Closed-loop neuromodulation of sleep-related oscillations

Abstract

Provided herein are systems and methods for inducing sleep in a subject using closed-loop neuromodulation. The methods can include determining a fundamental center frequency and an evoked response delay from oscillation data gathered by one or more sensors configured to detect brain waves of a subject. The methods can include determining a stimulation phase based on the evoked response delay and the fundamental center frequency. Using the stimulation phase, a phase-locked stimulus can be delivered to induce sleep in the subject.

Description

CLOSED-LOOP NEUROMODULATION OF SLEEP-RELATED OSCILLATIONSCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to and the benefit of U.S. Provisional Patent Application No. 63 / 432,891, filed December 15, 2022, the entire contents of which are hereby incorporated by reference.FIELD

[0002] This disclosure relates generally to neuromodulation of sleep-related oscillations, and more specifically to systems, methods, and devices for closed-loop neuromodulation of sleep-related oscillations to induce sleep in a subject.BACKGROUND

[0003] A significant fraction of an individual’s lifetime is occupied by sleep. The physiological role of sleep is not fully understood, but may encompass several processes necessary for life, including cellular and genomic maintenance, removal of waste products, memory consolidation, and synthesis of essential molecules. Despite its importance, the 12- month incidence of insomnia for adults in the United States, defined as repeated issues with sleep quality that result in daytime impairment, is 27.3% with an associated loss in quality- adjusted life-years (QUALYs) of 5.6 million. A number of pharmacological interventions exist to address insomnia, including benzodiazepines, antidepressants, and antihistamines. However, many such drugs are associated with a significant risk of harm relative to placebo and can be habit-forming. An additional barrier to the development of effective treatments for sleep disorders is the difficulty of objectively measuring sleep quality and quantity.Polysomnography is common for the diagnosis of several forms of disordered sleep, but requires experienced operators and often subjects must commit to a multi-night stay in a sleep laboratory. Miniaturization, wireless systems, ambulatory or “home testing”, and data-centric approaches have been identified as key features of new technologies to advance the field of sleep medicine.

[0004] Existing approaches have demonstrated sleep-related effects by delivering auditory stimulation locked to the phase of slow-wave oscillations (approximately 0.4-4 Hz), typically associated with N3 (“Slow Wave”) sleep. However, these approaches do not address sleep onset latency (SOL), which is a specific outcome measure identified by the American Association of Sleep Medicine as an important target for insomnia treatments.

[0005] Exemplary neuromodulation systems and methods are described in greater detail in U.S. Patent No. 11,166,632 and International Publication No. WO 2022 / 159428, the entire contents of each of which are incorporated herein by reference.SUMMARY

[0006] Described herein are systems, devices, and methods of use thereof for inducing sleep using auditory neurostimulation. The neuromodulation systems can measure and record neural activity, process the neural oscillation data to determine a fundamental center frequency and evoked response delay of the oscillation data, and determine a stimulation phase that is phase-locked to the oscillation data. The systems and methods may implement an endpoint-corrected Hilbert Transform (ecHT) algorithm to process the oscillation data and enables determination of key parameters from the data. The systems can deliver auditory stimulation based on the stimulation phase to induce sleep. The auditory stimulation can be delivered in-phase or anti-phase with the oscillations. In one example, the neurostimulation systems and methods described herein can be particularly used to measure and analyze alpha oscillations, which are understood to be related to insomnia. By delivering acoustic neuromodulation based on the intrinsic alpha frequency and evoked response delay of the subject, sleep onset latency can be reduced in subjects experiencing insomnia. The systems described herein can be embodied in an ambulatory wearable device, such as a headband, to track electroencephalography (EEG) data in real-time (or near-real time) and apply a phase- locked stimulus based on the data.

[0007] In some examples, a method for inducing sleep in a subject using neuromodulation is provided, comprising: receiving oscillation data from one or more sensors configured to detect brain waves from the subject; determining a fundamental center frequency of the oscillation data; determining an evoked response delay of the oscillation data; based on the determined fundamental center frequency and the determined evoked response delay, determining a stimulation phase for the subject; and delivering a phase- locked stimulus in accordance with the stimulation phase to induce sleep in the subject.

[0008] In some examples, a method for determining an intrinsic center frequency (iCF) of a subject is provided, comprising: receiving oscillation data from one or more sensors configured to detect brain waves; determining spectral data of the received data, wherein the spectral data comprises a noise floor; generating de-noised spectral data by subtracting the noise floor from the spectral data; and identifying a peak within a portion of the de-noised spectral data, the peak corresponding to the intrinsic center frequency (iCF) of the subject.

[0009] In some examples, a system for inducing sleep in a subject using neuromodulation is provided, comprising: one or more sensors configured to detect brain waves from the subject; a processing device configured to execute any one or more of the aforementioned methods, and one or more stimulators configured to deliver a phase-locked stimulus in accordance with the stimulation phase to induce sleep in the subject.

[0010] In some examples, a non-transitory computer-readable storage medium is provided, the non-transitory computer-readable storage medium storing instructions that, when executed by a processing device operatively coupled to one or more sensors configured to detect brain waves of a subject, cause the device to execute any one or more of the aforementioned methods to induce sleep in the subject.BRIEF DESCRIPTION OF THE FIGURES

[0011] Various aspects of the disclosed systems and methods are set forth with particularity in the appended claims. A better understanding of the features and advantages of the disclosed systems and methods will be obtained by reference to the detailed description of illustrative embodiments and the accompanying drawings.

[0012] FIG. 1A illustrates an exemplary method for inducing sleep in a subject using neuromodulation, in accordance with some embodiments.

[0013] FIG. IB illustrates an exemplary method for determining a fundamental center frequency of oscillation data from a subject, in accordance with some embodiments.

[0014] FIG. 1C illustrates another exemplary method for determining an evoked response delay of oscillation data from a subject, in accordance with some embodiments.

[0015] FIG. 2 illustrates EEG data for determining the fundamental center frequency, in accordance with some embodiments.

[0016] FIG. 3 illustrates an exemplary neuromodulation system, in accordance with some embodiments.

[0017] FIG. 4 illustrates a system architecture of an exemplary device for inducing sleep in a subject using neuromodulation, in accordance with some embodiments.

[0018] FIG. 5A illustrates an exemplary wearable neuromodulation device, in accordance with some embodiments.

[0019] FIG. 5B illustrates another view of the exemplary wearable neuromodulation device, in accordance with some embodiments.

[0020] FIG. 6 illustrates an example user interface for calibrating a wearable neuromodulation device, in accordance with some embodiments.

[0021] FIG. 7 illustrates a diagram demonstrating the process of the endpoint-corrected Hilbert Transform (ecHT), in accordance with some embodiments.

[0022] FIG. 8A illustrates a graph of the evoked response potential in-phase with alpha oscillations of a subject, in accordance with some embodiments.

[0023] FIG. 8B illustrates a graph of the evoked response potential anti-phase with alpha oscillations of a subject, in accordance with some embodiments.

[0024] FIG. 9 illustrates recorded electroencephalogram (EEG) data, including four sleep stages, in accordance with some embodiments.

[0025] FIG. 10 illustrates a chart representative of the distribution of time spent in each sleep stage for one subject during an overnight recording, in accordance with some embodiments.

[0026] FIG. 11 A illustrates a diagram of the distribution of the target onset and offset phases near the alpha peak phase, in accordance with some embodiments.

[0027] FIG. 1 IB illustrates a diagram of the mean onset and offset phase values for stimulation targeting the trough phase of the alpha cycle, in accordance with some embodiments.

[0028] FIG. 11C illustrates a diagram of the mean onset and offset phase values for stimulation targeting the peak phase of the alpha cycle, in accordance with some embodiments.

[0029] FIG. 12A illustrates a graph of the effect of stimulation on broadband auditory evoked potentials, in accordance with some embodiments.

[0030] FIG. 12B illustrates a graph of the effect of stimulation on bandpassed auditory evoked potentials, in accordance with some embodiments.

[0031] FIG. 13 illustrates an at-home sleep study design for testing the neurostimulation devices and methods described herein, in accordance with some embodiments.

[0032] FIG. 14A illustrates a box plot demonstrating changes in sleep onset latency for subjects in control, trough, or peak stimulation conditions, in accordance with some embodiments.

[0033] FIG. 14B illustrates a box plot demonstrating changes in sleep onset latency for subjects with scorable datasets in control, trough, or peak stimulation conditions, in accordance with some embodiments.

[0034] FIG. 14C illustrates a box plot demonstrating sleep onset latency for subjects classified as “good sleepers” in control, trough, or peak stimulation conditions, in accordance with some embodiments.

[0035] FIG. 14D illustrates a box plot demonstrating sleep onset latency for subjects classified as “poor sleepers” in control, trough, or peak stimulation conditions, as measured by actigraphy, in accordance with some embodiments.

[0036] FIG. 14E illustrates a box plot demonstrating sleep onset latency for subjects classified as “poor sleepers” in control, trough, or peak stimulation conditions, as measured using a morning survey, in accordance with some embodiments.

[0037] FIG. 15 illustrates a computing device, in accordance with some embodiments.DETAILED DESCRIPTION

[0038] Described herein are systems, methods, and devices for modulating neural activity to induce sleep. The methods can include receiving oscillation data of a subject, determining a fundamental center frequency and an evoked response delay based on the oscillation data, determining a stimulation phase based on the fundamental center frequency and the evoked response delay, and delivering a phase-locked stimulus in accordance with the stimulation phase to the subject. The systems and methods described herein may be embodied in (and executable by) a wearable device, such as a headband device, configured to deliver auditory stimulation via a bone conductor. The systems and methods described herein may be applicable in treating insomnia, for example by targeting alpha oscillation data and delivering phase-locked stimulation to disrupt the elevated alpha activity, thereby accelerating sleep initiation.

[0039] In healthy sleepers, cortical alpha oscillations are present during the transition from wakefulness to sleep, and then dissipate at sleep onset. For individuals with insomnia, alpha power is elevated during the wake-sleep transition and can persist throughout the night. Neuromodulation techniques using phase-locked auditory stimulation to augment or suppress oscillations have been put forth as alternatives to drugs for improving sleep quality. This approach has been applied to slow oscillations present during deep sleep, but historically exhibit limited real-time phase detection accuracy for biological rhythms with higher frequencies, including alpha.

[0040] The following disclosure describes acoustic neuromodulation and its use as a treatment regimen for sleep insomnia. Also described are exemplary acoustic neuromodulation systems and devices configurable for execution, for example for performing the methods described herein. Feasibility studies of the acoustic neuromodulation systems and devices provided herein that track neural oscillations and deliver phase-locked neurostimulation based on the oscillation, and is also described by way of example. Finally, ageneral computing system that can be configured to execute the methods described herein is described.

[0041] Acoustic neuromodulation involves the delivery of sound pulses to modulate brain signals (i.e., otherwise referred to herein as brain waves). A neuromodulator can generate exogenous signals (e.g., acoustic pulses) that stimulate the neural tissue to suppress, augment, or otherwise maintain neural oscillations in the subject. An electroencephalogram (EEG) may measure brain signals (e.g., electrophysiological signals representative of neural tissue activity), and the acoustic stimulation may be delivered in accordance with these brain signals. In the context of sleep insomnia, the brain signals of interest may comprise alpha signals. However, the disclosure is not intended to be limited to these high-frequency alpha signals and may be usable to determine a phase-locked stimulus in various stages, such as using alpha, theta, delta, beta, and gamma signals. By characterizing the oscillations of the signals and delivering stimuli in accordance with the phase of the oscillations, sleep can be induced. Particularly, the acoustic pulses may be delivered such that they are locked to the phase of the oscillations measured by an EEG, hereinafter referred to as “phase-locked acoustic neuromodulation.” For example, in phase-locked acoustic neuromodulation, a difference between the phase of a first signal and the phase of a second signal may be constant (i.e., the two signals have a consistent phase relationship with each other).Methods for Inducing Sleep Using Acoustic Neuromodulation

[0042] FIGS. 1A-1C illustrate methods for inducing sleep in a subject using neuromodulation. As will be described in greater detail herein, method 100 may be executed by an acoustic neuromodulation system and device, such as system 300 illustrated in FIG. 3 and device 500 illustrated in FIG. 5.

[0043] Method 100 may include, at block 110, receiving oscillation data from one or more sensors configured to detect brain waves from the subject. The oscillation data may comprise electrophysiological signals that represent brain waves of the subject. For example, the oscillation data may comprise electroencephalogram (EEG) signals received from one or more sensors (e.g., external electrodes) of an EEG system. In some embodiments, the brain signal received by the system may comprise substantially all of the oscillation data detected by the one or more sensors. In some embodiments, the system may receive a portion of oscillation data from the one or more sensors because the sensors may be configured to begin detecting (and / or transmitting) oscillation data upon a determination that the subject is in and / or approaching a specific stage of sleep. For example, the oscillation data detected and / ortransmitted by the one or more sensors may comprise signals within a frequency band, such as between about 0.1-150 Hz. In some embodiments, the oscillation data may be alpha oscillation data occupying a frequency band between about 8-12 Hz, 7.5-14 Hz, etc. In some embodiments, the oscillation data may comprise at least one of alpha, theta (e.g., about 4-7 Hz), delta (e.g., about 0.5-3 Hz), beta (e.g., about 13-30 Hz), and gamma (e.g., about 30-150 Hz) oscillations. The oscillation data may reasonably comprise any neural oscillation that can be reliably captured using EEG. The frequency band may comprise frequencies corresponding to one or more sleep stage(s) of interest.

[0044] In some embodiments, the oscillation data received at block 110 may be pre- processed signal data or natural signal data. In other words, the oscillation data transmitted to the system executing method 100 may receive processed brain signal or natural signal data. The method 100 may comprise processing the data. Processing the oscillation data (e.g., the natural signal data) may comprise creating a discrete signal that is a sampling of the received oscillation data and based on the discrete signal, determining an analytical signal and Hilbert transform signal for the discrete signal. Determining the analytical signal and Hilbert transform signal can include utilizing an endpoint-corrected Hilbert transform (ecHT) algorithm, described in greater detail in U.S. Patent No. 11,166,632, incorporated by reference herein in its entirety.

