LME model investigating spectral coherence (NREM vs. wakefulness) with a fixed effect of sleep stages revealed that, compared to wakefulness during onset stimulation, the total difference in spectral coherence in δ was increased (β = 0.05, 95% CI = [0.04, 0.06], p = 5e-11; n = 104) while β was decreased (β = -0.18; 95% CI = [-0.23, -0.14], p = 5.5e-13) during NREM sleep. Figures 5E-5F During single-night recordings of stimulation cutoff in PD participants, an increase in delta coherence and a decrease in beta coherence were also observed in PD participants during NREM. Figure 5G The comparison between PD and dystonia also showed that, compared with PD participants, NREM and the change in cortical-basal δ / β coherence were smaller in dystonia participants (LME model with PD / dystonia status as a fixed effect; δ coherence: β=0.05, 95%CI=[0.02, 0.08], p=0.0008; β coherence: β=-0.21, 95%CI=[-0.31, -0.11], p=0.0001).

[0433] In our on / off DBS analysis (PD participants only; n=4), we also noted a statistically significant further reduction in low β (13–15 Hz) coherence during on stimulation compared to off stimulation (LME model with on / off conditions as fixed effects and subjects as random effects; β = -0.012, 95% CI = [-0.021, -0.003], p = 0.015). Overall, these data suggest that functional connectivity between cortical and subcortical structures is modulated during NREM sleep and wakefulness. Increased delta coherence and decreased β coherence were observed in PD under both on and off stimulation conditions, with these effects enhanced under DBS on stimulation.

[0434] Interaction between cortical δ and subcortical β activity

[0435] The aforementioned spectral power and functional connectivity analyses revealed inverse changes in δ and β FP activity during NREM sleep and wakefulness. To further examine the direct relationship between these two rhythms, we specifically investigated the interaction between cortical δ and subcortical β FP activity within NREM (N2+N3) on short timescales. Here, we observed an inverse relationship between cortical δ power and subcortical β power during NREM sleep. Figure 6A To quantify this relationship, we first used standard correlation analysis, which revealed a negative correlation between subcortical β and subcortical δ FP power (5-second interval) in all PD participants during NREM, under both on and off stimulus conditions. Figures 6B-6C LME modeling of band power during NREM periods from all participants (n=232,064) showed an overall negative fixed effect of subcortical β power on cortical δ power (β=-0.24, 95%CI: [-0.28, -0.2], p=3.9e-30). Furthermore, we found a fixed effect of PD on dystonia status during NREM sleep-on stimulation (β=0.06, 95%CI: [0.01, 0.11], p=0.02), indicating that this effect was greater in PD patients than in our dystonia controls. During the stimulation-off condition, a negative fixed effect of subcortical β power on cortical δ power was also obtained in PD participants using the LME model (β = -0.38, 95% CI: [-0.44, -0.32], p = 1.9e-32; n = 17,518). These results indicate a negative correlation between subcortical β and cortical δ FP power in NREM sleep during both stimulation-on and stimulation-off conditions, and this effect is significantly stronger than in our contrasting dystonia patients.

[0436] Next, we used cross-correlation analysis to determine whether subcortical β led or lagged changes in cortical δ. We observed that during NREM sleep, in 3 out of 4 PD participants, an increase in subcortical β led to a decrease in cortical δ ( Figure 6D The mean lag over multiple nights with DBS on was: PD2: 4.5 s, n=11; PD3: -11.4 s, n=11; PD7: -7.5 s, n=10; PD9: -3 s, n=10. Finally, as a control analysis excluding the general inverse relationship between cortical-basal circuit δ and β, simply reflecting the depth of NREM sleep, we also correlated the power of cortical δ and cortical β from the same region. If the inverse relationship between cortical-basal circuit δ and β is simply a function of sleep stage depth, we could expect the inverse relationship between cortical δ and cortical β to be strongly reversed. Unlike the negative correlation between subcortical β and cortical δ for all PD participants, cortical δ and β showed a weak negative correlation only in 3 PD participants during NREM (on-stimulation), and in one PD participant ( Figure 6E The LME analysis did show a positive correlation between cortical β power and cortical δ power during NREM sleep (β ​​= -0.21, 95% CI: [-0.32, -0.1], p = 0.0002; n = 232,064), but no fixed-group effect was shown for PD / dystonia status (β = -0.02, 95% CI: [-0.06, 0.004], p = 0.1), indicating no evidence of difference between PD / dystonia status at the cortical level compared to the subcortical level. Furthermore, in direct model comparisons, the LME model of cortical δ with a fixed effect of subcortical β showed a statistically significant improvement over the model of cortical δ with a fixed effect of cortical β (simulated likelihood ratio test, repeated 100 times; p = 0.01). This indicates that subcortical β has a stronger effect on cortical δ than on the relationship between cortical β activity and cortical δ activity, supporting the view that this subcortical β-cortical δ effect is greater than any effect of sleep stage depth. Furthermore, our data show that the subcortical β-cortical δ effect is relatively specific to PD.

[0437] Changes in spectral power before spontaneous awakening

[0438] To better understand FP activity with a finer temporal resolution and to investigate the dynamics of intracranial FP leading to wakefulness, we analyzed changes in the spectral power of delta (δ) and beta (β) during the transition from NREM to spontaneous wakefulness. There were an average of 26.7 ± 1.5 awakening events per night, with a total duration of 52.7 ± 4.5 minutes for each participant with onset stimulation. During NREM, cortical delta power gradually increased with sleep depth across all participants (multi-night onset stimulation dataset) and decreased before wakefulness. Figure 7A The mean cortical delta power during the pre-awakening (-5 seconds) and post-awakening (+15 seconds) periods was lower than the mean spectral power found during the deep NREM phase. Figure 7A Before awakening: β = -1.3, 95% CI: [-1.9, -0.6], p = 7.4e-5; n = 446; After awakening: β = -2.9, 95% CI: [-3.5, -2.3], p = 7.7e-20), where the cortical delta power after awakening was lower than before awakening ( Figure 7A ;β=-2.9 and -1.3) and PD / dystonia status without fixed effect (before wakefulness: p=0.5; after wakefulness: p=0.06). This indicates that changes in cortical δ are not PD-specific, but rather a general characteristic of neurophysiological changes during sleep and wakefulness in NREM. Subcortical δ (before wakefulness: p=0.6; after wakefulness: p=0.002) and cortical β power (before wakefulness: p=0.00004; after wakefulness: p=0.98) did not show significant consistent changes between participants before and after spontaneous wakefulness. Figures 7B-7C However, subcortical β power showed an increase before wakefulness and persisted after wakefulness. Figure 7D Before awakening: β=0.6, 95%CI: [0.35, 0.84], p=2.5e-6; After awakening: β=1.4, 95%CI: [1.1, 1.6], p=1e-19; Power after awakening was higher than before awakening (β=1.4 vs. 0.6). Disease state (PD / dystonia) had a statistically significant fixed effect on the increase in subcortical β power before awakening (β=-0.9, 95%CI: [-1.36, -0.43], p=0.0002), but only one trend after awakening (after awakening: p=0.06), indicating that the increase in subcortical β is PD-specific.

