Systems and methods for identifying sleep states of a patient based on middle ear muscle activation

Abstract

Systems and methods for detecting the occurrence of middle ear muscle activation (MEMA) events in physiological signals obtained from one or more wearable sensors, e.g., in-ear sensors, worn by an individual, and identifying one or more sleep states of the individual, e.g., REM sleep, based on the detected MEMA events are provided. Additional physiological information measured via one or more additional sensors may be used in conjunction with the signals obtained from one or more wearable sensors to identify and remove non-MEMA events, e.g., artifacts, to more accurately identify MEMA events. The identified sleep states of the individual may be graphically presented to a user via a mobile device operatively coupled to the one or more wearable sensors.

Description

SYSTEMS AND METHODS FOR IDENTIFYING SLEEP STATES OF A PATIENT BASED ON MIDDLE EAR MUSCLE ACTIVATIONCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to European Patent Application No. 24305203.2, filed February 6, 2024, the entire contents of which are incorporated herein by reference.FIELD OF USE

[0002] The present disclosure is directed to systems and methods for detecting middle ear muscle activation (MEMA) events, and identifying one or more sleep states of an individual, e.g., a rapid eye movement (REM) sleep stage, based on the MEMA events.BACKGROUND

[0003] Polysomnography is the benchmark test for assessing an individual’s sleep and involves, for example, evaluating the stages of sleep during the night. Sleep is made up of 4 different sleep stages: N1 (sleep- wake transition), N2 (light sleep), N3 (deep sleep) and REM (REM sleep). Polysomnography is a standard but costly examination, involving a large number of parallel physiological measurements to enable sleep technicians to precisely identify the sleep stages in the recordings. Conventional examinations use 8 electroencephalography (EEG) electrodes, 2 electromyography (EMG) electrodes, 2 electro-oculography (EOG) electrodes, 1 electrocardiography (ECG) measurement, 1 or 2 reference electrodes, and a ground electrode. Overall, the set-up may be loaded onto the individual to be tested, with expert installation taking 15 to 30 minutes. This examination also may include respiratory measurements, e.g., measurement of respiratory flow by nasal cannula, plethysmography straps, etc., and / or blood oxygenation (SpO2) with a view to diagnosing sleep apnea. The central deliverable obtained at the end of a polysomnography examination is a hypnogram, as shown in FIG. 1, which consists of a chronological temporal representation of the individual’s sleep stages during the night.

[0004] One of the challenges of the last 30 to 40 years has been to take polysomnography out of the hospital and be able to obtain a reliable hypnogram of patients’ night-time sleep from a smaller, faster and easier-to-install set of measurements, potentially autonomously by the patient. Many industrial players and startups are trying to deploy such solutions, which generally may be described as sleep trackers such as smart watches (e.g., ScanWatch made available by Withings, Issy-les-Moulineaux, France; Apple Watch made available by Apple, Cupertino, California; Google Pixel Watch made available by Google, Menlo Park, California; Galaxy Watch made available by Samsung, Suwon, South Korea), rings (e.g., Oura Ring made available by Oura Health Ltd., Oulu, Finland; Circular Ring made available by Circular Paris, Ile-de-France, France), devices to be integrated into mattresses (Sleep made available by Withings, Issy-les- Moulineaux, France), lightweight EEG measurement headsets (Dreem headband made available by Beacon Biosignals, Boston, Massachusetts; Muse S headband made available by InteraXon Inc., Toronto, Ontario, Canada), etc. These sleep trackers generally manage to detect sleep reliably, but have known and criticized difficulties in reliably detecting the different sleep stages (Nl, N2, N3, REM). For example, Withings smart watches only distinguish between “light” sleep (a mixture of N2 and REM sleep) and “deep” sleep with no clarity as to whether this is N3 sleep.

[0005] In addition, solutions have been proposed to evaluate the sleep of individuals from physiological measurements made at the level of the ear, and more specifically the individual’s ear canal. For example, in-ear EEG measurement has been explored to measure sleep and other individual health data. The signal generated via in-ear EEG measurement provides a wealth of information as it contains access to brain activity (EEG), as well as ECG, EMG and horizontal eye movements (e.g., when 2 electrodes are present in both ears).