[0045] The ecHT algorithm may comprise a frequency domain ecHT, end-padded time domain ecHT, or front-padded time domain ecHT. The ecHT algorithms described herein can correct (e.g., correct, prevent, mitigate, ameliorate, or otherwise compensate for) distortion in the oscillation data due to Gibbs phenomenon that occurs at the ends of sample data. The ecHT algorithms can be particularly useful for determining instantaneous attributes (e.g., instantaneous phase, amplitude, etc.) of a signal using a fast Fourier transform (FFT) and / or inverse fast Fourier transform (IFFT) procedures. FIG. 7, described in greater detail below, illustrates an example diagram summarizing the ecHT algorithm for computing the magnitude and phase of an analytic signal.

[0046] Each of the frequency domain ecHT, end-padded time domain ecHT, and front- padded time domain ecHT algorithms may include applying causal filter(s) to the discrete signal. The causal filter may comprise a causal infinite impulse response (IIR) filter or causal finite impulse response (FIR) filter. For example, the IIR or FIR filter may include but is not limited to Butterworth filters, Chebyshev filters, Bessel filters, etc. The causal filter may comprise a linear and time-invariant (ETI) bandpass filter. As described herein, the ecHT method may be used to determine instantaneous attributes of the oscillation data. In thismanner, the causal filter may reduce distortion of the instantaneous attributes (e.g., amplitude and / or phase) of the signal inside a passband and suppress frequencies outside of the passband. For example, the passband may comprise the range of frequencies of interest in the received brain signal.

[0047] At block 120, the method 100 may comprise determining a fundamental center frequency of the oscillation data. Determining the fundamental center frequency is described in greater detail with respect to FIG. IB and FIG. 2. The fundamental center frequency may be determined based on oscillation data received from one or more sensors while the subject is awake and has their eyes closed for a duration of time. For example, the subject may be in a resting state, and oscillation data may be gathered for about 1, 2, or 3 minutes. The oscillation data (e.g., EEG data) may be received from one or more channels or locations on the scalp, such as Fp or Fpz. Plot 200 in FIG. 2 illustrates a multitaper spectrogram comprising oscillation data computed for one representative subject during an approximately 2 minute passive recording session. The method at 120 can include determining spectral data 112, including a noise floor, from the received oscillation data. The spectral content of the time-series data (plot 200) may be analyzed using a Fast Fourier transform (FFT) or other similar algorithm. The noise floor can be modeled with a 3rd-order polynomial fit. For example graph 210 in FIG. 2 illustrates a median spectrum over the duration of the recording session shown in graph 200. The method at 120 can include generating de-noised spectral data by subtracting the noise floor from the spectral data. The method at 120 can include identifying a peak within a portion of the de-noised spectral data 114, the peak corresponding to the fundamental center frequency of the subject. The prominent peak (i.e., spectral peak) may be identified within a predefined frequency range, such as an alpha range between about 8-12 Hz, 7.5-14 Hz, etc. In graph 210 of FIG. 2, the vertical dashed line shows the location of the fundamental center frequency peak for this subject.

[0048] In some embodiments, the ecHT algorithm may be used to determine the fundamental center frequency of a signal within a specific range. For example, the ecHT algorithm may determine the local maximum of the signal in the frequency domain. This center frequency value can then be used as the fundamental center frequency of the causal bandpass filter. By determining the fundamental center frequency of the filter, any phase shift that may be introduced by the filter can be minimized. In a non-limiting example, an example peak frequency in the frequency spectrum of a brain signal between about 7.5 Hz and 12.5 Hz may be about 10 Hz. 10 Hz may be the intrinsic alpha frequency (iAF). The center frequency of the bandpass filter would be centered at 10 Hz, for example, by setting cutoff values offrequency to 9.5 Hz and 10.5 Hz. These cutoff values can help minimize phase shift introduced by the filter.

[0049] In some embodiments, the fundamental center frequency may be determined by selecting a value in the middle of a range of frequency values of interest, such as the midpoint of the bandpass filter used to filter the brain signal. Alternatively, the fundamental center frequency may be determined from a measurement of the brain signal itself. This can comprise bandpass filtering the brain signal to isolate a frequency range of interest, and then computing the FFT of a segment of the filtered brain signal. The dominant frequency within this frequency range of interest can be determined by finding the frequency of the FFT bin with the largest magnitude. As described herein, the frequency spectrum may need to first be de-trended to remove noise present in EEG signals. In some embodiments, the oscillation frequency may be determined using a bank of finite impulse response (FIR) notch filters that span the frequency range of interest. The dominant frequency of the signal within this range can then be estimated by determining a weighted sum of each filter frequency value according to the degree of attenuation induced on the brain signal by the filter at its notch frequency.

[0050] In some embodiments, the method 100 may comprise determining an evoked response delay 125. Determining the evoked response delay is described in greater detail with respect to FIG. 1C. Determining the evoked response delay can include delivering an initial stimulus 105 to the subject. The initial auditory stimulation may comprise a plurality of noise pulses. In some embodiments, the noise pulses may be duration pink noise pulses (e.g., at approximately 78-82 DB). The noise pulses may be fixed-duration noise pulses. The noise pulses may be delivered at random phases, or in accordance with a predetermined pulse train.

[0051] The system may receive brain wave data based on the initial stimulus. Based on the received brain wave data, an auditory evoked response potential can be determined. To do so, the brain wave data may be filtered, such as by using a bandpass filter 116 (as described in greater detail above with respect to determining the fundamental center frequency). An example bandpass filter for filtering the brain wave data may be between 2 and 30 Hz. An analysis window of the filtered data may be generated for segmenting the data into epochs. For example, an analysis window that begins prior to the time of delivery of the initial stimulus and extends through the stimulus can be generated. From this analysis window, the filtered data may be segmented into epochs. The epochs that comprise greater than a threshold number of artifacts and / or have a signal greater than a threshold voltage (e.g.,greater than ±100 |aV) may be discarded. The remaining epochs may be averaged to determine an evoked response potential (ERP) delay.

[0052] In some embodiments, the evoked response delay may be measured across multiple individuals, and the mean or median of these measurements may be interpreted as the evoked response delay. In some embodiments, the evoked response delay may be independently measured for each individual, for example, at a time prior to initiation of sleep. The evoked response delay may in some embodiments be determined by recording segments of EEG activity following each acoustic stimulation pulse.

[0053] In some embodiments, the method 100 may comprise determining a stimulation phase 130 for the subject based on the fundamental center frequency and the evoked response delay. The stimulation phase may be defined as the phase at which to trigger delivery of an acoustic stimulation pulse. As described herein, the stimulation phase may be chosen to be the phase at which triggering an acoustic stimulation pulse alters subsequent oscillation activity. The stimulation phase may be programmable and unique for each subject. As will be described in greater detail below, the stimulation phase may be selected to induce sleep, for example to treat sleep insomnia. In insomnia, EEG power in the alpha band is typically elevated. Neurostimulation at the determined stimulation phase may disrupt elevated alpha activity by altering subsequent alpha wave amplitude, thereby accelerating sleep initiation.

[0054] In some embodiments, determining the stimulation phase may comprise multiplying the evoked response delay (in seconds) by the fundamental center frequency (in Hz), and multiplying this product by 360 degrees. This intermediate value can be inputted to an algorithm configured to map it into a suitable angular phase range (e.g., [0°, 360°]).

[0055] In some embodiments, the stimulation phase may be chosen to align an evoked response to a sound pulse with a specific up-state or down-state of the oscillation of the brain waves. For example, the stimulation phase may align the evoked response potential (ERP) of the subject with a peak (i.e., optimal phase) and / or trough (i.e., pessimal phase) of the oscillation.

[0056] In some embodiments, determining the stimulation phase 130 may comprise determining an onset stimulation phase and an offset stimulation phase. The onset stimulation phase may be determined based on the above-described evoked response delay and the fundamental center frequency of the target oscillation. For example, the onset stimulation phase may be selected to align the peak of the evoked response with the peak of the upcoming oscillation.

[0057] The offset stimulation phase may be based on the onset stimulation phase, a pulse duration, the fundamental center frequency, and the evoked response delay. The pulse duration may be between about 20-500 ms, such as about 50 ms, 75 ms, 100 ms, 125 ms, 150 ms, 175 ms, or any value therebetween. In some embodiments, one or more of the fundamental center frequency and the pulse duration may be predetermined. Using these characteristics in combination with the aforementioned onset stimulation phase, the offset stimulation phase may be determined empirically such that the amplitude of the evoked response can be altered.

[0058] Optionally, the method may comprise receiving biomarker data from one or more sensors configured to detect biomarkers from the subject. Biomarker data may comprise at least one of a signal entropy and a variability in the fundamental center frequency. In some embodiments, the biomarker data may comprise a measurement of cardiac function, the cardiac function comprising at least one of heart rate, SpO2, and heart rate variability. In some embodiments, the biomarker data may comprise a measurement of a soluble factor, the soluble factor comprising at least one of insulin, glucose, and cortisol. The biomarker data may be used to iteratively update stimulation parameters, such as the simulation phase (e.g., onset and / or offset stimulation phase), pulse duration, and / or pulse pattern (i.e., pulse duty cycle). In a non-limiting example, the pulse duration may be varied according to a gradient descent (or other) algorithm to minimize variability in the fundamental center frequency, maximize blood oxygen saturation, etc. Example algorithms that may utilize biomarker data can include a parameter search and optimization algorithm, such as a simulated annealing algorithm, particle swarm algorithm, genetic algorithm, etc. In some embodiments, the biomarker data may be used to determine when to deliver a phase-locked stimulus, as described in greater detail below.

[0059] Optionally, the method 100 may comprise determining at least one of an amplitude (e.g., an instantaneous amplitude) and / or a spectral power of the oscillation data. In some embodiments, determining the amplitude and / or the spectral power of the brain signal may comprise using an endpoint-correct Hilbert transform (ecHT). As mentioned above, an ecHT algorithm (e.g., frequency domain ecHT, front-padded time domain ecHT, or end- padded time domain ecHT) can be used to determine an analytic signal based on a discrete signal created for the natural signal (e.g., the oscillation data). The instantaneous amplitude, similar to the instantaneous phase as described herein, can be determined at the end of the analytic signal. The endpoint of the analytic signal may correspond in time to the end of the original natural signal. In some embodiments, the amplitude of an oscillation can bedetermined by bandpass filtering the signal within a frequency range of interest (e.g., about 7.15-14 Hz, as described herein) and identifying the maximum value of the signal for each cycle.

[0060] The spectral power may be determined using a frequency domain representation (e.g., Fast Fourier Transform, FFT) of a portion of the oscillation data, as described herein. In some embodiments, the spectral power may be determined by computing a spectrogram (illustrated in graph 200 of FIG. 2). For example, a spectrogram can be computed using a multi-taper method and sliding window approach. As described in greater detail below, at least one of the spectral power and amplitude of the oscillation data may be used to deliver the phase-locked stimulus.

[0061] Method 100 may comprise, at block 140, delivering a phase-locked stimulus based on the stimulation phase to induce sleep. In some embodiments, the acoustic stimulus may be delivered in accordance with a determined onset stimulation phase and a determined offset stimulation phase. As described in greater detail below, in some embodiments, the acoustic stimulus may be delivered only if the spectral power and / or the amplitude of the oscillation data meets or exceeds a predetermined threshold value.

[0062] In some embodiments, delivering the phase-locked acoustic stimulus may comprise delivering the stimulus beginning at the determined stimulation phase and sustaining the stimulus for a fixed duration. The fixed duration can be a fixed time duration and / or a fixed phase duration. The phase-locked stimulus may comprise a stimulation pulse. For example, a stimulation pulse may be delivered for about 50 ms. In some embodiments, the duration of the pulse may be greater than or equal to about 20 ms, 30 ms, 40 ms, 50 ms, 60 ms, 70 ms, 80 ms, 90 ms, 100 ms, 125 ms, 150 ms, 175 ms, 200 ms, 250 ms, 300 ms, 350 ms, 400 ms, or 450 ms, . In some embodiments, the duration of the pulse may be less than or equal to about 30 ms, 40 ms, 50 ms, 60 ms, 70 ms, 80 ms, 90 ms, 100 ms, 125 ms, 150 ms, 175 ms, 200 ms, 250 ms, 300 ms, 350 ms, 400 ms, 450 ms, or 500 ms.

[0063] In some embodiments, the stimulus may comprise a continuous auditory stimulation. The continuous auditory stimulation may be in addition to or instead of the above-described stimulation pulses. The continuous auditory stimulation may comprise at least one pitch. For example, the stimulation may be a single pitch, a pattern of different pitches, a song or tune, or other background noise. At least one of the volume and rate of the continuous auditory stimulation may be adjustable. For example, the volume and / or rate of the stimulation may be adjusted based on the received oscillation data such that the stimulation is at a rate synchronous with the brain waves.

[0064] In some embodiments, the acoustic stimulus may begin when a particular onset stimulation phase is detected, and end when an offset stimulation phase is detected. In some embodiments, a first stimulation pulse may have a first duration, and a second stimulation pulse may have a second duration. The duration of the first stimulation pulse may be greater than, less than, or about equal to the duration of the second stimulation pulse.

[0065] In some embodiments, phase-locked acoustic stimulation pulses may be delivered in accordance with a pulse train pattern. The pulse train pattern may be predetermined. In some embodiments, the acoustic pulse may be delivered intermittently relative to the oscillation cycle of the stage of sleep. For example, the acoustic stimulation pulse may be delivered every 2, 3, 4, 5, 6, 7, or more oscillation cycles of the stage of sleep. In some embodiments, the acoustic stimulation pulse may be delivered for 1, 2, 3, or more (consecutive) oscillation cycles, followed by 1, 2, 3, or more (consecutive) oscillation cycles with no stimulation pulse. This pulse train pattern may repeat.

[0066] In some embodiments, the phase-locked acoustic stimulation pulse may be delivered in each oscillation cycle of the stage of sleep. In some embodiments, the pulse train pattern may comprise delivering the stimulation pulse in each oscillation cycle for a plurality of oscillation cycles (e.g., 2, 3, 4, 5, or more consecutive oscillation cycles), and after a given number of oscillation cycles, the stimulation pulse may be delivered intermittently (e.g., at every other oscillation cycle, at every 2, 3, 4 or more oscillation cycles, etc.), or vice versa. The pulse duration of the acoustic stimulation pulse may be fixed or may be variable (as described above).