[0439] Finally, we investigated whether machine learning models could predict the transition to spontaneous arousal. To this end, we trained a subject-specific quadratic discriminant analysis (QDA) classifier characterized by cortical delta, cortical low gamma, subcortical β, and low gamma spectral power during the onset of stimulation. For each participant (4 PDs and 1 dystonia), we performed 10-fold cross-validation within a 5-second time window from deep NREM to the spontaneous arousal event. Figure 8A 8A). We found that, prior to full arousal, the arousal prediction provided by the subject-specific model increased its predictive power for arousal. Specifically, the QDA model was able to distinguish between deep NREM and pre-awake NREM (-5 seconds) with reasonable accuracy (approximately 70%) in most participants (accuracy of deep vs. pre-awake NREM: dystonia = 63.7%, PD3 = 73.7%, PD9 = 55.6%, PD2 = 68.9%, PD7 = 69.8%; Supplementary Table 2; Figures 8A-8B Deep NREM versus the waking phase (+15 seconds) showed higher accuracy than deep versus pre-awake NREM, as expected. The area under the curve (AUC) performance in the deep versus pre-awake NREM classification was promising among PD participants (>70% in PD3, PD2, and PD7; Supplementary Table 2); Figure 8B ROC-based optimization, which specifies the classification threshold, can further improve the prediction of arousal in PD. Despite using only four spectral power features as input, the performance of the QDA model demonstrates the feasibility and potential application of machine learning algorithms for identifying micro-stages of sleep and designing adaptive DBS therapies that can modulate stimulation to prevent arousal.

[0440] discuss

[0441] We remotely collected multi-night intracranial brain recordings from cortical and subcortical regions from four patients with PD and one with dystonia in their own homes over 58 nights, and paired them with polysomnography against both DBS on and off conditions. We found that during NREM, slow-wave activity in the delta band of the cortical-basal network increased and beta power and connectivity decreased, an effect amplified by DBS. In NREM, there was a direct inverse relationship between subcortical beta and cortical delta activity, and further, we found that subcortical beta power increased before spontaneous arousal. These data reinforce the hypothesis that subcortical beta is associated with nocturnal sleep disruption and spontaneous arousal in PD.

[0442] Our study advances our understanding of sleep neurophysiology in several areas. First, technically, we recorded high-resolution intracranial cortical and subcortical neural activity during sleep over multiple nights (n=58) in the participants' own homes, using a fully embedded, sense-enabled DBS device in their natural environment. Second, we provide evidence of subcortical β and subcortical δ interactions during NREM in PD participants and their modulation by DBS. This effect has previously been observed in primate models of PD. 17 However, to our knowledge, our study is the first to demonstrate this interaction between humans and PD. Third, this is supported by an analysis of NREM sleep neurophysiology under both on and off stimuli, revealing both stimulus-dependent (increased cortical delta and decreased α and β in NREM) and stimulus-independent (interaction between subcortical β and cortical delta) effects. In our comparison with our dystonia subjects, our cortical-basal delta-β interaction findings and the pre-awakening subcortical β elevation had a significantly stronger effect in the PD cohort.

[0443] It has now been established that excessive subcortical β-oscillations occur during the day in PD and may lead to circuit disruption and motor symptoms. 25,26 Here we show that subcortical β-oscillations also disrupt slow cortical oscillations during NREM sleep in humans with PD and partially contribute to wakefulness during the night, validating the findings from a primate model of PD. 17 Furthermore, we show that DBS stimulation, known to reduce subcortical β-oscillations during wakefulness. 27 Here, during NREM sleep, an increase in cortical delta power and a decrease in cortical alpha and low β power are observed. This finding aligns with previous research, which found an increase in EEG delta power accumulation during NREM sleep due to thalamic-depressive-diverticulum (DBS) in PD. 28 One current hypothesis is that DBS therapy improves subjective sleep by reducing nighttime discomfort through improved movement. Data from our study suggest that DBS therapy appears to further improve sleep in PD by directly modulating β and δ oscillations.

[0444] Although previous studies have documented subcortical β oscillations during sleep in both the STN and GPi. 18–22However, to date, these studies have all been single-night studies and have not explored the interaction between β and δ, spontaneous arousal during the night, or the effects of DBS stimulation on sleep physiology. Here, we show that in NREM sleep, subcortical β is negatively correlated with cortical δ power and precedes spontaneous arousal on rapid timescales. Our findings on the mechanism of cortical-subcortical interactions during sleep provide a foundation for developing closed-loop adaptive DBS methods to restore normal sleep patterns in individuals with PD.

[0445] The link between sleep dysfunction and daytime motor, mood, and cognitive symptoms makes sleep an attractive potential target for further research. 6–8 Furthermore, sleep disturbances, particularly a reduction in cortical slow-wave activity during NREM, are associated with faster disease progression. 3,14 Therefore, targeting β-oscillations during NREM sleep has the potential to reduce all-night insomnia, increase cortical slow waves, and improve motor and non-motor symptoms during wakefulness. This supports the view that daytime neural activity and nighttime sleep physiology are significantly separable and different strategies for aDBS are needed to optimize the rhythms during these two distinct phases. Implementing different aDBS algorithms around the circadian rhythm cycle can be achieved by introducing a daytime (compared to sleep) neural classifier, a circadian rhythm (clock)-based algorithm, and feedforward and feedback controllers that optimize the combination of both daytime and nighttime neurophysiology. 29,30 .

[0446] Our study has limitations worth discussing. First, our baseline true sleep stage labels were obtained using a portable polysomnography system and an automated sleep scoring algorithm, and validated on healthy controls. 31 Instead of conventional laboratory-based PSG, we used intracranial recordings grouped according to sleep stages defined by our portable PSG, which revealed expected and classic variations in ECoG activity across various stages (Figure 12). In particular, the increased delta power and decreased beta power observed during N3 sleep provided evidence for using this protocol to differentiate potential sleep stages in our patient group. Figure 4FFurthermore, our portable remote setup allowed us to collect multi-night recordings in a natural setting, which is advantageous compared to single-night PSG recordings (from a sleep lab) that can be affected by first-night adaptation effects. We also report a small sample size of participants, although it is noteworthy that we collected many nights of recordings for each subject (n=58 in total), which supported highly statistically powerful LME analyses that modeled effects both within and across subjects, similar to the advantages of primate studies. Our contrast participants were a single patient with cervical dystonia (rather than a formal control group), reflecting the uniqueness of this patient cohort with high-resolution, sense-enabled pulse generators and long-term implanted ECoG electrodes. Nevertheless, and given the large size of the intra-subject dataset and the linear mixture modeling, we were able to show differences between the dystonia patients and the PD group, which should be supported by larger and more balanced cohorts in the future. Finally, here we limit our analysis to NREM and canonical power bands, with a focus on β and δ. 24 We did not examine changes in other sleep stages, nor did we specifically analyze sleep spindle waves (which overlap with low beta in frequency) or other bands that will be reported separately.