[0006] In view of the foregoing drawbacks of previously known systems and methods, there exists a need for an improved sleep tracker that more accurately identifies an individual’s sleep state, including for example, REM sleep.SUMMARY

[0007] The present disclosure overcomes the drawbacks of previously-known systems and methods by providing an apparatus for identifying a sleep state of a patient. The apparatus mayinclude an earpiece sized and shaped to be disposed adjacent an ear canal of the patient, the earpiece comprising one or more sensors configured to measure physiological information indicative of middle ear muscle activation (MEMA) events of the patient and generate one or more signals indicative of the measured physiological information, and a controller operatively coupled to the one or more sensors. The controller may have instructions that, when executed by a processor of the controller, cause the controller to: detect an occurrence of one or more MEMA events in the one or more signals received from the one or more sensors; and identify the sleep state of the patient based on the occurrence of one or more MEMA events. For example, the one or more MEMA events may comprise muscle contractions of at least one of a tensor tympani muscle, a tensor veli palatine muscle, or a stapedian muscle.

[0008] The controller may be programmed to: detect a frequency of the occurrence of one or more MEMA events; and identify the sleep state of the patient if the frequency exceeds a predetermined threshold. For example, the controller may be programmed to identify the sleep state as an REM sleep stage of the patient if the frequency exceeds a predetermined REM threshold. Additionally or alternatively, the controller may be programmed to identify the sleep state as a sleep with mentation state if the frequency exceeds a predetermined sleep with mentation state threshold. In some embodiments, the predetermined threshold may be adaptive and selected based on a basal activity of the patient.

[0009] Moreover, the one or more sensors may comprise an electrode configured to measure an electric field propagated by muscle contractions of middle ear muscles, the electric field indicative of MEMA events of the patient. Additionally or alternatively, the one or more sensors may comprise a pressure transducer configured to measure movement of an air pocket trapped within the ear canal between an eardrum of the patient and the pressure transducer, the movement of the air pocket indicative of MEMA events of the patient. Additionally or alternatively, the one or more sensors may comprise an impedancemetry sensor configured to measure reverberation of sound by a tympanic membrane of the patient, the reverberation of sound by the tympanic membrane indicative of MEMA events of the patient. Additionally or alternatively, the one or more sensors comprise an otoscope camera configured to generate video data of an eardrum of the patient, the video data indicative of MEMA events of the patient.

[0010] In addition, the controller may be programmed to: preprocess the one or more signals to isolate the one or more MEMA events in the one or more signals; generate one or more preprocessed signals comprising the one or more isolated MEMA events; and detect the occurrence of one or more MEMA events in the one or more preprocessed signals. For example, the controller may be programmed to filter the one or more signals based on one or more specific signal characteristics of the one or more signals to preprocess the one or more signals to isolate the one or more MEMA events in the one or more signals. The one or more specific signal characteristics may comprise at least one of power on a predefined frequency band, temporal pattern, phase portrait, event amplitude, overall power, energy, root mean square, or entropy. The controller further may be programmed to generate a plot of the one or more signals as a function of density of the one or more MEMA events over time to preprocess the one or more signals to isolate the one or more MEMA events in the one or more signals

[0011] In some embodiments, the system further may include one or more additional sensors configured to measure additional physiological information and generate one or more additional signals indicative of the measured additional physiological information. Accordingly, the controller may be programmed to: temporally align the one or more additional signals with the one or more signals; and use the one or more additional signals to preprocess the one or more signals and distinguish non-MEMA events to facilitate isolation of the one or more MEMA events in the one or more signals. For example, the one or more additional sensors may be configured to measure at least one of acceleration, audio, airflow, muscle atony, rapid eye movement, cardiac activity, EEG brain activity, or blood oxygen concentration. For example, the one or more additional sensors may comprise an accelerometer configured to measure acceleration and generate accelerometer data, such that the controller may be programmed to use the accelerometer data to preprocess the one or more signals and distinguish non-MEMA events to facilitate isolation of the one or more MEMA events in the one or more signals. Moreover, the measured additional physiological information may be indicative of at least one of activation of a jaw of the patient, movement of an eye of the patient, or a bruxism event. Accordingly, the controller may be programmed to use the physiological information indicative of activation of the jaw of the patient to preprocess the one or more signals and distinguish combined MEMA events from pure MEMA events to facilitate isolation of the one or more MEMA events in the one or more signals.

[0012] The controller further may be programmed to execute a machine learning algorithm to detect the occurrence of one or more of MEMA events in the one or more signals and to identify the sleep state of the patient based on the occurrence of one or more MEMA events, the machine learning algorithm trained on a temporal record of MEMA events. In addition, the controller may be programmed to generate a hypnogram illustrating a plurality of sleep states of the patient over time. Further, the controller may be programmed to identify a sleep pathology, e.g., sleep apnea or bruxism, of the patient based on the occurrence of one or more of MEMA events.