[0067] The phase-locked auditory stimulus may be delivered for substantially all of the duration of a stage of sleep, including prior to sleep when the subject is awake. In some embodiments, auditory stimulus may only be delivered for a portion of a given stage of sleep (e.g., about 25%, about 50%, about 75%, about 90%, etc.). In some embodiments, as described herein, the stimulus may only be delivered when the subject is awake and the amplitude and / or spectral power of the oscillation data is above a predetermined threshold value. The system may be configured in a closed-loop manner in that the oscillation data may be continuously or iteratively monitored to selectively deliver acoustic stimulation based on, for example, the stage of sleep (e.g., including whether the subject is awake), one or more characteristics of the stage of sleep (e.g., spectral power and / or amplitude), etc. As used herein, closed-loop may refer to the use or implementation of a closed-loop control system with one or more feedback loops to modulate system outputs and automatically regulateprocess variables to a desired state or set point. In some cases, the closed-loop control system can comprise a proportional-integral-derivative controller (PID controller).

[0068] In some embodiments, the phase-locked acoustic stimulation pulse may be delivered only if the subject is awake. In some embodiments, a first type of acoustic stimulation may be delivered if the subject is awake, and a second type of acoustic stimulation may be delivered if the subject is determined to be in a different stage of sleep (e.g., N1 sleep). The first and second types of acoustic stimulation may differ in, for example, stimulation phase (e.g., onset stimulation phase and / or offset stimulation phase), pulse duration, pulse train pattern, etc. As described herein, the system executing method 100 may be configured in a closed-loop manner such that the progression between the stages of sleep can be monitored and the acoustic stimulation can be updated in accordance with the detected stage of sleep.

[0069] In some embodiments, auditory stimulation may comprise broad- spectrum sound pulses. For example, the acoustic stimulation pulses may comprise pink and / or white noise pulses. The acoustic stimulation pulses may be in the range of, for example, about 20 decibels (DB) to about 60 DB, such as 20 DB, 25 DB, 30 DB, 35 DB, 40 DB, 45 DB, 50 DB, 55 DB, or 60 DB.

[0070] In some embodiments, delivering a phase-locked stimulus may be based on biomarker data. For example, delivering the phase-locked stimulus may be in accordance with at least one biomarker reaching a pre-determined threshold.

[0071] As described herein, delivering a phase-locked stimulus may be based on an amplitude and / or spectral power of the oscillation data. In some embodiments, the determined amplitude and / or spectral power of the oscillation data may be compared to a threshold (e.g., an amplitude threshold and a spectral power threshold, respectively). The threshold may be indicative of the sleep stage of the subject. These thresholds can vary, for example, based on electrode position, contact, etc. For example, the threshold may be represented by an increase in amplitude and / or spectral power of the brain signals within about a 0.1 Hz to 4 Hz range, as compared to a previously measured baseline amplitude or spectral power for the individual. The baseline may be determined during a resting stage of the subject. In the instance the determined amplitude and / or spectral power of the oscillation data meets or exceeds the predetermined threshold value(s), a phase-locked acoustic stimulation pulse may be delivered. In the instance the determined amplitude and / or spectral power of the brain signal does not meet or exceed the predetermined threshold value(s), the phase-lockedstimulation pulse may not be delivered, and the system may be configured to continue monitoring the amplitude and / or spectral power of the brain signal.

[0072] The stimulation phase may be selected to accelerate sleep initiation for subjects with insomnia. To do so, elevated alpha activity (e.g., the amplitude of the alpha oscillation) may be modulated in comparison to a baseline alpha amplitude determined prior to the start of stimulation. By altering EEG power in the alpha band, sleep initiation can be accelerated, thereby reducing insomnia in subjects. For example, phase-locked acoustic pulses can be delivered at a stimulation phase corresponding to the subject being awake. Once it is determined that the subject is in a sleep stage (e.g., REM, Nl, N2, etc.), stimulation can be terminated. In some embodiments, once it is determined that the subject is in a sleep stage, phase-locked acoustic stimulation (e.g., of a different stimulation phase, for example) can be delivered to maintain the sleep. As described herein, in some embodiments the acoustic stimuli may be delivered for the duration of a given sleep stage to maintain the sleep. Alternatively, in some embodiments, the acoustic stimulus may be delivered as needed, that is, upon a threshold (e.g., a brain signal amplitude threshold and / or spectral power threshold) being met or exceeded. In some embodiments, acoustic stimulation may be delivered for a predetermined amount of time upon the subject entering a given sleep stage.Neuromodulation Systems and Devices

[0073] FIG. 3 illustrates a schematic of an exemplary acoustic neuromodulation system 300 that can be configured to execute method 100 illustrated across FIGS. 1A-1C. The system 300 may be embodied in a wearable medical device and / or a tabletop device, described in greater detail below. The wearable medical device may be an electroencephalogram (EEG)-based device. In some embodiments, the wearable medical device may be a headband and / or a CPAP mask. For example, a subject may wear the headband device configured to measure brain (EEG) signals from one or more positions on the scalp.

[0074] The acoustic neuromodulation system 300 may comprise one or more hardware components known to one of ordinary skill in the art for signal processing, including but not limited to one or more processors (e.g., microcontrollers), an analog-to-digital converter (ADC), a digital signal processor (DSP), a digital-analog-converter (DAC), and a transducer configured to provide a stimulation signal to an output device.

[0075] The system 300 may comprise one or more processors 305 embodied within the tabletop device and / or wearable medical device. The one or more processors 305 may beconfigured to receive inputs such as oscillation data (e.g., electrophysiological signals, such as EEG signals) from one or more sensors 310 (e.g., electrodes). For example, the one or more sensors 310 may be sensors of the wearable medical device (e.g., a headband, a CPAP mask, etc.), an electroencephalography (EEG) device, and / or the external tabletop device. The one or more sensors may be configured to detect neural activity, such as neural activity associated with sleep. Based on the detected neural activity, the neuromodulation device may be configured to determine the fundamental center frequency, the evoked response delay, and in turn a stimulation phase for delivering a phase-locked stimulus to induce sleep in the subject.

[0076] In some embodiments, the processor(s) 305 may be configured to receive biomarker data from one or more biomarker sensors 315. The biomarker sensors may be sensors of an EEG machine, a heart rate monitor, a pulse oximeter, an audio recording device, etc. The one or more sensors may be configured to detect and record physiological data, such as EEG signals, heart rate, oxygen saturation, respiratory rate and effort, motion, sound (e.g., snoring sound), echocardiogram (ECG), and / or electromyogram (EMG). For example, one or more of the aforementioned signals may be used to monitor symptoms of insomnia experienced by the subject. In some embodiments, the biomarker sensors may be embodied in the aforementioned headband device and / or CPAP mask.

[0077] The system 300 may be configured to receive one or more audio files 320 containing sounds to be delivered as acoustic stimulation. The system 300 may be configured to receive configuration scripts 325. The configurations scripts 325 may be usable by system 300 to define initial input and output parameters, such as initial trigger phase, audio volume, pulse duration, etc. The configuration scripts 325 and audio files 320 may be stored, for example in a memory 330 of system 300. The memory 330 may be communicatively coupled (e.g., via a wired or wireless connection, such as a USB connection, a Bluetooth connection, etc.) to the one or more processors 305 to allow data to be passed between the memory 330 and the processors 305. Memory 330 of system 300 may be of the aforementioned external tabletop device and / or of the wearable device.

[0078] The one or more processors 305 may be configured to receive configuration scripts stored in the memory to configure the acoustic neuromodulation system. The configuration scripts usable by the one or more processors 305 may in some embodiments initially be set based on a user input indicating a desired output type. For example, the user may request that the system 300 be configured to deliver acoustic neurostimulation for inducing sleep and treating sleep insomnia. The active configuration scripts may be used toconfigure the system for determining the parameters (e.g., fundamental center frequency, evoked response delay) for inducing sleep, thereby treating sleep insomnia. In some embodiments, the active configuration scripts may be based on population normative brain signals and corresponding biomarkers. For example, the configuration scripts may be set initially based on parameters that may be, on average, optimal for a population of users. As described herein, the system 300 may be configured to iteratively refine the configuration scripts, for example, based on received oscillation data and / or biomarker data. For example, the received oscillation data may be processed using one or more causal filters, an endpoint- corrected Hilbert transform algorithm (ecHT), etc. as described in greater detail above. Based on this processed signal, parameters of the oscillation data, such the fundamental center frequency, evoked response delay, etc. may be identified and used to determine a stimulation phase for acoustic neurostimulation. The neuromodulating audio signal may be phase-locked with the endogenous signals (e.g., brain wave signals) of the subject. In some embodiments, the configuration scripts can be refined over the course of a single neuromodulation session, or over the course of a plurality of sessions.

[0079] In some implementations, the one or more processors 305 may be configured to generate the acoustic signal and drive a closed-loop playback system. The generated acoustic signal may be transmitted as instructions to an output device 335 (e.g., an auditory stimulator, such as a speaker). The auditory stimulator may be embodied in a wearable device (e.g., a headband, CPAP mask, etc.) described herein and / or provided as an external tabletop device. In some embodiments, the acoustic stimulation may be provided via one or more boneconduction drivers (e.g., embodied in a wearable headband). As described herein, the delivered acoustic stimulation may be phase-locked to phases of oscillatory activity.

[0080] In some embodiments, the one or more processors may be configured to record and transmit sensor data (e.g., brain signal sensors, biomarker sensors, environmental sensors, etc.) to the memory 330 and / or an external storage or device. For example, sensor data (e.g., biomarker data, oscillation data, etc.) may be transmitted to a smartphone, tablet, desktop, or other web application for presentation to the subject, caregiver, and / or a clinician.Channel Switching on Wearable Devices

[0081] The disclosure provided herein also pertains to wearable devices (e.g.,“wearables”, such as watches, rings, headbands, etc.) having multiple sets of electrodes in different locations.

[0082] The wearables described herein (e.g., wearable device 500 illustrated in FIG. 5 and described in greater detail below) can have multiple sets of electrodes in different locations and also can have real-time channel switching functionality to switch between electrode channels to get the best signal from the wearer of the wearable. For example, a wearer of an EEG headband with electrodes on the inside of the headband can sleep on one side, enabling one set of electrodes to be used (e.g., Fpl / F7, a Ref, and GND G of wearable device 500 in FIGS. 5A-5B). The wearer may rouse a little and turn to the other side. Suddenly, the electrode channel that was being used now has worsened data (caused, e.g., by an intermittent physical connection due to less pressure being applied by the wearer’s head). The headband device described herein can automatically detect this and switch to a different electrode channel (e.g., Fp2 / F8, the Ref on the other side, and GND G of wearable device 500 in FIGS. 5A-5B) that has better signal quality. In addition, the wearable devices described herein may use this measurement to change the functional identity of electrodes. For example, if neither Ref has acceptable impedance, another electrode may be chosen as Ref instead. The wearable devices provided herein may use impedance measurements over time to deliver information about the lifetime of the electrodes to the user, and alert of a need for replacement.

[0083] As illustrated in FIGS. 5A-5B, an exemplary EEG headband 500 can have multiple sets of electrodes 506, 508, 510 and a bone transducer / driver 512 touching the forehead of the wearer (e.g., located on the forehead between the wearer’s pupils) to apply audio signals directly to the wearer’s forehead. The bone transducer 512 is shown in FIGS. 5A-5B a little to the right of center. In some embodiments, placement of electrodes across the forehead allows for very similar brain waves to be measured by each electrode, but the physical separation of electrodes allows for redundancy in case one electrode loses contact with the head.

[0084] In an exemplary embodiment, a controller or processor (e.g., processor 502) can algorithmically combine more than one (e.g., three) channels of EEG measurements into one channel, effectively choosing the “best” channel, also called channel switching. The best channel can mean one or any two or more of the following: the channel with the lowest impedance, the channel with smallest deviation from the expected EEG amplitude, the channel that is within min-max amplitude bounds (where the min-max bounds can be bounds can be determined from the user’s historical data from the current sleep session or previous sleep sessions), the channel with the highest RMS value for the EEG signal, and the channel with least motion artifacts from the hypnogram scores for the preceding epoch(s). Regardingthe channel with the highest RMS value for the EEG signal, the RMS values for the EEG signal (e.g., 0.2-40 Hz) can be computed for the current channel for a window of time (e.g., 5 seconds). If the RMS for the current channel drops below a fixed threshold, the RMS values for the remaining 2 channels can be calculated and compared, and the channel with the highest RMS value can be selected to be active.

[0085] In some embodiments, electrode impedance can be measured for each electrode when the device is initially put in place on a wearer and prepared for use by the wearer. In some embodiments, electrode impedance can additionally or alternatively be measured when in use by a wearer, for example to dynamically choose active recording electrodes to optimize the quality of a recorded signal. Additionally or alternatively, impedance measurements may be used to alert a user about specific adjustments that should be made, for example to improve contact of a particular electrode.

[0086] In some embodiments, the wearable devices described herein may have a processor-controlled current source (AC or DC) to measure the impedance of each electrode against the skin. For example, a voltage for each electrode induced by the current source may be measured (e.g., via an ADC in the processor or external to the processor) and the voltages may be compared by the processor to determine the lowest-impedance electrode. The channel(s) with the lowest impedance can be used. In the event that the impedance test introduces noise in the recording, the processor may automatically switch away from an electrode that is actively running an impedance test to the next best electrode without an impedance test.

[0087] In exemplary embodiments, the channels can be switched in the processor by selecting a different data stream to process. It is envisaged that physical switching, logical switching, or some other switching can also be used. During the channel switching operation, artifacts can be introduced due to mismatches in the amplitude or phase of oscillation between the signals being switched. In some embodiments, the processor can mute the stimulation output (the audio signals applied by the device through the bone conductor) during the switch for a period of time until the artifact from the switch becomes insignificant.Computing Device

[0088] FIG. 14 depicts a computing device 1400, according to one or more examples of the disclosure. In one or more examples, device 1400 may be understood to encompass one or more components of the neuromodulation system 300 illustrated in FIG. 3. In one or moreexamples, computing device 1400 may be configured to implement a method for inducing sleep using neuromodulation, such as the method 100 of FIGS. 1A-1C.

[0089] Device 1400 can be a host computer connected to a network. Device 1400 can be a client computer or a server. As shown in FIG. 14, device 1400 can be any suitable type of microprocessor-based device, such as a personal computer, workstation, server, or handheld computing device (portable electronic device) such as a phone or tablet. The device can include, for example, one or more of processors 1402, input device 1406, output device 1408, storage 1410, and communication device 1404. Input device 1406 and output device 1408 can generally correspond to those described above and can either be connectable or integrated with the computer. For example, the one or more processors 1402 may comprise processors 305. The input device 1406 may comprise the one or more sensors (e.g., brain signal sensors 310 and / or biomarker sensors 315 described herein with respect at least to system 300 and FIG. 3). The output device 1408 may comprise the output device 335. The storage 1410 may comprise at least the memory 330 described herein with respect to system 300 illustrated in FIG. 3.