[0447] in conclusion

[0448] In this study, we recorded and analyzed intracranial FP (presumably sleep-related activity) at home over multiple nights in PD participants with and without DBS (polysomnography). Our data revealed that cortical-basal network power and connectivity in the delta and beta bands were increased and decreased, respectively, in NREM (normative cerebral echocardiography) compared to wakefulness, effects amplified by DBS. Furthermore, in NREM, cortical delta slow-wave activity was negatively correlated with subcortical beta, which was also elevated prior to spontaneous wakefulness. These findings reveal the role of subcortical beta in sleep dysfunction in PD and provide a target for future personalized sleep-specific adaptive DBS.

[0449] method

[0450] Participants, Demography and Ethics

[0451] We recruited four participants with idiopathic PD for this study (Table 1). Movement disorder physicians diagnosed them according to the Movement Disorders Association's PD diagnostic criteria. 32Each individual was diagnosed with PD. The motor component of the Combined Parkinson's Disease Rating Scale (UPDRS) score was administered by a trained rater. We also recruited a participant with cervical dystonia as a control subject. Participants were recruited from a parent study that focused on investigating closed-loop DBS for daytime motor symptoms. The implanted electrodes were connected to an investigational sense-enabled Summit RC+S DBS implantable pulse generator provided by Medtronic. Figure 4A ) 23 This study was reviewed by our Institutional Review Board and registered on clinicaltrials.gov (NCT0358289; IDE G180097). The study was also reviewed by the Office of Human Resource Protection (HRPO) of the Defense Advanced Research Projects Agency (DARPA). Written informed consent was provided by all participants. All participants had chronic bilateral cortical ECoG electrodes, and two PD participants had bilateral electrodes implanted in the subthalamic nucleus (STN; PD2 and PD7), and two PD and one dystonia participant had bilateral electrodes implanted in the globus pallidus nucleus (GPi; PD3, PD9, and the dystonia participant). Figure 4B The DBS electrode implantation target is determined by the clinical team. Movement disorder specialists program the patient using standard DBS settings to optimize stimulation to address daytime motor symptoms.

[0452] Experimental Design and Scheme

[0453] We collected data using two protocols: a long-term, multi-night data collection with on-stimulation plus separate two-night comparative recordings, one night with DBS on and one night with DBS off. During the long-term, multi-night data collection, each participant (n=5) was equipped with a portable PSG (Dreem 2) headset, and overnight intracranial and polysomnography data were recorded for approximately 10 nights (Supplementary Table 1), which were primarily continuous. In the on / off protocol, overnight data were collected from PD participants (n=4) over two consecutive days. DBS was on on day 1 (3.075 ± 0.65 mA), and DBS was off on day 2. PD participants received their routine clinical dopaminergic replacement therapy during both data collection protocols. At the patients' request, on / off recordings were not performed in patients with cervical dystonia. All data recordings were performed remotely from the patients' homes.

[0454] Polysomnography

[0455] Extracranial polysomnography (PSG) is recorded using the Dreem headband, which includes an automated sleep staging algorithm with extracranial electroencephalogram (EEG) data (Dreem2 headband, Dreem, Paris, France). 31,33 The Dreem2 headband provides a sleep stage classification sleep structure map based on the American Academy of Sleep Medicine (AASM) scoring method (NREM: N3, N2, N1, and REM), which has been validated in healthy subjects. Figure 4C ) 31,33 Sleep staging was performed using EEG data in 30-second intervals. Sleep onset was defined as the beginning of NREM sleep (requiring three consecutive intervals to classify N1). Waking after sleep onset (WASO) was calculated as the total awake time from sleep onset to the last sleep interval. Because N1 is difficult to detect and physiologically distinct, we focused our analysis on N2 and N3 stages of NREM sleep.

[0456] Intracranial data collection

[0457] For each participant, a Summit RC+S device was bilaterally implanted and connected to a quadrupole lead with bilateral sensing and stimulation capabilities in a basal ganglia target (GPi in 2 PD patients or GPi in 2 PD patients and 1 cervical dystonia patient) plus a quadrupole sensorimotor chronic cortical electroencephalogram (ECoG) that only sensed the bands, with 4 electrode contacts spanning the central gyrus ( Figure 4B In addition to data from bilateral accelerometers embedded in a chest-mounted pulse generator device, overnight intracranial data were collected from cortical and subcortical structures in both the left and right hemispheres. Figure 4D Time series FP data are recorded at a sampling rate of 250 Hz or 500 Hz.

[0458] Data preprocessing

[0459] Accelerometer data was used to validate intracranial recordings and synchronize them with PSG recordings. Cross-correlation was applied to accelerometer data from both the Dreem2 band and the RC+S neurostimulator to determine the delay between the PSG and RC+S time series (Fig. 11). Since sleep stage estimations from the PSG sleep structure map were performed in 30-second intervals, we also used post-awake motion (measured via accelerometer) to further realign with wakefulness on a scale less than 30 seconds (Fig. 11). All intracranial data were downsampled to 250 Hz and filtered through a 0.8–100 Hz zero-phase IIR elliptical bandpass filter with 1 dB passband ripple and 100 dB attenuation (using the “filtfilt” and “designfilt” functions in Matlab). Large artifact spikes in the subcortical intracranial data were removed along with the corresponding cortical data (Fig. 11). To identify artifacts, the absolute square subcortical data were first smoothed with a Gaussian kernel with a 1-s window, and any time interval greater than 5 times the median throughout the night was considered an artifact spike. A combination of two ECG data removal algorithms (“PerceptHammer” and “Perception” libraries; Matlab; Figure 11) was used to remove ECG artifacts from subcortical data. 34,35 .

[0460] Power spectral analysis

[0461] To calculate the power spectrum, z-scores were performed on intracranial data from each night for each location. The NREM data segments (N2+N3) were then aggregated according to PSG sleep structure map labels. The selected data were segmented into 5-second intervals, and the power spectrum for each interval was calculated using the Welch method (“pwelch” in Matlab) with a Hamming window of a 1-second 512-point FFT with 50% overlap, normalized to the total power from 0–50 Hz. The calculated power spectra for each interval were then aggregated across the two hemispheres of the subject. To calculate the variation in the power spectrum during NREM with wakefulness as the baseline, the power spectrum for wakefulness periods was calculated in a similar manner to that during NREM, and the difference between the average wakefulness power spectrum and the NREM power spectrum for each night was calculated. To calculate the power spectrum for on and off nights, the average power spectrum was calculated for on and off nights, and their differences were obtained. The average was calculated on the logarithmically transformed power spectrum.

[0462] Spectral coherence analysis

[0463] To compute spectral coherence, intracranial data obtained from each night were normalized using z-scores for each location. NREM data segments incorporating N2 and N3 sleep stages were then extracted and segmented into 5-second intervals. For each interval, 5 seconds of cortical and subcortical data were used to estimate unilateral amplitude-squared coherence using a multi-tapered method (“mscohere”; Matlab) with a 1-second Hamming window and a 512-point FFT. Inter-interval spectral coherence was then pooled across both hemispheres. Similar to power spectral analysis, spectral coherence during wakefulness was computed, and the difference between the nightly wakefulness and the mean NREM spectral coherence was calculated to obtain the variation in spectral coherence in the NREM with wakefulness as a baseline.