[0090] Input device 1406 can be any suitable device that provides an input, such as a touch screen, keyboard or keypad, mouse, or voice-recognition device. Output device 1408 can be any suitable device that provides output, such as a touch screen, haptics device, or speaker.

[0091] Storage 1410 can be any suitable device that provides storage, such as an electrical, magnetic, or optical memory, including a RAM, cache, hard drive, or removable storage disk. Communication device 1404 can include any suitable device capable of transmitting and receiving signals over a network, such as a network interface chip or device. The components of the computer can be connected in any suitable manner, such as via a physical bus or wirelessly.

[0092] Software 1412, which can be stored in storage 1410 and executed by processor 1402, can include, for example, the programming that embodies the functionality of the present disclosure (e.g., as embodied in the devices as described above).

[0093] Software 1412 can also be stored and / or transported within any non-transitory computer-readable storage medium for use by or in connection with an instruction execution system, apparatus, or device, such as those described above, that can fetch instructions associated with the software from the instruction execution system, apparatus, or device and execute the instructions. In the context of this disclosure, a computer-readable storagemedium can be any medium, such as storage 1410, that can contain or store programming for use by or in connection with an instruction execution system, apparatus, or device.

[0094] Software 1412 can also be propagated within any transport medium for use by or in connection with an instruction execution system, apparatus, or device, such as those described above, that can fetch instructions associated with the software from the instruction execution system, apparatus, or device and execute the instructions. In the context of this disclosure, a transport medium can be any medium that can communicate, propagate, or transport programming for use by or in connection with an instruction execution system, apparatus, or device. The transport readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, or infrared wired or wireless propagation medium.

[0095] Device 1400 may be connected to a network, which can be any suitable type of interconnected communication system. The network can implement any suitable communications protocol and can be secured by any suitable security protocol. The network can comprise network links of any suitable arrangement that can implement the transmission and reception of network signals, such as wireless network connections, T1 or T3 lines, cable networks, DSL, or telephone lines.

[0096] Device 1400 can implement any operating system suitable for operating on the network. Software 1412 can be written in any suitable programming language, such as C, C++, Java, or Python. In various embodiments, application software embodying the functionality of the present disclosure can be deployed in different configurations, such as in a client / server arrangement or through a Web browser as a Web-based application or Web service, for example.Definitions & Conclusion

[0097] As used herein, the singular forms “a”, “an”, and “the” include the plural reference unless the context clearly dictates otherwise.

[0098] Reference to “about” a value or parameter herein includes (and describes) variations that are directed to that value or parameter per se. For example, description referring to “about X” includes description of “X”.

[0099] It is understood that aspects and variations of the invention described herein include “consisting” and / or “consisting essentially of’ aspects and variations.

[0100] The term “real time” or “real-time,” as used interchangeably herein, generally refers to an event (e.g., an operation, a process, a method, a technique, a computation, acalculation, an analysis, a visualization, an optimization, etc.) that is performed using recently obtained (e.g., collected or received) data. In some cases, a real time event may be performed almost immediately or within a short enough time span, such as within at least 1 millisecond (ms), 5 ms, 0.01 seconds, 0.05 seconds, 0.1 seconds, 0.5 seconds, 1 second, 0.1 minute, 0.5 minutes, 1 minute, or more. In some cases, a real time event may be performed almost immediately or within a short enough time span, such as within at most 1 second, 0.5 seconds, 0.1 seconds, 0.05 seconds, 0.01 seconds, 5 ms, 1 ms, or less.

[0101] The term “treat,” “treatment,” and / or “treating” is used herein generally refers to reducing the severity of sleep apnea. Reducing the severity of sleep apnea may include reducing the frequency or duration of apnea or hypopnea events and / or reducing the severity of symptoms relating to or resulting from sleep apnea. Reducing the severity of sleep apnea may in some instances involve measurable changes in diagnostic markers used to detect the presence of sleep apnea.

[0102] When a range of values or values is provided, it is to be understood that each intervening value between the upper and lower limit of that range, and any other stated or intervening value in that stated range, is encompassed within the scope of the present disclosure. Where the stated range includes upper or lower limits, ranges excluding either of those included limits are also included in the present disclosure.

[0103] The entire disclosure of the patents and publications referred in this application are hereby incorporated herein by reference for all purposes. To the extent that any reference incorporated by reference conflicts with the instant disclosure, the instant disclosure shall control.

[0104] The section headings used herein are for organization purposes only and are not to be construed as limiting the subject matter described. The description is presented to enable one of ordinary skill in the art to make and use the invention and is provided in the context of a patent application and its requirements. Various modifications to the described embodiments will be readily apparent to those persons skilled in the art and the generic principles herein may be applied to other embodiments. Thus, the present invention is not intended to be limited to the embodiment shown but is to be accorded the widest scope consistent with the principles and features described herein.

[0105] The figures illustrate processes according to various embodiments. In the exemplary processes, some blocks are, optionally, combined, the order of some blocks is, optionally, changed, and some blocks are, optionally, omitted. In some examples, additional steps may be performed in combination with the exemplary processes. Accordingly, theoperations as illustrated (and described in greater detail below) are exemplary by nature and, as such, should not be viewed as limiting.

[0106] Although the disclosure and examples have been fully described with reference to the accompanying figures, it is to be noted that various changes and modifications will become apparent to those skilled in the art. Such changes and modifications are to be understood as being included within the scope of the disclosure and examples as defined by the claims.EXEMPLARY EMBODIMENTS

[0107] The following embodiments are exemplary and are not intended to limit the scope of any invention described herein.

[0108] Embodiment 1. A method for inducing sleep in a subject using neuromodulation, comprising: receiving oscillation data from one or more sensors configured to detect brain waves from the subject; determining a fundamental center frequency of the oscillation data; determining an evoked response delay of the oscillation data; based on the determined fundamental center frequency and the determined evoked response delay, determining a stimulation phase for the subject; and delivering a phase-locked stimulus in accordance with the stimulation phase to induce sleep in the subject.

[0109] Embodiment 2. The method of embodiment 1, wherein the fundamental center frequency is an intrinsic center frequency (iCF) of the subject, and the oscillation data comprises at least one of alpha, theta, delta, beta, and gamma oscillation data.

[0110] Embodiment 3. The method of embodiment 1 or 2, wherein the oscillation data occupies a frequency band between about 0.1-150 Hz.

[0111] Embodiment 4. The method of any one of embodiments 1-3, wherein determining the fundamental center frequency of the oscillation data comprises: receiving oscillation data from the one or more sensors while the subject is awake and with closed eyes for a duration of time; determining spectral data of the received data, wherein the spectral data comprises a noise floor; generating de-noised spectral data by subtracting the noise floor from the spectral data; andidentifying a peak within a portion of the de-noised spectral data, the peak corresponding to the fundamental center frequency of the subject.

[0112] Embodiment 5. The method of any one of embodiments 1-4, wherein determining the evoked response delay of the oscillation data comprises: delivering an initial auditory stimulation to the subject, the initial auditory stimulation comprising a plurality of noise pulses delivered at random phases; receiving brain wave data based on the initial auditory stimulation; determining an auditory evoked response potential, comprising: filtering the received brain wave data using a bandpass filter; segmenting the filtered data into a plurality of epochs; and averaging at least a portion of the epochs that do not comprise artifacts or poor signals to determine the auditory evoked delay.

[0113] Embodiment 6. The method of any one of embodiments 1-5, wherein determining the stimulation phase comprises: multiplying the evoked response delay by the fundamental center frequency; multiplying a product of the evoked response delay and the fundamental center frequency by 360 degrees to obtain an intermediate value; and inputting the intermediate value to an algorithm that maps into an angular range of [0°, 360°].

[0114] Embodiment 7. The method of any one of embodiments 1-6, wherein the determined stimulation phase aligns an evoked response with an up-state of the oscillation of the brain waves or a down-state of the oscillation of the brain waves.

[0115] Embodiment 8. The method of any one of embodiments 1-7, wherein determining the stimulation phase comprises determining an onset stimulation phase and an offset stimulation phase, the offset stimulation phase based on a pulse duration, the fundamental center frequency, and the evoked response delay.

[0116] Embodiment 9. The method of any one of embodiments 1-8, wherein delivering the phase-locked stimulus comprises delivering the stimulus beginning at the determined stimulation phase and sustaining the stimulus for a fixed duration, and wherein the fixed duration comprises at least one of a fixed time duration and a fixed phase duration.

[0117] Embodiment 10. The method of any one of embodiments 1-9, wherein delivering the phase-locked stimulus comprises delivering a phase-locked stimulation pulse intermittently in accordance with a predetermined pulse train pattern defined relative to an oscillation cycle of the received oscillation data.

[0118] Embodiment 11. The method of any one of embodiments 1-10, wherein delivering the phase-locked stimulus comprises delivering a phase-locked stimulation pulse in each oscillation cycle of the received oscillation data.

[0119] Embodiment 12. The method of any one of embodiments 1-11, comprising determining at least one of an amplitude and a spectral power of the oscillation data.

[0120] Embodiment 13. The method of embodiment 12, wherein delivering the phase- locked stimulus comprises delivering the phase-locked stimulus in accordance with at least one of the amplitude and the spectral power reaching a predetermined threshold value.

[0121] Embodiment 14. The method of any one of embodiments 1-13, wherein the stimulus comprises a broad- spectrum sound pulse.

[0122] Embodiment 15. The method of any one of embodiment 1-14, wherein the stimulus comprises a continuous auditory stimulation, the continuous auditory stimulation including at least one pitch, and the continuous auditory stimulation deliverable in at least one volume.

[0123] Embodiment 16. The method of any one of embodiments 1-15, comprising receiving biomarker data from one or more sensors configured to detect biomarkers from the subject.

[0124] Embodiment 17. The method of embodiment 16, wherein the received biomarker data comprises a measurement based on an electroencephalogram (EEG) signal, the measurement including at least one of an entropy and a variability in the fundamental center frequency.

[0125] Embodiment 18. The method of embodiment 16 or 17, wherein the biomarker data comprises a measurement of cardiac function, the cardiac function comprising at least one of heart rate, SpO2, and heart rate variability.

[0126] Embodiment 19. The method of any one of embodiments 16-18, wherein the biomarker data comprises a measurement of a soluble factor, the soluble factor comprising at least one of insulin, glucose, and cortisol.

[0127] Embodiment 20. The method of any one of embodiments 16-19, wherein delivering the phase-locked stimulus comprises delivering the phase-locked stimulus in accordance with at least one biomarker reaching a pre-determined threshold.

[0128] Embodiment 21. The method of any one of embodiments 16-20, comprising updating the stimulation phase based on the received biomarker data to maximize or minimize at least one biomarker.

[0129] Embodiment 22. The method of any one of embodiments 1-21, comprising using a wearable device on the subject that comprises the one or more sensors and one or more stimulators to deliver the phase-locked stimulus.

[0130] Embodiment 23. The method of any one of embodiments 1-22, wherein inducing sleep in the subject reduces sleep onset latency and / or treats sleep insomnia.

[0131] Embodiment 24. A method for determining an intrinsic center frequency (iCF) of a subject, comprising: receiving oscillation data from one or more sensors configured to detect brain waves; determining spectral data of the received data, wherein the spectral data comprises a noise floor; generating de-noised spectral data by subtracting the noise floor from the spectral data; and identifying a peak within a portion of the de-noised spectral data, the peak corresponding to the intrinsic center frequency (iCF) of the subject.

[0132] Embodiment 25. The method of embodiment 24, comprising delivering a stimulus based on the determined iCF of the subject.

[0133] Embodiment 26. The method of embodiment 24 or 25, wherein the received oscillation data comprises brain wave data detected while the subject is awake and with closed eyes for a duration of time.

[0134] Embodiment 27. The method of any one of embodiments 24-26, wherein determining the spectral data comprises using a Fourier transform.

[0135] Embodiment 28. The method of any one of embodiments 24-27, wherein determining the spectral data comprises modeling the noise floor of the spectral data with a 3rd-order polynomial fit.

[0136] Embodiment 29. The method of any one of claims 24-28, wherein identifying the peak within the portion of the de-noised spectral data comprises identifying a peak within a frequency range of about 0.1 to 150 Hz.

[0137] Embodiment 30. A system for inducing sleep in a subject using neuromodulation, comprising: one or more sensors configured to detect brain waves from the subject; a processing device configured to: receive oscillation data from the one or more sensors; determine a fundamental center frequency of the oscillation data; determine an evoked response delay of the oscillation data; andbased on the determined fundamental center frequency and evoked response delay, determine a stimulation phase for the subject; and one or more stimulators configured to deliver a phase-locked stimulus in accordance with the stimulation phase to induce sleep in the subject.

[0138] Embodiment 31. The system of embodiment 30, wherein at least the one or more sensors and the stimulators are provided in a wearable device wearable by the subject.

[0139] Embodiment 32. The system of embodiment 30, wherein the fundamental center frequency is an intrinsic center frequency (iCF) of the subject, and the oscillation data comprises at least one of alpha, theta, delta, beta, and gamma oscillation data.

[0140] Embodiment 33. The system of embodiment 31 or 32, wherein the oscillation data occupies a frequency band between about 0.1-150 Hz.

[0141] Embodiment 34. The system of any one of embodiments 30-33, wherein determining the fundamental center frequency of the oscillation data comprises: receiving oscillation data from the one or more sensors while the subject is awake and with closed eyes for a duration of time; determining spectral data of the received data, wherein the spectral data comprises a noise floor; generating de-noised spectral data by subtracting the noise floor from the spectral data; and identifying a peak within a portion of the de-noised spectral data, the peak corresponding to the fundamental center frequency of the subject.

[0142] Embodiment 35. The system of any one of embodiments 30-34, wherein determining the evoked response delay of the oscillation data comprises: delivering an initial auditory stimulation to the subject, the initial auditory stimulation comprising a plurality of noise pulses delivered at random phases; receiving brain wave data based on the initial auditory stimulation; determining an auditory evoked response potential, comprising: filtering the received brain wave data using a bandpass filter; segmenting the filtered data into a plurality of epochs; and averaging at least a portion of the epochs that do not comprise artifacts or poor signals to determine the auditory evoked delay.

[0143] Embodiment 36. The system of any one of embodiments 30-35, wherein determining the stimulation phase comprises: multiplying the evoked response delay by the fundamental center frequency; 1multiplying a product of the evoked response delay and the fundamental center frequency by 360 degrees to obtain an intermediate value; and inputting the intermediate value to an algorithm that maps into an angular range of [0°, 360°].

[0144] Embodiment 37. The system of any one of embodiments 30-36, wherein the determined stimulation phase aligns an evoked response with an up-state of the oscillation of the brain waves or a down-state of the oscillation of the brain waves.

[0145] Embodiment 38. The system of any one of embodiments 30-37, wherein determining the stimulation phase comprises determining an onset stimulation phase and an offset stimulation phase, the offset stimulation phase based on a pulse duration, the fundamental center frequency, and the evoked response delay.

[0146] Embodiment 39. The system of any one of embodiments 30-38, wherein delivering the phase-locked stimulus comprises delivering the stimulus beginning at the determined stimulation phase and sustaining the stimulus for a fixed duration, and wherein the fixed duration comprises at least one of a fixed time duration and a fixed phase duration.

[0147] Embodiment 40. The system of any one of embodiments 30-39, wherein delivering the phase-locked stimulus comprises delivering a phase-locked stimulation pulse intermittently in accordance with a predetermined pulse train pattern defined relative to an oscillation cycle of the received oscillation data.

[0148] Embodiment 41. The system of any one of embodiments 30-40, wherein delivering the phase-locked stimulus comprises delivering a phase-locked stimulation pulse in each oscillation cycle of the received oscillation data.

[0149] Embodiment 42. The system of any one of embodiments 30-41, comprising determining at least one of an amplitude and a spectral power of the oscillation data.

[0150] Embodiment 43. The system of embodiment 42, wherein delivering the phase- locked stimulus comprises delivering the phase-locked stimulus in accordance with at least one of the amplitude and the spectral power reaching a predetermined threshold value.

[0151] Embodiment 44. The system of any one of embodiments 30-43, wherein the stimulus comprises a broad- spectrum sound pulse.

[0152] Embodiment 45. The system of any one of embodiment 30-44, wherein the stimulus comprises a continuous auditory stimulation, the continuous auditory stimulation including at least one pitch, and the continuous auditory stimulation deliverable in at least one volume.

[0153] Embodiment 46. The system of any one of embodiments 30-45, comprising receiving biomarker data from one or more sensors configured to detect biomarkers from the subject.

[0154] Embodiment 47. The system of embodiment 46, wherein the received biomarker data comprises a measurement based on an electroencephalogram (EEG) signal, the measurement including at least one of an entropy and a variability in the fundamental center frequency.

[0155] Embodiment 48. The system of embodiment 46 or 47, wherein the biomarker data comprises a measurement of cardiac function, the cardiac function comprising at least one of heart rate, SpO2, and heart rate variability.

[0156] Embodiment 49. The system of any one of embodiments 46-48, wherein the biomarker data comprises a measurement of a soluble factor, the soluble factor comprising at least one of insulin, glucose, and cortisol.

[0157] Embodiment 50. The system of any one of embodiments 46-49, wherein delivering the phase-locked stimulus comprises delivering the phase-locked stimulus in accordance with at least one biomarker reaching a pre-determined threshold.

[0158] Embodiment 51. The system of any one of embodiments 46-50, comprising updating the stimulation phase based on the received biomarker data to maximize or minimize at least one biomarker.

[0159] Embodiment 52. The system of any one of embodiments 30-51, comprising using a wearable device on the subject that comprises the one or more sensors and one or more stimulators to deliver the phase-locked stimulus.

[0160] Embodiment 53. The system of any one of embodiments 30-52, wherein inducing sleep in the subject reduces sleep onset latency and / or treats sleep insomnia.

[0161] Embodiment 54. A non-transitory computer-readable storage medium storing instructions that, when executed by a processing device operatively coupled to one or more sensors configured to detect brain waves of a subject, cause the device to: receive oscillation data from the one or more sensors; determine a fundamental center frequency of the oscillation data; determine an evoked response delay of the oscillation data; based on the determined fundamental center frequency and evoked response delay, determine a stimulation phase for the subject; and deliver a phase-locked stimulus in accordance with the stimulation phase to induce sleep in the subject.EXAMPLES

[0162] The feasibility of using an endpoint-corrected version of the Hilbert Transform (ecHT) algorithm implemented efficiently on-device to measure alpha phase and deliver phase-locked stimulation in the form of pink noise sound bursts to modulate ongoing alpha oscillations and promote healthy sleep initiation is examined herein. First, the ecHT algorithm is implemented on a tabletop electroencephalogram (EEG) device and used to measure the timing of the auditory evoked response and its delivery at precise phases of the alpha oscillation. Secondly, a pilot at-home clinical trial tests feasibility to use a headband wearable version of the neuromodulation device for real-time phase-locked stimulation in the alpha (8-12 Hz) frequency range.

[0163] Auditory stimulation was delivered at the intended phases of the alpha oscillation with high precision, and alpha oscillations were affected differently by stimuli delivered at opposing phases. The wearable system described herein was capable of measuring sleep micro- and macro-events present in the EEG that were appropriate for clinical sleep scoring during the at-home study. Moreover, sleep onset latencies were reduced for a subset of subjects displaying sleep onset insomnia symptoms in the stimulation condition. This study demonstrates the feasibility of closed-loop real-time tracking and neuromodulation of alpha oscillations using a wearable EEG device. Preliminary results suggest that this approach could be used to accelerate sleep initiation in individuals with objective insomnia symptoms.

[0164] The current study explores the feasibility to physiologically track and modulate brain oscillations associated with insomnia utilizing an EEG-based neuromodulation device. Elevated EEG power in the alpha band (8-12 Hz) is a physiologic biomarker of insomnia disorder as demonstrated with meta analyses of multiple EEG studies. Though EEG alpha power is elevated throughout several sleep stages in insomnia disorder, it is most pronounced when subjects are awake and trying to initiate sleep. Hence, elevated alpha may be a sign of cortical hyper-arousal associated with insomnia. Sleep-inducing medications, such as benzodiazepines, have been shown to decrease alpha power at doses that induce sleep and treat insomnia. Outside of pharmacological approaches, physiological interventions to disrupt elevated alpha activity are lacking. Though presentation of visual stimuli out-of-phase with alpha rhythms decreases alpha power, visual stimulation may not be optimal for sleep health. More importantly, phase-estimation algorithms and their associated processing rates historically have limited real-time phase detection accuracy for biological rhythms with higher frequencies, including alpha. When implemented as part of an embedded system, theend point corrected Hilbert transform (ecHT) algorithm overcomes this limitation, allowing stimulus phase-locking to fast oscillation frequencies. Here, the ecHT algorithms are implemented on a wearable EEG neuromodulation device. Using this approach, instantaneous alpha phase is tracked and audible pink noise sound pulses are delivered at physiologically opposing phases of alpha. The neuromodulatory performance of this technology is demonstrated, including the ability to phase align auditory-evoked response potentials (ERPs) in the EEG to modulate downstream alpha oscillations. An in-home ambulatory study confirms this device accurately tracks EEG signals for effective sleep staging and identification of the sleep micro- and macro-events. Finally, this pilot study supports the feasibility of using EEG-based audible phase-locked sound stimulation to reduce sleep onset latency in individuals with sleep initiation problems.Part 1 : Feasibility study of the ecHT on-device to measure alpha phase, deliver phase-locked stimulation, and in turn modulate brain activityMethods

[0165] Tabletop device design: For lab-based studies including measurement of auditory-evoked potentials, a tabletop version of the wearable neuromodulation device described herein was developed. The EEG front end included amplifier stages with lOOOx gain, anti-aliasing and powerline filtering, as well as a 24-bit analog to digital converter. Two processing cores were used; the first of which was responsible for sampling the EEG (at 250- 500 Hz, selectable) and computation of the endpoint-corrected Hilbert Transform (ecHT) for phase estimation. The other processor / core generated the audio stimulus (>44.1kHz, 16bit stereo) and controlled the audio amplifier. The system architecture 200 for this device is displayed in simplified form in FIG. 4.

[0166] As shown in the architecture 400 in FIG. 4, the tabletop system featured a 2 channel EEG as input (shown), with expansion ports that include 3x GPIO pins, 2x 12-bit analog outputs, 2x 12-bit analog inputs, and 2x serial ports (not shown). The expansion ports allow researchers to add input and output devices as needed. This figure depicts the major components of the primary neuro-feedback path: the EEG, processing, and audio playback. Each component can introduce some delay, which contributes to an end-to-end system delay of ~ 1.4ms. For example, the EEG can have a delay from the electrode voltage to the digital sample of about 0.002 ms. The processor can have a delay in processing time for the ecHT of about 1-2 ms, which can be selectable based on the resolution of the phase computation. Theaudio playback can have a delay between the digital trigger to audio to sound pressure of about 0.36 ms.

[0167] Wearable neuromodulation device design: A wearable version of the tabletop research device, wearable device 500 illustrated in FIGS. 5A-5B, was produced with dimensions of 70 mm x 45 mm x 20 mm and a weight of 43.9 g, including a single-cell, 1000 mAh capacity rechargeable lithium-polymer battery. The processor unit 502 was designed to attach to a commercially-available fabric headband 504 (Muse-S Gen-2, InteraXon, Toronto, Ontario, Canada) that contained three (3) flexible dry recording electrodes 506 at positions approximating Fpl, Fpz, and Fp2, two linked reference electrodes (L Ref) 508 positioned at the skin just above the ears, and a ground electrode (G) 510 adjacent to Fpz. To eliminate the need for headphones, auditory stimulation was delivered via a bone-conduction driver 512 with a 22 kHz sample rate.

[0168] FIG. 5A illustrates a front % view of the device 500 comprising processor unit 502 (“puck”) attached to the Muse S (2nd gen) headband 504. In FIG. 5A, the left over-the-ear reference electrode 508 is shown (there is also a right over-the-ear reference electrode, not shown).

[0169] FIG. 5B illustrates a back view of the device 500 showing the left over-the-ear reference electrode (E Ref) 508, recording electrodes 406 Fpl, Fpz, and Fp2, ground electrode (G) 510, and the bone conduction driver 512. The bone conduction driver 512 was added to the Muse S headband 404. The puck 502 is electrically connected to each of the electrodes 506, 508, 510 of the Muse S headband 504 and drives the bone conduction driver 512 (e.g., provides the audio signal that the bone conduction driver 512 applies to the user’s forehead).

[0170] Electrode quality check, EEG recording, initiation of experimental conditions and data transfer were all controlled through a smartphone and custom application provided for each subject. An example graphical user interface (GUI) 600 for this is illustrated in FIG. 6. FIG. 6 illustrates a screenshot of the smartphone application interface 600 after a successful (8 out of 10 identified blinks) and 2 out of 3 scalp electrodes (Fpl and Fpz) with adequate signal strength (>5 pV RMS). If six or fewer blinks are detected by the processor, or only one scalp electrode has adequate signal strength, the wearer is prompted to move the band around on her / his head a little and try the blink test again.

[0171] Device control with the smartphone was mediated through Bluetooth Eow Energy(BLE). Data transfer was possible through Bluetooth or a direct USB-C connection, which was also used for charging. Apart from the app, the user interface on the device was a singleRGB LED indicator light, volume up and down buttons, and an “activity” button which was used to perform a hardware reset. The system had other sensors which were not used in this study, including a 3-axis accelerometer, microphone to measure ambient volume, and an ambient light sensor. The system architecture for the ambulatory system (and the software architecture for this device) is shown in FIG. 3 (the larger block represents a multi-core processor and Memory represents NAND flash memory).

[0172] As described herein, FIG. 3 illustrates the software architecture for the ambulatory device. The left side of the diagram shows the inputs: EEG 310 and a variety of ambient sensors 315. Only EEG sensors were active for this study. Configuration scripts 325 define the parameters used to alter the stimulation paradigm. During the at-home study the scripts represent different study conditions under test (Stim or Control). Core 1 of the processor 305 is responsible for computing when to stimulate, and runs the ecHT algorithm. Core 2 generates the audio 335 and drives the low-latency playback system. Finally, all the sensor data as well as many intermediate computational derivations are logged in memory 330 and can be downloaded or live- streamed for additional (computer, phone, or cloud based) analysis.

[0173] EEG experimental procedures and measures: Through a data sharing agreement, EEG measures of auditory evoked response potential (ERP) latencies were obtained from a study conducted at the University of Connecticut. Scalp EEG was recorded with wet electrodes from Fpz with reference at the mastoid (Ml) using the tabletop device (ENModvl, Elemind Technologies, Inc). First, in a 3-minute recording session with eyes closed, the EEG power spectrum was obtained and 1 / f detrended to estimate the individualized alpha frequency (IAF) using standard methods. Secondly, in a 15-minute EEG recording session with eyes open, pink-noise sound pulses were played at random phases of ongoing alpha to measure the frontal (Fpz) auditory ERP delay. Finally, in a 15-minute EEG recording session with eyes closed, subjects were instructed to ignore the sounds while rehearsing multiplication tables for a subsequent verbal test. Pink-noise sound pulses were played to determine whether individually tailored peak and trough phase-locked auditory stimulation had distinct effects on alpha oscillations.

[0174] For the in-home study, EEG data collection and real-time phase calculation and sound delivery were performed on the wearable version of the ecHT devices. EEG signals were sampled at 250 Hz, and then bandpass-filtered with a lowpass cutoff of 35 Hz and a highpass cutoff frequency of 2.5 Hz. The channel with the highest recording quality at a given time was chosen algorithmically to be used for phase estimation. In order to determinethe optimal channel, RMS values for the EEG signal were computed for 5-second windows. If the RMS for the current channel dropped below a fixed threshold, the RMS values for the remaining 2 channels were compared, and the channel with the highest value was selected to be active.

[0175] Android App: A mobile application (e.g., an Android app) could be remotely programmed by research staff to schedule the different stimulation conditions. In exemplary embodiments a file with numerous configuration parameters (e.g., alpha frequency range and many others: virtually any parameter mentioned herein) is generated by research staff and saved to a file in AWS, and the device processor accesses the file via AWS, reads the file, and configures itself accordingly.