[0464] β-δ correlation analysis

[0465] To analyze the interaction between subcortical β and cortical δ activity during NREM sleep in intracranial signals, we applied z-scores, power spectrum calculations, and normalization techniques as previously described. However, there was an exception regarding the normalization of the cortical power spectrum, where instead of normalizing it by dividing by the total power (0–50 Hz), we divided it by the total power excluding the β range (0–13 and 31–50 Hz). This adjustment was necessary to avoid detecting spurious negative correlations that might be introduced by the normalization procedure itself. Both subcortical β and cortical δ were calculated for 5-second intervals, which were logarithmically transformed for each night and each hemisphere. The band power was then pooled across both hemispheres. Subsequently, for each participant, we calculated Spearman's rho correlation coefficient between subcortical β and cortical δ power across all 5-second intervals each night. Similar results were obtained when various other normalization methods were employed. To calculate the lag between subcortical β power and cortical δ power, normalized cross-correlation between these frequency bands was calculated from the aforementioned 5-second intervals each night (using the "xcorr" function in Matlab). The lag was calculated by finding the minimum (valley, reflecting a negative relationship) normalized cross-correlation between the two frequency bands. A 20-point Gaussian kernel was used to smooth the frequency band power for each night. The nightly data were subtracted from the mean and aggregated from both hemispheres. To investigate the interaction between cortical δ and β power, we applied the same power spectrum calculation technique described previously in the β-δ correlation analysis to the 5-second intervals. The only exception was the power spectrum normalization step, which was not used to avoid detecting artificial negative correlations that might be imposed by the power spectrum normalization. Spearman's rho correlation coefficient was calculated for all participants across all 5-second intervals each night between cortical δ and β power.

[0466] NREM to lucidity transition analysis

[0467] To investigate the variation of spectral power during the transition from NREM to wakefulness, all intracranial data were bandpass filtered using a zero-phase IIR elliptic bandpass filter. Next, z-scores were performed on data from each hemisphere for each night at all locations. A Hilbert transform was applied to the z-score data, and the absolute square of the results was converted to a decibel scale for specific frequency band power. Events with NREM sleep durations of less than 85 seconds and wakefulness durations of less than 25 seconds were ignored after all NREM-to-wake transitions were detected. A maximum total discontinuity of 1 second in sleep events was allowed, and the frequency band power was averaged for each 5-second interval. The power of deep NREM (slow-wave sleep; SWS) was calculated by averaging all intervals in the NREM data 40 seconds after the start of NREM and 40 seconds before wakefulness. The power of the wakefulness phase was calculated by averaging all intervals in the wakefulness phase data 25 seconds after the start of wakefulness. All data were analyzed from the on-stimulation multinight dataset.

[0468] Awake Prediction Model

[0469] For each participant, a separate QDA model (“fitcdiscr” in Matlab) was trained with four intracranial spectral power features: cortical delta, subcortical beta, cortical low gamma (31–50 Hz), and subcortical low gamma power from the NREM-to-wake transition data. Data processing was the same as described previously for the NREM-to-wake transition analysis. During each NREM-to-wake transition, NREM data from the start of NREM sleep to 40 seconds after the wakefulness state and all data after the wakefulness phase were used for model training. The average power for deep NREM and the wakefulness phase was calculated as previously described. Five QDA models were trained for all five participants. A uniform prior distribution was assumed during training. No rating transformation was applied. 10x cross-validation was performed to observe performance. To bias the model towards predicting pre-wake events, data were labeled as follows: 0 represents all data 5 seconds before wakefulness, and 1 represents the remaining data, with data from -5 seconds (5 seconds before wakefulness) given a greater weight (10x) compared to all other data during QDA training. The threshold for binary classification is 0.5, and no further threshold optimization was performed.

[0470] Statistical methods

[0471] A significance threshold of 0.05 was used to determine statistical significance. A linear mixed-effects model (LME) ("fitlme" in Matlab) was used to investigate differences in spectral power and coherence, as well as the interaction between cortical and subcortical β and cortical δ power. Theoretical likelihood ratio tests ("compare" in Matlab) were used to compare LME models. The Wilcoxon rank-sum test ("ranksum" in Matlab) was used to measure group-level differences in wakefulness prediction. All analyses were performed using Matlab 2022a (Mathworks).

[0472] Table 1: Demographic characteristics of participants

[0473] The UPDRS score was preoperative, Dx = duration of disease (years), RBD = REM sleep behavior disorder, obstructive sleep apnea = OSA, PD = Parkinson's disease, C-Ldopa = carbidopa-levodopa (Sinemet), A-HCl = amantadine hydrochloride (Symmetrel). None of them had dementia.

[0474] Supplementary Table 1: Sleep Statistical Data

[0475] SO = Time of sleep onset; WASO = Wakefulness after sleep onset; Total duration of N1, N2, N3, REM, and N2+N3 per night, in minutes; TST = Total sleep time; Wakefulness events = Total wakefulness events during a night.

[0476] Stimulation on includes the average sleep measurement over 11 nights recorded at home. Stimulation off includes one night recorded at home without stimulation.

[0477] Supplementary Table 2: Performance of Awake Prediction

[0478] Performance of individual QDA models for binary classification. PPV = Positive Predictive Value, NPV = Negative Predictive Value, U-test = Wilcoxon rank-sum test, AUC = Area under the acceptor operating characteristic curve.

[0479] References

[0480] 1. Videnovic A, Golombek D. Circadian and sleep disorders in Parkinson’s disease. Exp Neurol. 5 / 2013;243:45-56.

[0481] 2. Barone P, Antonini A, Colosimo C et al., The PRIAMO study: A multicenter assessment of nonmotor symptoms and their impact on quality of life in Parkinson’s disease. Mov Disord. 2009;24(11):1641-1649.

[0482] 3. Schreiner SJ, Imbach LL, Werth E, et al. Slow‐wave sleep and motor progression in Parkinson disease. Ann Neurol. 05 / 2019;85(5):765-770.

[0483] 4. Bartolomei L, Pastore A, Meligrana L, et al. Relevance of sleep quality on caregiver burden in Parkinson’s disease. Neurol Sci. 5 / 2018;39(5):835-839.

[0484] 5. Karlsen KH, Larsen JP, Tandberg E, Maeland JG. Influence of clinical and demographic variables on quality of life in patients with Parkinson’s disease. J Neurol Neurosurg Psychiatry. 1999;66(4):431-435.

[0485] 6. Rutten S, Vriend C, van der Werf YD, Berendse HW, Weintraub D, vanden Heuvel OA. The bidirectional longitudinal relationship between insomnia, depression and anxiety in patients with early-stage, medication-naïve Parkinson’s disease. Parkinsonism Relat Disord. 2017;39:31-36.

[0486] 7. Suzuki K, Miyamoto M, Miyamoto T, et al. Correlation between depressive symptoms and nocturnal disturbances in Japanese patients with Parkinson’s disease. Parkinsonism Relat Disord. 2009;15(1):15-19.

[0487] 8. Stavitsky K, Neargarder S, Bogdanova Y, McNamara P, Cronin-Golomb A. The impact of sleep quality on cognitive functioning in Parkinson’s disease. J Int Neuropsychol Soc. 2012;18(1):108-117.