[0176] Subjects were asked to start stimulation / data recording on the appropriate nights. It is worth noting that the exemplary design of the prototype headband used in these examples did not have the ability to measure channel impedances, as discussed above. Subjects instead used the “Electrode Quality Check” smartphone app (FIG. 6) to obtain two different measures of signal strength including: detection of peak voltage activity to cued eyeblinks and signal RMS energy to resting state activity. For eyeblink detection, subjects were guided via auditory cues to blink 10 times, and voltages were measured at one or more channels timed after to the blink cues (blinking induces an easily measurable voltage).Voltage peaks exceeding +100 pV within a 1-second window of the auditory cue were scored as a detected blink. A successful blink test was defined as 7 or more detect blinks. Following the blink test, the app interface 600 showed a simplified graphical representation of the headband recording channels’ 5-second broadband RMS signal strength. Channels with RMS signal strength <1 pV were shown in red (illustrated by a dashed outline circle in FIG. 6), signal strength >1 pV and <5 pV in orange (not illustrated in FIG. 6), and >5 pV in green (illustrated by a solid outlined circle in FIG. 6).

[0177] Data were uploaded and analyzed once all study equipment was returned back to the research staff at the end of the study. Although the headband did not have the ability to upload the large EEG data logs remotely, it could remotely upload small (1 KB) metadata files to a secure cloud server detailing data log recording times and file sizes. Research staff monitored study progress through these metadata logs. Subjects were contacted directly when any lapse in protocol or schedule was suspected.

[0178] Real-time phase estimation with the endpoint-corrected Hilbert Transform (ecHT): A custom endpoint-corrected Hilbert Transform (ecHT) algorithm, described in greater detail above, was implemented on the tabletop and wearable devices. FIG. 7summarizes how HT is modified to generate the ecHT. The ecHT uses an infinite impulse response causal bandpass filter, whose parameters (gain, poles, and zeros) were originally empirically optimized on sine-waves to preserve the phase of the most recent sample point. This computation estimates the instantaneous phase of an oscillation at the leading edge of a sample window, overcoming the phase error inherent to existing approaches, such as the Hilbert Transform or the phase-locked loop. The methods and error corrections made possible with the ecHT computation have been described in detail previously in the context of determining the phase of a tremor oscillation. As depicted by the diagram in FIG. 7, the EEG signal is sampled and decomposed into spectral components with a fast Fourier transform (FFT). A causal bandpass filter is used to filter the frequency-domain representation of the data to correct for endpoint effects. Then an inverse FFT is used to compute the complexvalued analytic signal. At this step, the instantaneous phase and amplitude of the most recent sample is equal to the phase and magnitude of the last complex number in the analytic signal. The amplitude and phase at the beginning of the sample window is warped to be continuous with the end. The end amplitude and phase is untouched.

[0179] Auditory stimulation for evoked response measures: To estimate population mean ERP latencies and effects on alpha oscillations, EEG was recorded in a sound isolated room and sound delivery was controlled by the tabletop device with audio output through Etymotic ER-4P earphones. Short duration (12 ms), high intensity (e.g., 78-82 dB, such as 85 dB) pink noise sound pulses were played at random phases relative to alpha with variable inter- stimulus intervals. The noise pulses were presented while the subjects sat with eyes open watching a silent movie with subtitles to pass time. To assess the phase-dependent effects on alpha oscillations, the short duration (12 ms), high intensity (85 dB), pink noise sound pulses were played at peak and trough phases aligning the early ERP (Pl) component to the individual subject’s alpha trough and peak phase, respectively, with variable interstimulus intervals. For example, at least 750 ms passed between pulses. Each session lasted approximately 8 minutes, during which approximately 1080 pulses were presented. EEG data was bandpass-filtered between 2 and 30 Hz, and an analysis window was drawn from -250 to 500 ms from the time of sound onset to sort the data into epochs. Any epoch with a peak signal >±100 pV was discarded. The auditory evoked response potential (ERP) was then computed by averaging across epochs. From this average, a Pl search window was defined between 35-75 ms. Pooling data from all subjects, a mean P50 latency was determined to be approximately 62.4 ms. Additional details of this computation are described in greater detail above (e.g., block 125 in FIGS. 1A and 1C).

[0180] Computation of auditory stimulation timing for optimal and pessimal phase: To determine the optimal and pessimal stimulation times, the individual alpha center frequencies (IAF) and P50 latencies were measured on a separate cohort of subjects (n=21) in a laboratory setting, using the benchtop version of the device. EEG was recorded from Fpz and Fz and sampled at 500 Hz. Subjects were instructed to remain seated with eyes closed while EEG was passively recorded for two minutes. From the timeseries EEG data, a 4-taper spectrogram was computed using a 6-second sliding analysis window and step size of 150 ms (FIG. 2). From these data, the IAF peak was determined by fitting a 3rd-order polynomial to the across-time median spectrum and subtracting the fit from the power spectral density plot to remove 1 / f noise. From here, the peak could be identified as the most prominent peak in the 7.5 - 14.0 Hz range. Pooling data from all subjects, a median IAF was determined to be about 10 Hz. Using the computed population average values for IAF and Pl, the optimal and pessimal phases were computed according to the formulas below:Trough Phasedegrees)= wrapTo360[360° X Plseconds)X MF(Hz)]where wrapTo360 is a MATEAB function mapping angles into the range [0, 360].

[0181] Auditory stimulation for in-home sleep study: Auditory stimulation on the wearable device used in at-home testing was provided through a bone conduction driver (bone conduction driver 512) with 22 kHz mono WAV playback positioned in the middle of the user’s forehead. The phase-locked auditory stimulus the subjects received was a combination of phase-locked pink noise sound pulses combined with a background natural rain sound. Through prior pilot studies, we determined that subjects best tolerated the high intensity pink noise stimulation when it was presented against a background of a gentle rain sound (Eight of Mind ©). The phase-locked pink noise sounds were played at 18 dB above the background sound to ensure minimal masking and an effective signal-to-noise ratio. Accordingly, the sound pressure level (SPE) calculated from the root mean square (RMS) energy was 47 dB and 65 dB for the background rain and pink noise pulses, respectively. This sound combination maintains alpha phase-locking precision (described in greater detail below with respect to FIGS. 11A-11C). Because the perceived loudness of the audio depended on both the tightness of the headband fit and background noise levels, subjects were instructed to adjust the volume of the background rain until it was just audible.

[0182] FIGS. 8A-8B illustrate how the auditory stimulus is delivered to align the early (Pl) auditory evoked response potential (ERP) in-phase (800a) or anti-phase (800b) with thesubsequent peak or trough of alpha, respectively. As described above, aligning the ERP latency with these two target phases of alpha requires estimating the ERP latency as well as the alpha frequency. The two different phase conditions were based on timing of the Pl peak component of the evoked response to a single 12-ms pink noise burst such that the auditory evoked response arrived in-phase or out-of-phase with a 10-Hz alpha oscillation. The pink noise phase-locking was determined by the instantaneous phase of the subject’s alpha oscillations, such that the onset and offset of the stimulus appeared at 134° to 224° (alpha trough) and 314° to 44° (alpha peak). As shown in FIG. 8A, accounting for an average ERP- P1 latency of 62 ms, the auditory stimulus was delivered at a phase of 134° so the ERP-P1 component coincided with the alpha peak. As shown in FIG. 7B, accounting for the same ERP-P1 latency of 62 ms, an auditory stimulation at 314° produced an ERP-P1 component coincided with alpha trough phase.

[0183] Subjects and study design for ERP study: Data for the ERP study was collected at the University of Connecticut and reported here as part of a data sharing agreement with Elemind Technologies, Inc. Subjects were recruited locally to take part in a daytime ERP study consisting of three phases: 1) an individual alpha frequency (IAF) measurement, 2) and ERP delay measurement, and 3) a two-phase auditory ERP measurement. The total experimental time was roughly 50-60 minutes per participant. 22 participants took part in the study, of which 21 were included for analysis. Of those included were 11 males and 10 females, average age 22.75 years (range 18-38). All procedures were approved by the Institutional Review Board at the University of Connecticut.

[0184] Sleep Study Recruitment, Inclusion and Exclusion Criterion: Potential study candidates were recruited using advertisements and postings on social media sites (e.g., Linkedln, Twitter) targeting individuals who regularly report having problems initiating sleep within 30 minutes. Respondents were directed to an online screening survey to determine study eligibility. To be included in this study, adults between the ages of 25 and 55 years of age had to meet the following criteria: 1) fluency in English, 2) access to the internet or other cellular data services, and 3) evidence of moderately severe clinical insomnia as determined by an Insomnia Severity Index (ISI) score > 21 (Morin et al., 2011) and a Pittsburgh Sleep Quality Index (PSQI) score > 5. Candidates that were excluded met any one or more of the following criteria: 1) with clinically-confirmed sleep apnea, 2) a current or past history of any neurological or psychiatric disorder, 3) at high risk for generalized anxiety disorder (GAD-7 > 15), 4) at moderate to high risk for alcohol abuse disorder (AUDIT-C > 6), 5) diagnosed with a hearing impairment 6) with a BMI > 33, 7) night shift workers, and 8) pregnantwomen. Subjects who reported taking antidepressants, stimulants, medication for hypo / hypertension, cannabis or cannabis-derived products, or consuming more than 4 caffeinated beverages per day were also excluded. In total, 24 subjects (13 male, 11 female, average age 33.0 ± 6.6 yrs (mean ± SD), median 31 yrs, range 26 to 55 yrs) met the study criteria. All subjects received informed consent following the guidelines outlined by Elemind Technologies’ institutional review board, Solutions IRB (Yarnell, AZ).

[0185] At-home data collection: All data collected in this study was performed by the subjects themselves while at home. Study hardware and materials were delivered via FedEx or UPS courier service and included a prototype EEG-based neurostimulating headband, an Android smartphone programmed with a customized companion app, and a wrist- worn activity tracker (Philips Respironics Actiwatch). All subjects were trained on the use of all study equipment and procedures during a 30-40 minute secure video call. Research personnel were available through email, text, or phone to address any questions or problems.

[0186] Sleep Study Design: The pilot feasibility sleep study was a single arm, randomized, control, subject-blind, at-home study comparing outcome measures across 3 experimental conditions including 1) “No Audio” , 2) “Alpha Peak Phase-locked Audio”, and 3) “Alpha Trough Phase-locked Audio.” Both experimental alpha phase-locked sounds were combined with continuous broadband rain sound and delivered for 30 minutes following lights out.

[0187] Following the video training call (Wednesday), all subjects were instructed to wear the headband without any of the data recording or stimulation features enabled for two consecutive nights (Wednesday and Thursday night) to acclimate to the experience of sleeping with the study equipment. Subjects also started wearing their activity trackers and marked their bedtime and wake times throughout the remainder of the study. Beginning on the following week, subjects were randomly assigned one of three different experimental conditions for four consecutive nights (Monday-Thursday night): 1) a control condition with no auditory stimulation, 2) a stimulation condition with pink noise pulses presented during peak alpha phase, and 3) a stimulation condition with pink noise pulses presented during the trough alpha phase. Phase-locked stimulation was presented continuously for 30 minutes for both stimulation conditions (peak and trough).

[0188] For each night of the three-week data collection series (FIG. 13), subjects were given detailed instructions regarding their nightly routine. Each night, subjects placed the headband on their head before conducting their pre-bedtime routine to ensure the electrodes had enough time to establish contact to their skin. Once in bed, subjects checked the electrodeconnection strength using the “blink test” and “signal strength” tests (mentioned above). Following a successful signal quality test, subjects were instructed to turn off all other technology devices, shut off the lights, mark their bedtime on the activity watch, start the headband EEG / neurostimulation session, close their eyes, and go to sleep. Researchers stressed the importance of following this exact routine each night in order to ensure consistency of the data collection across nights and subjects.

[0189] On the mornings following each data recording session, subjects plugged in their headbands and phones for recharging, uploaded their EEG metadata to a secure cloud server, and completed a morning survey questionnaire. The morning survey consisted of 11 questions taken from Consensus Sleep Diary and additional questions related to their subjective experience wearing the neurostimulation headband. At the conclusion of the three- week study, subjects returned all study equipment and were financially compensated for the participation time.

[0190] Sleep Study Data analysis: Sleep staging was performed by a 3rd-party registered polysomnographic technologist (Sleep Strategies, Inc. Ottawa, Canada). EEG data logs were qualitatively assessed and ranked for signal quality. These data logs were subjectively sorted by quality, data conditions blinded to the scorer and delivered in random order in batches of 10 data logs. Visual scoring was complete / stopped when both the registered sleep tech and Elemind staff scientist agreed that the signal quality of the EEG data no longer contained information consistent with the AASM scoring guidelines. For the analysis, weekly average sleep onset latency times to the first epoch of N2 sleep were calculated from all available datasets (i.e., not all subjects had a complete complement of 4 nights of EEG data per condition).Results

[0191] The results demonstrated that EEG signals collected from the headband showed measurable neural oscillations and sleep microevents that were suitable for tracking sleep stage. First the ability of the ambulatory EEG system to record neurological signals was assessed with sufficient fidelity for effective measurement of sleep-related activity. As shown in FIG. 9, datasets collected from sleeping subjects showed typical spectrotemporal features of sleep, including oscillatory activity in the alpha, beta, theta, and delta bands. Additionally, sleep-related microevents were also visible. Activity characteristic of sleep spindles, k- complexes, and vertex waves was also observed. Though spectral distortions were observedat approximately 2 Hz due to a high-pass filter, data were suitable for standard visual sleep stage scoring by independent certified technicians.

[0192] The wearable device records EEG with sufficient fidelity to distinguish oscillatory activity characteristic of sleep. The time-frequency plot 902 in FIG. 9 illustrates spectral power (yellow max, blue minimum) recorded in the first two hours of overnight sleep with the wearable headband in one subject. The overlaid hypnogram in the plot 902 (black line, y- axis on right) indicates four sleep stages scored by a clinically certified sleep technician. Sleep stages include Awake (eyes closed) behavior, Rapid Eye Movement (REM), and three progressively deeper non-REM stages of sleep (i.e., stages 1, 2, 3 or Nl, N2, N3). The timevoltage signals in graphs 904, 906, 908, 910, and 912 were suitable for standard visual sleep scoring, as shown for data from the same recording shown in plot 902. “Awake” stage 904 was dominated by EEG oscillations in the alpha (8-10 Hz) spectral range. Stage- 1 (Nl) sleep was associated with increased theta (4-7 Hz) activity (906) and vertex waves (908). Stage-2 (N2) sleep 910 was associated with increased incidence of k-complex and spindle oscillations micro-events. Stage-3 deep sleep (N3) 912 was dominated by delta slow wave (0.5-4 Hz) oscillations.