[0488] 9. Breen DP, Vuono R, Nawarathna U, et al. Sleep and Circadian Rhythm Regulation in Early Parkinson Disease. JAMA Neurol. 2014;71(5):589.

[0489] 10. Postuma RB, Bertrand JA, Montplaisir J, et al. Rapid eye movementsleep behavior disorder and risk of dementia in Parkinson’s disease:Aprospective study:RBD and Parkinson Dementia.Mov Disord.05 / 2012;27(6):720-726.

[0490] 11. Diederich N, Vaillant M, Mancuso G, Lyen P, Tiete J. Progressivesleep “destructuring” in Parkinson’s disease.A polysomnographic study in 46patients.Sleep Med.07 / 2005;6(4):313-318.

[0491] 12. Deane KHO, Flaherty H, Daley DJ, et al. Priority settingpartnership to identify the top 10 research priorities for the management ofParkinson’s disease.BMJ Open.12 / 2014;4(12):e006434.

[0492] 13. Berry RB, Brooks R, Gamaldo CE, Harding SM, Lloyd RM, Marcus CL andVaughn BV on behalf of the American Academy of Sleep Medicine.The AASM Manual for theScoring of Sleep and Associated Events:Rules, Terminology and TechnicalSpecifications, Version 2.6,.Published online in 2020. aasmnet.org

[0493] [ PubMed ] 14.Amato N, Manconi M, Möller JC. Levodopa-induced dyskinesiain Parkinson disease:Sleep matters:L-dopa-Induced Dyskinesia.Ann Neurol.12 / 2018;84(6):905-917.

[0494] 15.Nassan M, Videnovic A. Circadian rhythms in neurodegenerative disorders.Nat Rev Neurol. doi:10.1038 / s41582-021-00577-7

[0495] [ PubMed ] 16.Beudel M, Oswal A, Jha A, et al. Oscillatory beta power correlateswith akinesia-rigidity in the parkinsonian subthalamic nucleus.MovDisord.2017;32(1):174-175.

[0496] 17.Mizrahi-Kliger AD, Kaplan A, Israel Z, Deffains M, Bergman H.Basal ganglia beta oscillations during sleep underlying Parkinsonian insomnia.:10.

[0497] 18.Urrestarazu E, Iriarte J, Alegre M, Other. Beta activity in the subthalamic nucleus during sleep in patients with Parkinson's disease.MovDisord.2009;24(2):254-260.

[0498] 19. Thompson JA, Tekriwal A, Felsen G, et al. Sleep patterns in Parkinson’s disease: direct recordings from the subthalamic nucleus. J Neurol Neurosurg Psychiatry. 2018;89(1):95 - 104.

[0499] 20. Christensen E, Abosch A, Thompson JA, Zylberberg J. Inferring sleep stage from local field potentials recorded in the subthalamic nucleus of Parkinson’s patients. J Sleep Res. 2019;28(4):1 - 6.

[0500] 21. Yin Z, Jiang Y, Merk T, et al. Pallidal activities during sleep and sleep decoding in dystonia, Huntington’s, and Parkinson's disease. Neurobiol Dis. 2023;182:106143.

[0501] 22. van Rheede JJ, Feldmann LK, Busch JL, et al. Diurnal modulation of subthalamic beta oscillatory power in Parkinson’s disease patients during deep brain stimulation. NPJ Parkinsons Dis. 2022;8(1):88.

[0502] 23. Gilron R ’ee, Little S, Perrone R, et al. Long-term wireless streaming of neural recordings for circuit discovery and adaptive stimulation in individuals with Parkinson’s disease. Nat Biotechnol. 09 / 2021;39(9):1078-1085.

[0503] 24. Mizrahi-Kliger AD, Kaplan A, Israel Z, Bergman H. Desynchronization of slow oscillations in the basal ganglia during natural sleep. Proc Natl Acad Sci U S A. 2018;115(18):E4274-E4283.

[0504] 25. Hammond C, Bergman H, Brown P. Pathological synchronization in Parkinson’s disease: networks, models and treatments. Trends Neurosci. 2007;30(7):357-364.

[0505] 26. Turner RS, Desmurget M. Basal ganglia contributions to motor control: a vigorous tutor. Curr Opin Neurobiol. 2010;20(6):704-716.

[0506] 27. Quinn EJ, Blumenfeld Z, Velisar A, et al. Beta oscillations in freely moving Parkinson’s subjects are attenuated during deep brain stimulation. Mov Disord. 2015;30(13):1750-1758.

[0507] 28. Baumann-Vogel H, Imbach LL, Sürücü O, et al. The Impact of Subthalamic Deep Brain Stimulation on Sleep-Wake Behavior: A Prospective Electrophysiological Study in 50 Parkinson Patients. Sleep. 2017;40(5). doi:10.1093 / sleep / zsx033

[0508] 29. Fleming JE, Kremen V, Gilron R, et al. Embedding Digital Chronotherapy into Bioelectronic Medicines. iScience. Published online March 4, 2022:104028.

[0509] 30. Gilron R ’ee, Little S, Wilt R, Perrone R, Anso J, Starr PA. Sleep-Aware Adaptive Deep Brain Stimulation Control: Chronic Use at Home With Dual Independent Linear Discriminate Detectors. Front Neurosci. 2021;15:732499.

[0510] 31. Arnal PJ, Thorey V, Debellemaniere E, et al. The Dreem Headband compared to polysomnography for electroencephalographic signal acquisition and sleep staging. Sleep. 2020;43(11):zsaa097.

[0511] 32. Postuma RB, Berg D, Stern M, et al. MDS clinical diagnostic criteria for Parkinson’s disease. Mov Disord. 2015;30(12):1591 - 1601.

[0512] 33. Arnal PJ, Thorey V, Ballard ME, et al. The Dreem Headband as an Alternative to Polysomnography for EEG Signal Acquisition and Sleep Staging. Neuroscience; 2019. Accessed October 2, 2021. biorxiv.org / lookup / doi / 10.1101 / 662734

[0513] 34. Neumann WJ, Memarian Sorkhabi M, Benjaber M, et al. The sensitivity of ECG contamination to surgical implantation site in brain computer interfaces. Brain Stimul. 09 / 2021;14(5):1301 - 1306.

[0514] 35. Hammer LH, Kochanski RB, Starr PA, Little S. Artifact Characterization and a Multipurpose Template - Based Offline Removal Solution for a Sensing - Enabled Deep Brain Stimulation Device. Stereotact Funct Neurosurg. 2022;100(3):168 - 183.

[0515] 36. Debellemaniere E, Chambon S, Pinaud C, et al. Performance of an Ambulatory Dry - EEG Device for Auditory Closed - Loop Stimulation of Sleep Slow Oscillations in the Home Environment. Front Hum Neurosci. 2018;12:88.

Claims

1. A method for treating sleep dysfunction in a subject, the method comprising: The first electrode is placed at a first location in the basal ganglia region or cortical region of the subject's brain to deliver electrical stimulation to the basal ganglia region or the cortical region; The second electrode is placed at a second location in the subcortical or cortical region of the subject's brain to record electroencephalogram (EEG) signal data while the subject is sleeping. The second electrode is used to detect electroencephalogram (EEG) signals associated with sleep features or sleep stages of interest; as well as When the second electrode detects an EEG signal associated with the sleep feature or sleep stage of interest, electrical stimulation is applied to the basal ganglia region or cortical region of the subject's brain using the first electrode in a manner that effectively treats the subject's sleep dysfunction.

2. The method according to claim 1, wherein the electroencephalogram (EEG) signal data includes field potential data.

3. The method according to claim 1 or 2, wherein the basal ganglia region is the subthalamic nucleus region, the globus pallidus region, or the thalamic region.

4. The method according to any one of claims 1-3, wherein the cortical region is the precentral gyrus or the postcentral gyrus.

5. The method according to any one of claims 1-4, wherein the sleep stage of interest is N2, N3 or REM.

6. The method according to any one of claims 1-5, the method further comprising using an accelerometer in conjunction with the electroencephalogram (EEG) signals to identify the sleep feature or sleep stage of interest.

7. The method according to any one of claims 1-6, the method further comprising using autonomic nervous data in conjunction with the electroencephalogram (EEG) signals to identify the sleep feature or sleep stage of interest.

8. The method according to any one of claims 1-7, the method further comprising using electroencephalography, polysomnography, noninvasive sleep monitoring device, wearable sleep monitoring device, photoplethysmography (PPG) based sleep monitoring device or radar-based sleep monitoring device to identify the sleep feature or sleep stage of interest.

9. The method according to any one of claims 1-8, the method further comprising generating a sleep structure map.

10. The method according to any one of claims 1-9, the method further comprising, when an EEG signal associated with the sleep feature of interest or the sleep stage is detected, using a control algorithm to automate the application of electrical stimulation.

11. The method of claim 10, wherein the control algorithm uses a machine learning algorithm for sleep feature or sleep stage classification.

12. The method of claim 11, wherein the machine learning algorithm is a supervised machine learning algorithm.

13. The method according to any one of claims 10-12, wherein the control algorithm further adjusts one or more programmed stimulation parameters to maximize slow-wave activity.

14. The method of claim 13, wherein the slow wave activity is in the frequency range of 0.5 Hz to 4 Hz.

15. The method according to any one of claims 10-14, wherein the control algorithm further uses linear discriminant analysis (LDA) to adjust the stimulation amplitude or frequency of the electrical stimulation.

16. The method of claim 15, wherein the stimulation amplitude is optimized during the N3 sleep stage to maximize slow-wave activity.

17. The method according to any one of claims 1-16, wherein the electrical stimulation is applied unilaterally or bilaterally.

18. The method according to any one of claims 1-17, wherein the electroencephalogram (EEG) signal comprises neural oscillations of beta frequency, gamma frequency, delta frequency, or theta frequency.

19. The method according to any one of claims 5-18, wherein the N3 sleep stage is identified by an increase in δ power during the N3 sleep stage compared to when the subject is awake.

20. The method according to any one of claims 5-19, wherein the N2 sleep stage or the N3 sleep stage is identified by one or more spectral power variations selected from: a decrease in β power in a frequency range of 12 Hz to 30 Hz compared to β power when the subject is awake; a decrease in γ power in a frequency range of 30 Hz to 60 Hz compared to γ ​​power when the subject is awake; an increase in θ power in a frequency range of 5 Hz to 10 Hz compared to θ power when the subject is awake; and an increase in δ power in a frequency range of 0.5 Hz to 4.5 Hz compared to δ power when the subject is awake.

21. The method of claim 20, wherein the N2 sleep stage or the N3 sleep stage is identified by means of the one or more spectral power changes combined with the detection of one or more changes in cortical-subcortical spectral coherence, the one or more changes in cortical-subcortical spectral coherence being selected from: an increase in the δ-cortical-subcortical spectral coherence compared to the δ-cortical-subcortical spectral coherence when the subject is awake; and a decrease in the β-cortical-subcortical spectral coherence compared to the β-cortical-subcortical spectral coherence when the subject is awake.

22. The method according to any one of claims 1-21, wherein the second electrode is placed on the surface of the subcortical region or the cortical region.

23. The method according to any one of claims 1-22, wherein the first electrode is a non-brain-penetrating surface electrode array or a brain-penetrating electrode array.

24. The method according to any one of claims 1-23, wherein the second electrode is a non-brain-penetrating surface electrode array or a brain-penetrating electrode array.

25. The method according to any one of claims 1-24, wherein the second electrode is an electroencephalogram (EEG) electrode array, a neurostimulator electrode installed under the galea aponeurotica or through a burr hole in the skull or installed on the cranial side, or an electrocorticography (ECoG) electrode array.

26. The method of claim 25, wherein the ECoG electrode array spans the anterior central gyrus and the posterior central gyrus.

27. The method according to any one of claims 1-26, wherein the sleep dysfunction is caused by a motor disorder or a neurological condition, wherein applying the electrical stimulation improves sleep.

28. The method of claim 27, wherein the movement disorder is Parkinson's disease.

29. The method of claim 27 or 28, wherein the subject is further subjected to daytime neural stimulation.

30. The method of claim 28 or 29, wherein the subject is further administered a dopaminergic drug.

31. The method according to any one of claims 1-30, the method further comprising evaluating the efficacy of treatment for the sleep dysfunction of the subject.

32. The method of claim 31, wherein the assessment comprises using the Visual Analogue Scale (VAS), Likert Scale, Stanford Somnolence Scale (SSS), Maintain Wakefulness Test (MWT), Appleworth Somnolence Scale (ESS), Multiple Sleep Latency Test (MSLT) or Athens Insomnia Scale.

33. The method of claim 31 or 32, wherein the assessment includes monitoring the subject using a motion recorder, electroencephalogram (EEG), or polysomnography.

34. The method according to any one of claims 1-33, the method further comprising mapping the subject's brain to identify optimal locations in the subcortical region or the cortical region to detect the electroencephalogram (EEG) signals associated with the sleep characteristics or the sleep stage.

35. The method of claim 34, wherein the cortical region is the precentral gyrus or the postcentral gyrus.

36. The method according to any one of claims 1-35, the method further comprising segmenting the recorded electroencephalogram (EEG) signal data into consecutive time periods.

37. The method of claim 36, the method further comprising assigning sleep characteristics or sleep stage tags to each time period.

38. The method of claim 36 or 37, wherein each time period comprises 0.5 seconds to 1 minute of recorded EEG signal data.

39. The method according to any one of claims 1-38, wherein the method is performed while the subject is sleeping at home, in a sleep laboratory, or in a hospital.

40. The method according to any one of claims 1-39, wherein the sleep stage is N1, N2, N3 or phase- or tonic-torsional rapid eye movement (REM).

41. The method according to any one of claims 1-40, wherein the sleep characteristics are slow waves, sleep spindle waves, K complex waves, beta bursts, pre-wake period, wake period, post-wake period, or sleep stage transitions.