[0193] Some datasets collected by subjects at home did not yield high-quality data. These datasets were typically characterized by broadband spectral activity measured on all 3 electrodes or a lack of any obvious neural activity. Of the 257 recorded datasets, 77 were found to be unusable due to signal quality issues, outlined below in Table 1. Hardware malfunctions were ruled out in the majority of cases; therefore, these issues likely resulted from poor electrode contact with the subjects’ scalp or degradation of the electrode fabric material. Similar kinds of data loss were observed on specific channels on the comparator commercial system when the associated electrodes became disconnected while the subject was asleep, suggesting that loss of sufficient electrode contact was at issue in both cases.Table 1. Tabulation of datasets from all nights (288 data collection nights) of sleep recorded during the at-home study

[0194] 31 of 288 missing data logs reported in Table 1 were never recorded due to user error or technical failures. 180 of 257 RPSGT scored data logs were of sufficient quality for a registered sleep technician to score. 77 of 257 unscorable recorded data logs had insufficient data quality for sleep scoring.

[0195] The results demonstrated that on-device implementation of ecHT results in minimal phase error when tracking fast frequency neural oscillations. To evaluate the performance of the ecHT algorithm for tracking phase with minimal delay, the precision of the phase estimate was computed using a post-hoc analysis of recorded EEG activity. FIGS. 11A-11C demonstrate mean phase angles that are represented by dark grey (onset) and light grey (offset) vector lines; arcs at the edge of the polar axis represent one standard deviation. The phase locking value, a normalized metric of phase coherence, is represented by the length of the vector line relative to the radius of the polar axis. The radial axis labels denote the probability distribution values for the radial histogram bins. The bin size for FIG. 11 A is 5°, and the bin size for FIGS. 1 IB-11C is 20°.

[0196] FIG. 11 A demonstrates the phase accuracy of phase-locked pink noise onset and offset events for a single representative 30-minute stimulation session targeting the peak phase of alpha. In FIG. 11A, distribution of target onset (dark grey: 314°) and offset (light grey: 44°) phases for pink noise pulses near the alpha peak phase for a single 30-minute phase-locked stimulation interval are shown. The average phase error for the onset stimulus events was -1.07° ± 43.8° and -8.87° ± 49.2° for offset events (mean ± SD). These results are also summarized below in Table 2A.Table 2A. Single-session phase accuracy summary

[0197] FIGS. 1 IB- 11C summarize the across-session average phase accuracy for stimulation to alpha at trough and peak phase. FIG. 1 IB demonstrates the session-averaged (n = 86 datasets) mean onset and offset phase values for stimulation targeting the trough phase of the alpha cycle. For stimulation targeting the trough phase of alpha (FIG. 1 IB), the average across-session (n = 86 sessions) phase error was -5.39° ± 12.6° for onset events and - 15.5° ± 18.7° for offset events (mean ± SD). FIG. 11C demonstrates the session-averaged (n= 88 datasets) mean onset and offset phase values for stimulation targeting the peak phase of the alpha cycle. For stimulation targeting the peak phase of alpha (FIG. 11C), the average across-session (n = 88 sessions) phase error was -4.04° ± 25.7° for onset events and -12.1° ± 29.7° for offset events (mean ± SD). These results are also summarized below in Table 2B.Table 2A. Single-session phase accuracy summary

[0198] The results demonstrated that auditory stimulation locked to alpha oscillations alters alpha activity in a phase-dependent manner. Based on known neurophysiology, it was hypothesized that audio stimulation would differentially modulate alpha when the early ERP component (Pl) arrives during the excitable trough phase compared to the inhibited peak phase of alpha. In an initial session, the latency of the ERP Pl was measured in response to pink-noise sound pulses delivered at random phases with eyes open (FIGS. 8A-8B). In a second session, the alpha spectral band evoked response oscillation (ERO) was measured by using the predetermined ERP delay to deliver alpha peak or trough phase-locked pink-noise sound pulses. For both sessions, the auditory stimulus ERP Pl component delay was the same (64.2 ± 8.8 ms). As shown in FIGS. 12A-12B, stimulation timing relative to phase of alpha rhythms alters subsequent alpha wave amplitude. On average, alpha oscillation amplitude following auditory stimulation with the peak phase-locked sound (FIG. 12B, solid line) was reduced relative to the alpha oscillation amplitude with trough phase-locked sound stimulation (FIG. 12B, dashed line). In FIG. 12A, the P50 and N100 components of the grand average broadband auditory evoked potentials (AEPs) are reduced when pink noise pulses are presented at the trough phase of the alpha cycle (dashed). In FIG. 12B, bandpassed AEPs to the alpha band (7.5-12 Hz) show stimulation to alpha peak disrupts the post-stimulus alpha oscillations. Stimulation at the trough of the pre-stimulus alpha cycle does not impede the subsequent post-stimulus alpha cycles. These data provided a population estimate of average pink-noise sound evoked potential delays to use in the sleep study in order to influence alpha.Moreover, this supported the hypothesis that auditory stimulation modulates alpha oscillations in a phase-dependent manner.Part 2: Feasibility study of phase-locked acoustic stimulation to reduce sleep onset latency Methods

[0199] Alpha oscillations are elevated in wakefulness and throughout multiple sleep stages in subjects with insomnia disorders including those with sleep initiation problems. Given the observation that peak and trough phase-locked sounds differentially impacted alpha oscillations (FIGS. 12A-12B), this feasibility study set out to compare stage-2 sleep onset latency (SOL-N2) under three sleep conditions including: 1) no sound stimulation; 2) peak (314°) and 3) trough (134°) phase-locked sound stimulation.

[0200] To minimize potential effects of home environment sounds and to make the sounds more soothing, phase-locked pink-noise sounds were played jointly with background natural rain sound at a high (18 dB) signal-to-noise ratio. Subjects were screened and included in the study based on scores assessed with the insomnia severity index, Pittsburgh sleep quality index, and self-report that it took them 30 minutes or more to fall asleep regularly (Table 1). Following screening, 24 subjects (11 male, 13 female; Age: 33.0 ± 6.6 yrs (mean ± SD), Median age: 31 yrs, Range 26-55 yrs) completed a 3-week, randomized, cross-over in-home sleep study (FIG. 11). In a video meeting, study personnel walked subjects through the daily routine for operating the wearable headband device with a smartphone application (FIG. 6). As described above, in the at-home sleep study, subjects wore the headband without audio stimulation or EEG data recording for two adaptation nights prior to data collection. Four nights (Mon-Thu) of EEG data were collected each week for 3 weeks of testing across 3 separate conditions: control (no audio stimulation), phase- locked stimulation to alpha peak, and phase-locked stimulation to alpha trough. Conditions were randomized and counter-balanced across weeks. On the randomly assigned weeks with either type of experimental phase-locked auditory stimulation, the sounds were played for 30 minutes during sleep initiation. Actigraphy data collection began on the day of the first adaptation night and continued throughout the duration of the study.Results

[0201] To quantify the sleep stages and sleep onset latency outcome measures, deidentified EEG data was scored by independent sleep scoring technicians across all recording nights. FIGS. 14A-14E illustrate changes in sleep onset latency for subjects in control,trough, and peak stimulation conditions. In FIGS. 14A-14E, box boundaries show the interquartile range, while dashed horizontal lines depict the median. Solid horizontal lines represent the means of each population. Filled (black) circles represent poor sleepers. Unfilled (white) circles represent good sleepers.

[0202] FIG. 14A illustrates graph 1400a demonstrating sleep onset latencies for all subjects with scorable datasets (n=24), and graph 1400b demonstrating a reduction in sleep onset latencies for each stimulation condition relative to control. As shown in FIG. 14A, for data pooled across all (n=24) subjects, there was no significant difference in stage-2 sleep onset latency (SOL-N2) across the three experimental conditions (no audio: 16.0 ±14.6 mins; trough phase (134°) 13.3 ±8.1 mins; peak phase (314°) : 12.3 ±8.7 mins; one-way ANOVA(2,66) = 0.9938, p = 0.3756). Despite the inclusion criteria, 71% (n=17 of 24) of subjects did not display objective EEG verified sleep onset latencies greater than 30 minutes. The lack of effect on the full population of participants could reflect the variation in sleep initiation problems across weeks or alternatively a subjective misperception of sleep onset problems in 71% of the subjects included in the study.

[0203] FIG. 14B illustrates graphs 1400b and 1410b demonstrating sleep onset latencies for individuals with at least one night in which sleep onset latency was measured to be greater than 30 mins by visual EEG scoring (n=7; p < 0.05). As shown in FIG. 14B, 29% (n=7) of subjects were objectively defined as “poor sleepers” with sleep initiation problems confirmed by at least one night of EEG verified SOL-N2 of 30 minutes or more. During weeks when the neuromodulation headband was programmed to deliver alpha phase-locked sound stimulation, these “poor sleepers” had a significantly reduced average SOL-N2, as shown in FIG. 14B. A one-way analysis of variance found weekly average SOL-N2 was significantly reduced by phase-locked auditory stimulation versus no-audio conditions (F(2, 18) = 5.8478, p = 0.011). Tukey’s HSD test for multiple comparisons found that phase-locked stimulation to both the peak cycle of alpha (p = 0.0125, 95% C.I. = [-35.2614, -4.0958]) and trough cycle of alpha (p = 0.0453, 95% C.I. = [-31.4697 -0.3041]) were significantly different from the control condition without stimulation. There was no statistically significant difference between the peak and trough stimulation conditions (p = 0.8106). Without stimulation, average weekly sleep onset time was 35.3 (± 13.0) minutes. With phase-locked auditory stimulation to alpha trough and peak sleep onset time was 19.4 (± 10.8) and 15.6 (±10.3) minutes, reflecting a reduction in SOL-N2 by 15.9 (±22) and 19.7 (± 20.4) minutes, respectively (FIG. 14B).

[0204] FIG. 14C illustrates graph 1400c demonstrating a difference in sleep onset latency for individuals who did not have any observable sleep onsets greater than 30 min (“good sleepers”), and graph 1410c demonstrating a reduction in sleep onset latencies for each stimulation condition relative to control.

[0205] While onset times to non-REM Stage 2 (N2) sleep were improved with phase- locked auditory stimulation, the other two measures of sleep onset latency (actigraphy and subjective morning surveys) failed to reflect similar improvements. FIG. 14D illustrates sleep onset latencies for poor sleepers, as measured by actigraphy (Philips Actiwatch). Poor sleepers were identified as those subjects whose visually-scored N2 onset latencies were greater than 30 minutes for at least one of the four control (no audio) nights. As shown in FIG. 14D, weekly average sleep onset as measured by actigraphy was 19.2 ± 13.1 minutes for the baseline (no audio) condition, 13.2 ± 6.5 minutes for phase-locked stimulation to alpha trough (134°), and 10.2 ± 7.6 minutes for stimulation to alpha peak (314°). A one-way analysis of variance confirmed no significant effect of stimulation condition [F(2, 18) = 1.6070, p = 0.2280],

[0206] FIG. 14E illustrates subjective sleep onset latency for poor sleepers as measured by a morning survey. Poor sleepers were identified as those subjects whose visually- scored N2 onset latencies were greater than 30 minutes for at least one of the four control (no audio) nights. As shown in FIG. 14E, weekly average subjectively-reported sleep onset times were 30.2 ± 10.4 minutes the baseline (no audio) condition, 25.3 ± 7.5 minutes for stimulation to alpha trough, and 21.4 ± 8.9 minutes for stimulation to alpha peak [one-way ANOVA: F(2, 18) = 1.6598, p = 0.2180],Discussion (Part 1 & 2)

[0207] This feasibility study developed and tested a wearable device for accurate closed- loop auditory stimulation phase-locked to alpha oscillations. The device employed an endpoint-corrected Hilbert Transform algorithm implemented with efficient EEG signal processing and device control operations. First, these results demonstrate feasibility for realtime EEG signal processing technology to accurately track instantaneous alpha phase and deliver phase-locked audible sound stimulation to modulate alpha in a phase-dependent manner. This modulation is consistent with the neurophysiology of synchronous rhythmic neural oscillations such as alpha and supports the hypothesis that sounds phase-locked to the peak of alpha such that the ERP arrives during the trough are more likely to disrupt alpha oscillations (FIGS. 12A-12B). Secondly, the current results demonstrate feasibility toimplement this alpha phase-locking technology on a wearable, non-invasive, EEG-based neuromodulation device for home use. During the overnight sleep study, objective EEGbased measures of sleep onset latency (SOL-N2) were reduced for poor sleepers by up to ~20 minutes with peak phase-locked auditory stimulation versus a no auditory control condition (FIG. 14B). Importantly, the magnitude of these effects are within the range of sleep onset latency reductions (i.e., 10 to 17 minutes) observed across multiple studies with sleep hypnotics (e.g. zolpidem) versus placebo drugs and for an FDA certified thermal cooling headband versus placebo. This is the rationale for continued testing and refinement of phasedependent audible sound stimulation for disrupting alpha and promoting healthy sleep initiation. Finally, subjective reports and wrist- worn accelerometry measures of sleep onset latency were not different across control “no audio” versus either phase-locked audio condition. However, subjective measures and non-EEG-enabled wearables have been shown to be poor estimates of sleep stage, highlighting the need for wearable devices capable of recording neural activity.

[0208] Beyond demonstrating the feasibility of this approach, this study yielded additional insight into features that could improve future iterations of wearable EEG devices to help ameliorate hyperarousal and promote healthy sleep. Although many of the subjects’ recordings contained EEG signals that were scorable by an RPSGT, in the at-home study, it was observed that 30% of datasets did not contain usable data. This outcome was primarily related to issues around poor contact between electrodes and the skin or degradation of the electrodes themselves. Without the ability to effectively record neural data, closed- loop approaches like the one presented here are not possible. Long-term durability or easy replacement of electrodes is therefore an important area of improvement, and several groups are currently working on this problem. Finally, 10.8% of subject-nights in the study did not yield datasets at all. Though it was hard to identify a single issue that resulted in these losses, many lost nights of data were related to user error. Designing neuromodulation systems that are intended for consumer use will need to address issues of usability and stability in an at- home setting. One such approach may be to provide easy-to-understand feedback about device function and performance, such as signal quality or electrode impedance, so that users can better understand how to achieve optima. As part of the evaluation of the device, an in- home sleep study was performed in which subjects applied the device and recorded data on their own after a virtual training session with researchers. This study yielded insight into the usability of the neuromodulation system described herein, as well as tested the use of closed- loop auditory neuromodulation as an intervention for accelerating sleep initiation. Afterrestricting the analysis to datasets with scorable EEG data, a significant effect of stimulation (either optimal or pessimal) was not observed on sleep onset latency (SOL) across the whole subject population, relative to sham stimulation (no sound). Interestingly, despite self-report of symptomatic insomnia, including an Insomnia Severity Index (ISI) score of > 21 and PSQI score > 5, the majority of the subjects did not display sleep initiation difficulties as scored by EEG. These observations are not inconsistent with what is known about insomnia patients, of which up to 50% display discrepancies between subjective reports of sleep and what can be measured objectively. Indeed, a recent study found that insomnia can be subdivided into at least 5 different subtypes with unique characteristics that remain stable over time. In the data, it was also observed that a subset of the subjects, specifically those with observable SOLs > 30 min, appeared to respond favorably to the stimulation protocol, while those with shorter SOLs did not. These differences in response may be reflective of unique features or etiologies underpinning the various subtypes of insomnia, and suggests the possibility that alpha rhythms may be more important in certain forms of insomnia than others. Alternatively, the lack of effect observed in subjects with low SOLs may be due to a floor effect, in which the approach described herein cannot accelerate sleep onset below a certain threshold.