42. The method of claim 41, wherein the pre-wake period or the wake period is identified by one or more spectral power variations selected from: a decrease in cortical delta power in a frequency range of 1 Hz to 4 Hz compared to the average cortical delta power during deep non-rapid eye movement (NREM) sleep; an increase in cortical gamma power in a frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power during deep NREM sleep; an increase in subcortical gamma power in a frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power during deep NREM sleep; and an increase in subcortical beta power in a frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power during deep NREM sleep.

43. The method of claim 42, wherein the increase in subcortical β power precedes the decrease in cortical δ power.

44. The method according to any one of claims 41-43, wherein the post-awakening time period is identified by one or more spectral power variations selected from: a decrease in cortical delta power in a frequency range of 1 Hz to 4 Hz compared to the average cortical delta power in the pre-awakening time period; an increase in cortical gamma power in a frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power in the pre-awakening time period; an increase in subcortical gamma power in a frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power in the pre-awakening time period; and an increase in subcortical beta power in a frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power in the pre-awakening time period.

45. The method according to any one of claims 1-44, wherein the electrical stimulation increases cortical delta power, decreases cortical alpha power, decreases cortical beta power, and decreases cortical σ power.

46. ​​The method according to any one of claims 1-45, wherein the electrical stimulation reduces cortical-subcortical σ-spectral coherence.

47. The method according to any one of claims 1-46, the method further comprising: The second electrode is used to detect EEG signals associated with one or more additional sleep features or sleep stages of interest; as well as When the second electrode detects the EEG signal associated with one or more additional sleep features or sleep stages of interest, the first electrode is used to apply electrical stimulation to the basal ganglia region or the cortical region of the subject's brain in a manner that effectively treats the subject's sleep dysfunction.

48. A computer-implemented method for programming a deep brain stimulation (DBS) device to treat sleep dysfunction in a subject, the computer performing the following steps: a) While the subject is sleeping, receive recorded electroencephalogram (EEG) signal data from the subcortical or cortical regions of the subject's brain; b) Use a classification model to analyze the recorded EEG signal data, the classification model identifying patterns of electrical signals in the recorded EEG signal data that are associated with sleep features or sleep stages of interest; c) Adjust one or more programmed stimulation parameters based on the recorded EEG signal data according to the algorithmic control rules; and d) When the sleep feature or sleep stage of interest is detected, instruct the DBS device to apply electrical stimulation to the basal ganglia region or cortical region of the subject's brain to treat the subject's sleep dysfunction.

49. The computer-implemented method of claim 48, wherein the electroencephalogram (EEG) signal data includes field potential data.

50. The computer-implemented method of claim 48 or 49, wherein a machine learning algorithm is used to generate the classification model.

51. The computer-implemented method of claim 50, wherein the machine learning algorithm is a supervised machine learning algorithm.

52. The computer-implemented method according to any one of claims 48-51, wherein the basal ganglia region is the subthalamic nucleus (STN) region, the globus pallidus region, or the thalamic region.

53. The computer-implemented method according to any one of claims 48-52, wherein the cortical region is the precentral gyrus or the postcentral gyrus.

54. The computer-implemented method according to any one of claims 48-53, wherein the sleep stage is N2, N3, or rapid eye movement (REM).

55. The computer-implemented method according to any one of claims 48-54, wherein the computer-implemented method further comprises: Accelerometer data of the subject are received while the subject is sleeping; as well as The classification model is used to analyze the accelerometer data combined with the recorded electroencephalogram (EEG) signal data to identify the sleep characteristics or sleep stages.

56. The computer-implemented method according to any one of claims 48-55, wherein the computer-implemented method further comprises: The subject's autonomic nervous data was received while the subject was sleeping; as well as The classification model is used to analyze the autonomic nervous data combined with the recorded electroencephalogram (EEG) signal data to identify the sleep characteristics or sleep stages.

57. The computer-implemented method according to any one of claims 48-56, wherein the computer-implemented method further comprises: The subject's electroencephalogram (EEG) or polysomnography (PSG) is received while the subject is sleeping. as well as The classification model is used to analyze the electroencephalogram (EEG) or polysomnography (PSG) combined with the recorded EEG signal data to identify the sleep characteristics or sleep stages.

58. The computer-implemented method according to any one of claims 48-57, the computer-implemented method further comprising receiving data from a non-invasive sleep monitoring device, a wearable sleep monitoring device, a photoplethysmography (PPG)-based sleep monitoring device, or a radar-based sleep monitoring device; and using the classification model to analyze the data to identify the sleep characteristics or the sleep stages.

59. The computer-implemented method according to any one of claims 48-58, the computer-implemented method further comprising generating a sleep structure diagram.

60. The computer-implemented method according to any one of claims 48-59, wherein the classification model is trained to identify the sleep characteristics or the sleep stages by analyzing electroencephalogram (EEG) signal data recorded over multiple nights while the subject is sleeping.

61. The computer-implemented method according to any one of claims 48-60, wherein the computer-implemented method further comprises: a) Using a linear classification model, the predicted stimulus efficacy of the available settings of the DBS device is ranked based on the classifier score of the stimulus efficacy for each setting. b) Select the stimulus setting predicted to have the highest stimulus efficacy based on the linear classification model; c) After applying electrical stimulation to the basal ganglia region or the cortical region of the subject's brain using the DBS device in the setting predicted to have the highest stimulating efficacy, receiving the recorded electroencephalogram (EEG) signal data from the subcortical region or the cortical region of the subject's brain. d) Analyze the recorded electroencephalogram (EEG) data to assess the subject's neural response to the electrical stimulation; e) Update the linear classification model based on the subject's neural response to the electrical stimulation to generate an updated linear classification model; f) Use the updated linear classification model to update the ranking of the predicted stimulus efficacy for the available settings of the DBS device; g) Select the stimulus setting predicted to have the highest stimulus efficacy based on the updated linear classification model; h) After applying the electrical stimulation to the basal ganglia region or the cortical region of the subject's brain using the DBS device based on the updated linear classification model and the setting predicted to have the highest stimulation efficacy, the recorded electroencephalogram (EEG) signal data is received from the subcortical region or the cortical region of the subject's brain. as well as i) Repeat e)-h) to adjust the available settings of the DBS device to optimize stimulation efficacy.

62. The computer-implemented method of claim 61, wherein the linear classification model uses linear discriminant analysis (LDA) to adjust the current amplitude and frequency of the electrical stimulation.

63. The computer-implemented method of claim 62, wherein the stimulation amplitude is optimized during the N3 sleep stage to maximize slow-wave activity.

64. The computer-implemented method of claim 63, wherein the slow-wave activity is in the frequency range of 0.5 Hz to 4 Hz.

65. The computer-implemented method according to any one of claims 48-64, the computer-implemented method further comprising segmenting the recorded electroencephalogram (EEG) signal data into consecutive time periods.

66. The computer-implemented method of claim 65, further comprising assigning sleep characteristics or sleep stage tags to each time period.

67. The computer-implemented method of claim 65, wherein each time period comprises 0.5 seconds to 1 minute of recorded EEG signal data.

68. The computer-implemented method according to any one of claims 61-67, the computer-implemented method further comprising training a linear model by analyzing the recorded electroencephalogram (EEG) signal data using a nonlinear model during all sleep stages while the subject is sleeping, to classify each time period as an N3 sleep stage period or a non-N3 sleep stage period.

69. The computer-implemented method of claim 68, wherein the standardized δ and β power frequency bands are used as feature inputs to train the linear classification model using linear discriminant analysis to classify each time period as an N3 sleep stage period or a non-N3 sleep stage period.

70. The computer-implemented method of claim 68, wherein subcortical field potentials are used as feature inputs to train the linear classification model using linear discriminant analysis to classify each time period as an N3 sleep stage period or a non-N3 sleep stage period.

71. The computer-implemented method according to any one of claims 48-70, wherein the electroencephalogram (EEG) signal data includes field potential data.

72. The computer-implemented method according to any one of claims 48-71, the computer-implemented method further comprising storing a user profile of the subject, the user profile including information about recorded electroencephalogram (EEG) signal data associated with the sleep characteristics or the sleep stage.

73. The computer-implemented method according to any one of claims 48-72, the computer-implemented method further comprising storing a user profile of the subject, the user profile including information about the programmed stimulation parameters for applying electrical stimulation to the basal ganglia region or the cortical region of the subject's brain based on recorded electroencephalogram (EEG) signal data to treat the subject's sleep dysfunction.

74. The computer-implemented method according to any one of claims 48-73, wherein the classification model identifies the N2 sleep stage or the N3 sleep stage by one or more spectral power variations, the one or more spectral power variations being selected from: a decrease in β power in a frequency range of 12 Hz to 30 Hz compared to β power when the subject is awake; a decrease in γ power in a frequency range of 30 Hz to 60 Hz compared to γ ​​power when the subject is awake; an increase in θ power in a frequency range of 5 Hz to 10 Hz compared to θ power when the subject is awake; and an increase in δ power in a frequency range of 0.5 Hz to 4.5 Hz compared to δ power when the subject is awake.

75. The computer-implemented method of claim 74, wherein the classification model identifies the N2 sleep stage or the N3 sleep stage by combining the detection of one or more variations in cortical-subcortical spectral coherence with the one or more variations in cortical-subcortical spectral coherence, wherein the one or more variations in cortical-subcortical spectral coherence are selected from: an increase in δ-cortical-subcortical spectral coherence compared to δ-cortical-subcortical spectral coherence when the subject is awake; and a decrease in β-cortical-subcortical spectral coherence compared to β-cortical-subcortical spectral coherence when the subject is awake.

76. The computer-implemented method according to any one of claims 48-75, wherein the sleep characteristics are slow waves, sleep spindle waves, K-complex waves, beta bursts, pre-wake period, wake period, post-wake period, or sleep stage transitions.

77. The computer-implemented method of claim 76, wherein the classification model identifies the pre-wake period or the wake period by one or more spectral power variations, the one or more spectral power variations being selected from: a decrease in cortical delta power in a frequency range of 1 Hz to 4 Hz compared to the average cortical delta power during deep non-rapid eye movement (NREM) sleep; an increase in cortical gamma power in a frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power during deep NREM sleep; an increase in subcortical gamma power in a frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power during deep NREM sleep; and an increase in subcortical beta power in a frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power during deep NREM sleep.

78. The computer-implemented method of claim 77, wherein the increase in subcortical β power precedes the decrease in cortical δ power.

79. The computer-implemented method according to any one of claims 76-78, wherein the classification model identifies the post-awakening time period by one or more spectral power variations, the one or more spectral power variations being selected from: a decrease in cortical delta power in a frequency range of 1 Hz to 4 Hz compared to the average cortical delta power in the pre-awakening time period; an increase in cortical gamma power in a frequency range of 31 Hz to 50 Hz compared to the average cortical gamma power in the pre-awakening time period; an increase in subcortical gamma power in a frequency range of 31 Hz to 50 Hz compared to the average subcortical gamma power in the pre-awakening time period; and an increase in subcortical beta power in a frequency range of 13 Hz to 31 Hz compared to the average subcortical beta power in the pre-awakening time period.

80. A non-transitory computer-readable medium comprising program instructions that, when executed by a processor in a computer, cause the processor to perform the method according to any one of claims 48-79.

81. A kit comprising the non-transitory computer-readable medium according to claim 80 and instructions for using a deep brain stimulation device to treat a subject's sleep dysfunction.

82. A system for treating sleep dysfunction in a subject, the system comprising: A first electrode, adapted to be placed in a location in the basal ganglia region or cortical region of the subject's brain, to deliver electrical stimulation to the basal ganglia region or the cortical region; A second electrode, adapted to be placed in a subcortical or cortical region of the subject's brain, to record electroencephalogram (EEG) signal data while the subject is sleeping; and A processor, programmed to perform a computer-implemented method according to any one of claims 48-79, for instructing the first electrode to apply electrical stimulation to the basal ganglia region or the cortical region of the subject's brain in a manner that effectively treats the subject's sleep dysfunction when an electroencephalogram (EEG) signal associated with the sleep feature or sleep stage of interest is detected using the second electrode.

83. The system of claim 82, wherein the electroencephalogram (EEG) signal data includes field potential data.

84. The system according to claim 82 or 83, wherein the basal ganglia region is the subthalamic nucleus region, the globus pallidus region, or the thalamic region.

85. The system according to any one of claims 82-84, wherein the cortical region is the precentral gyrus or the postcentral gyrus.

86. The system according to any one of claims 82-85, wherein the sleep stage of interest is N2, N3 or REM.

87. The system according to any one of claims 82-86, the system further comprising an accelerometer for recording the subject's motion while the subject is sleeping.

88. The system according to any one of claims 82-87, the system further comprising a non-invasive sleep monitoring device, a wearable sleep monitoring device, a sleep monitoring device based on photoplethysmography (PPG), or a radar-based sleep monitoring device.

89. The system according to any one of claims 82-88, wherein the first electrode is a non-brain-penetrating surface electrode array or a brain-penetrating electrode array.

90. The system according to any one of claims 82-89, wherein the second electrode is a non-brain-penetrating surface electrode array or a brain-penetrating electrode array.

91. The system according to any one of claims 82-90, wherein the second electrode is an electroencephalogram (EEG) electrode array, a neurostimulator electrode installed under the galea aponeurotica or through a drill hole in the skull or installed on the cranial side, or an electrocorticography (ECoG) electrode array.

92. The system of claim 91, wherein the ECoG electrode array spans the anterior central gyrus and the posterior central gyrus.

93. The system according to any one of claims 82-92, wherein the sleep dysfunction is caused by a motor disorder or a neurological condition, wherein applying the electrical stimulation improves sleep.

94. The system of claim 93, wherein the movement disorder is Parkinson's disease.

95. The system according to any one of claims 82-94, wherein the system further comprises a user interface, the user interface including an input terminal electrically coupled to the processor for instructing the first electrode to apply electrical stimulation to the basal ganglia region or the cortical region to treat the sleep dysfunction of the subject.

96. The system of claim 95, wherein the user interface is password protected and can be operated by a healthcare practitioner.

Images