[0209] The study provided herein included 3 test conditions: a control condition in which no audio was played, one condition in which auditory pulses were delivered at the optimal phase of alpha, and one where pulses were delivered at the pessimal phase of alpha. Based on previous observations that elevated alpha power is associated with delays in sleep onset, it was expected that auditory stimulation delivered at the pessimal alpha phase would be the most efficacious for reducing sleep onset. However, in the subset of individuals that exhibited a reduction in SOL, both stimulation conditions appeared to be equivalent. At least three possible explanations for this outcome are hypothesized. The first is that the audio alone was sufficient to affect sleep onset in these subjects. While broadband sounds have been reported to decrease sleep onset latency in insomniacs, the magnitude of the effect is less than that reported here. Another possibility is that auditory stimulation, regardless of phase, disrupts alpha enough to accelerate sleep onset. Indeed, prior studies have shown that salient and high intensity sensory stimuli can decrease the late ERP components and desynchronize alpha oscillations independent of alpha phase. Lastly, subtypes of insomniacs have been observed to differ in response to auditory-evoked neural responses. Individuals may respond differently to stimulation at particular phases of alpha, and an approach tailored to individual phenotypes may be required to achieve the maximum effect. A better understanding of the relationshipbetween elevated alpha power and sleep onset as well as additional experiments would be necessary to evaluate these hypotheses.Conclusion (Parts 1 & 2)

[0210] While previous reports have identified closed-loop auditory stimulation as a potential mechanism for improving slow-wave sleep, the device studied herein is uniquely designed to provide closed-loop auditory stimulation timed to particular phases of oscillations (e.g., fast, > 8 Hz) with high phase precision. The system was operable by subjects in an at- home setting, allowing individuals to administer therapy and collect data independently. Further developments will improve usability and possibly include the flexibility to tailor the stimulation to individual phenotypes. This approach may also have applications outside of sleep, in conjunction with other stimulus modalities by targeting conditions associated with faster oscillations such as essential tremor, or other conditions associated with measurable differences in neural oscillations. By enabling real-time analysis of oscillatory signals, wearable-based approaches such as this one could be explored as potential alternatives to invasive neuromodulation strategies.

Claims

CLAIMSWhat is claimed is:

1. A method for inducing sleep in a subject using neuromodulation, comprising: receiving oscillation data from one or more sensors configured to detect brain waves from the subject; determining a fundamental center frequency of the oscillation data; determining an evoked response delay of the oscillation data; based on the determined fundamental center frequency and the determined evoked response delay, determining a stimulation phase for the subject; and delivering a phase-locked stimulus in accordance with the stimulation phase to induce sleep in the subject.

2. The method of claim 1, wherein the fundamental center frequency is an intrinsic center frequency (iCF) of the subject, and the oscillation data comprises at least one of alpha, theta, delta, beta, and gamma oscillation data.

3. The method of claim 1 or 2, wherein the oscillation data occupies a frequency band between about 0.1-150 Hz.

4. The method of any one of claims 1-3, wherein determining the fundamental center frequency of the oscillation data comprises: receiving oscillation data from the one or more sensors while the subject is awake and with closed eyes for a duration of time; determining spectral data of the received data, wherein the spectral data comprises a noise floor; generating de-noised spectral data by subtracting the noise floor from the spectral data; and identifying a peak within a portion of the de-noised spectral data, the peak corresponding to the fundamental center frequency of the subject.

5. The method of any one of claims 1-4, wherein determining the evoked response delay of the oscillation data comprises: delivering an initial auditory stimulation to the subject, the initial auditory stimulation comprising a plurality of noise pulses delivered at random phases;receiving brain wave data based on the initial auditory stimulation; determining an auditory evoked response potential, comprising: filtering the received brain wave data using a bandpass filter; segmenting the filtered data into a plurality of epochs; and averaging at least a portion of the epochs that do not comprise artifacts or poor signals to determine the auditory evoked delay.

6. The method of any one of claims 1-5, wherein determining the stimulation phase comprises: multiplying the evoked response delay by the fundamental center frequency; multiplying a product of the evoked response delay and the fundamental center frequency by 360 degrees to obtain an intermediate value; and inputting the intermediate value to an algorithm that maps into an angular range of [0°, 360°].

7. The method of any one of claims 1-6, wherein the determined stimulation phase aligns an evoked response with an up-state of the oscillation of the brain waves or a downstate of the oscillation of the brain waves.

8. The method of any one of claims 1-7, wherein determining the stimulation phase comprises determining an onset stimulation phase and an offset stimulation phase, the offset stimulation phase based on a pulse duration, the fundamental center frequency, and the evoked response delay.

9. The method of any one of claims 1-8, wherein delivering the phase-locked stimulus comprises delivering the stimulus beginning at the determined stimulation phase and sustaining the stimulus for a fixed duration, and wherein the fixed duration comprises at least one of a fixed time duration and a fixed phase duration.

10. The method of any one of claims 1-9, wherein delivering the phase-locked stimulus comprises delivering a phase-locked stimulation pulse intermittently in accordance with a predetermined pulse train pattern defined relative to an oscillation cycle of the received oscillation data.

11. The method of any one of claims 1-10, wherein delivering the phase-locked stimulus comprises delivering a phase-locked stimulation pulse in each oscillation cycle of the received oscillation data.

12. The method of any one of claims 1-11, comprising determining at least one of an amplitude and a spectral power of the oscillation data.

13. The method of claim 12, wherein delivering the phase-locked stimulus comprises delivering the phase-locked stimulus in accordance with at least one of the amplitude and the spectral power reaching a predetermined threshold value.

14. The method of any one of claims 1-13, wherein the stimulus is a broad-spectrum sound pulse.

15. The method of any one of claim 1-13, wherein the stimulus is a continuous auditory stimulation, the continuous auditory stimulation including at least one pitch, and the continuous auditory stimulation deliverable in at least one volume.

16. The method of any one of claims 1-15, comprising receiving biomarker data from one or more sensors configured to detect biomarkers from the subject.

17. The method of claim 16, wherein the received biomarker data comprises a measurement based on an electroencephalogram (EEG) signal, the measurement including at least one of an entropy and a variability in the fundamental center frequency.

18. The method of claim 16 or 17, wherein the biomarker data comprises a measurement of cardiac function, the cardiac function comprising at least one of heart rate, SpO2, and heart rate variability.

19. The method of any one of claims 16-18, wherein the biomarker data comprises a measurement of a soluble factor, the soluble factor comprising at least one of insulin, glucose, and cortisol.

20. The method of any one of claims 16-19, wherein delivering the phase-locked stimulus comprises delivering the phase-locked stimulus in accordance with at least one biomarker reaching a pre-determined threshold.

21. The method of any one of claims 16-20, comprising updating the stimulation phase based on the received biomarker data to maximize or minimize at least one biomarker.

22. The method of any one of claims 1-21, comprising using a wearable device on the subject that comprises the one or more sensors and one or more stimulators to deliver the phase-locked stimulus.

23. The method of any one of claims 1-22, wherein inducing sleep in the subject reduces sleep onset latency and / or treats sleep insomnia.

24. A method for determining an intrinsic center frequency (iCF) of a subject, comprising: receiving oscillation data from one or more sensors configured to detect brain waves; determining spectral data of the received data, wherein the spectral data comprises a noise floor; generating de-noised spectral data by subtracting the noise floor from the spectral data; and identifying a peak within a portion of the de-noised spectral data, the peak corresponding to the intrinsic center frequency (iCF) of the subject.

25. The method of claim 24, comprising delivering a stimulus based on the determined iCF of the subject.

26. The method of claim 24 or 25, wherein the received oscillation data comprises brain wave data detected while the subject is awake and with closed eyes for a duration of time.

27. The method of any one of claims 24-26, wherein determining the spectral data comprises using a Fourier transform.

28. The method of any one of claims 24-27, wherein determining the spectral data comprises modeling the noise floor of the spectral data with a 3rd-order polynomial fit.

29. The method of any one of claims 24-27, wherein identifying the peak within the portion of the de-noised spectral data comprises identifying a peak within a frequency range of about 0.1 to 150 Hz.

30. A system for inducing sleep in a subject using neuromodulation, comprising: one or more sensors configured to detect brain waves from the subject; a processing device configured to: receive oscillation data from the one or more sensors; determine a fundamental center frequency of the oscillation data; determine an evoked response delay of the oscillation data; and based on the determined fundamental center frequency and evoked response delay, determine a stimulation phase for the subject; and one or more stimulators configured to deliver a phase-locked stimulus in accordance with the stimulation phase to induce sleep in the subject.

31. The system of claim 30, wherein at least the one or more sensors and the stimulators are provided in a wearable device wearable by the subject.

32. The system of claim 30, wherein the fundamental center frequency is an intrinsic center frequency (iCF) of the subject, and the oscillation data comprises at least one of alpha, theta, delta, beta, and gamma oscillation data.

33. The system of claim 31 or 32, wherein the oscillation data occupies a frequency band between about 0.1-150 Hz.

34. The system of any one of claims 30-33, wherein determining the fundamental center frequency of the oscillation data comprises: receiving oscillation data from the one or more sensors while the subject is awake and with closed eyes for a duration of time; determining spectral data of the received data, wherein the spectral data comprises a noise floor;generating de-noised spectral data by subtracting the noise floor from the spectral data; and identifying a peak within a portion of the de-noised spectral data, the peak corresponding to the fundamental center frequency of the subject.

35. The system of any one of claims 30-34, wherein determining the evoked response delay of the oscillation data comprises: delivering an initial auditory stimulation to the subject, the initial auditory stimulation comprising a plurality of noise pulses delivered at random phases; receiving brain wave data based on the initial auditory stimulation; determining an auditory evoked response potential, comprising: filtering the received brain wave data using a bandpass filter; segmenting the filtered data into a plurality of epochs; and averaging at least a portion of the epochs that do not comprise artifacts or poor signals to determine the auditory evoked delay.

36. The system of any one of claims 30-35, wherein determining the stimulation phase comprises: multiplying the evoked response delay by the fundamental center frequency; multiplying a product of the evoked response delay and the fundamental center frequency by 360 degrees to obtain an intermediate value; and inputting the intermediate value to an algorithm that maps into an angular range of [0°, 360°].

37. The system of any one of claims 30-36, wherein the determined stimulation phase aligns an evoked response with an up-state of the oscillation of the brain waves or a downstate of the oscillation of the brain waves.

38. The system of any one of claims 30-37, wherein determining the stimulation phase comprises determining an onset stimulation phase and an offset stimulation phase, the offset stimulation phase based on a pulse duration, the fundamental center frequency, and the evoked response delay.

39. The system of any one of claims 30-38, wherein delivering the phase-locked stimulus comprises delivering the stimulus beginning at the determined stimulation phase and sustaining the stimulus for a fixed duration, and wherein the fixed duration comprises at least one of a fixed time duration and a fixed phase duration.

40. The system of any one of claims 30-39, wherein delivering the phase-locked stimulus comprises delivering a phase-locked stimulation pulse intermittently in accordance with a predetermined pulse train pattern defined relative to an oscillation cycle of the received oscillation data.

41. The system of any one of claims 30-40, wherein delivering the phase-locked stimulus comprises delivering a phase-locked stimulation pulse in each oscillation cycle of the received oscillation data.

42. The system of any one of claims 30-41, comprising determining at least one of an amplitude and a spectral power of the oscillation data.

43. The system of claim 42, wherein delivering the phase-locked stimulus comprises delivering the phase-locked stimulus in accordance with at least one of the amplitude and the spectral power reaching a predetermined threshold value.

44. The system of any one of claims 30-43, wherein the stimulus comprises a broadspectrum sound pulse.

45. The system of any one of claim 30-44, wherein the stimulus comprises a continuous auditory stimulation, the continuous auditory stimulation including at least one pitch, and the continuous auditory stimulation deliverable in at least one volume.

46. The system of any one of claims 30-45, comprising receiving biomarker data from one or more sensors configured to detect biomarkers from the subject.

47. The system of claim 46, wherein the received biomarker data comprises a measurement based on an electroencephalogram (EEG) signal, the measurement including at least one of an entropy and a variability in the fundamental center frequency.

48. The system of claim 46 or 47, wherein the biomarker data comprises a measurement of cardiac function, the cardiac function comprising at least one of heart rate, SpO2, and heart rate variability.

49. The system of any one of claims 46-48, wherein the biomarker data comprises a measurement of a soluble factor, the soluble factor comprising at least one of insulin, glucose, and cortisol.

50. The system of any one of claims 46-49, wherein delivering the phase-locked stimulus comprises delivering the phase-locked stimulus in accordance with at least one biomarker reaching a pre-determined threshold.

51. The system of any one of claims 46-50, comprising updating the stimulation phase based on the received biomarker data to maximize or minimize at least one biomarker.

52. The system of any one of claims 30-51, comprising using a wearable device on the subject that comprises the one or more sensors and one or more stimulators to deliver the phase-locked stimulus.

53. The system of any one of claims 30-52, wherein inducing sleep in the subject reduces sleep onset latency and / or treats sleep insomnia.

54. A non-transitory computer-readable storage medium storing instructions that, when executed by a processing device operatively coupled to one or more sensors configured to detect brain waves of a subject, cause the device to: receive oscillation data from the one or more sensors; determine a fundamental center frequency of the oscillation data; determine an evoked response delay of the oscillation data; based on the determined fundamental center frequency and evoked response delay, determine a stimulation phase for the subject; and deliver a phase-locked stimulus in accordance with the stimulation phase to induce sleep in the subject.

Images