[0013] In accordance with another aspect of the present invention, a method for identifying a sleep state of a patient is provided. The method may include: measuring, via one or more sensors of an earpiece disposed adjacent an ear canal of the patient, physiological information indicative of middle ear muscle activation (MEMA) events of the patient, and generating one or more signals indicative of the measured physiological information; detecting, via a controller operatively coupled to the one or more sensors, an occurrence of one or more MEMA events in the one or more signals received from the one or more sensors; and identifying, via the controller, the sleep state of the patient based on the occurrence of one or more MEMA events. For example, detecting the occurrence of one or more MEMA events in the one or more signals may comprise detecting, via the controller, a frequency of the occurrence of one or more MEMA events, such that identifying the sleep state of the patient based on the occurrence of one or more MEMA events may comprise identifying, via the controller, the sleep state of the patient if the frequency exceeds a predetermined threshold.

[0014] The method further may include selecting the predetermined threshold based on a basal activity of the patient. Moreover, the method may include: preprocessing, via the controller, the one or more signals to isolate the one or more MEMA events in the one or more signals; and generating, via the controller, one or more preprocessed signals comprising the one or more isolated MEMA events, such that detecting the occurrence of one or more MEMA events in the one or more signals may comprise detecting, via the controller, the occurrence of one or more MEMA events in the one or more preprocessed signals. Additionally, the method may include: measuring, via one or more additional sensors operatively coupled to the controller, additional physiological information and generating one or more additional signals indicative ofthe measured additional physiological information; and temporally aligning, via the controller, the one or more additional signals with the one or more signals, such that preprocessing the one or more signals may comprise using, via the controller, the one or more additional signals to distinguish non-MEMA events to facilitate isolation of the one or more MEMA events in the one or more signals.

[0015] Detecting the occurrence of one or more MEMA events in the one or more signals and identifying the sleep state of the patient based on the occurrence of one or more MEMA events may comprise executing, via the controller, a machine learning algorithm to detect the occurrence of one or more MEMA events in the one or more signals and to identify the sleep state of the patient based on the occurrence of one or more MEMA events, the machine learning algorithm trained on a temporal record of MEMA events. In addition, the method may include generating, via the controller, a hypnogram illustrating a plurality of sleep states of the patient over time, and / or identifying, via the controller, a sleep pathology of the patient based on the occurrence of one or more MEMA events.BRIEF DESCRIPTION OF THE DRAWINGS

[0016] FIG. 1 illustrates a conventional hypnogram.

[0017] FIG. 2 illustrates an exemplary sleep state determination system in accordance with the principles of the present disclosure.

[0018] FIG. 3 shows some example components that may be included in a sleep state determination platform in accordance with the principles of the present disclosure.

[0019] FIG. 4 is a graph illustrating various sensor signals over time including a MEMA event in accordance with the principles of the present disclosure.

[0020] FIG. 5 is a graph illustrating a density function of muscle activations of the middle ear overlaid with time spent in REM sleep.DETAILED DESCRIPTION

[0021] It is known in the scientific literature that the muscles of the middle ear (the tensor tympani, the tensor veli palatini and / or the stapedian muscle), which support the chain of the middle ear bones between the eardrum and the cochlea, activate during REM sleep. See e.g., Pessah, M A, and H P Roffwarg. “Spontaneous Middle Ear Muscle Activity in Man: A Rapid Eye Movement Sleep Phenomenon” Science (New York, N.Y.) vol. 178,4062 (1972): 773-6. doi:10.1126 / science.l78.4062.773; see also, Benson, K, and V P Zarcone Jr. “Phasic Events of REM Sleep: Phenomenology of Middle Ear Muscle Activity and Periorbital Integrated Potentials in the Same Normal Population” Sleep vol. 2,2 (1979): 199-213. doi:10.1093 / sleep / 2.2.199; Duane E. Siegel et al., “Middle-Ear Muscle Activity (MEMA) and Its Association with Motor Activity in the Extremities and Head in Sleep” Sleep, vol. 14,5 (1991): 454-459. doi: 10.1093 / sleep / l 4,5,454. However, it has not been proposed to automatically detect an individual’s sleep state, e.g., REM sleep, from a reading of activations of the individual’s middle ear muscles.

[0022] Accordingly, disclosed are systems and methods for detecting the occurrence of middle ear muscle activation (MEMA) events in physiological signals obtained from one or more wearable sensors, e.g., in-ear sensors, and identifying one or more sleep states of a patient / individual based on the detected MEMA events. In addition to identifying sleep according to sleep stages, one also may distinguish, in parallel, the sleep states where the individual is experiencing mental content (e.g., oneiric activity, involving internally generated sensory content like visual, tactile or auditory content) or not, designated as “sleep with mentation state” and “sleep without mentation state.” Indeed, until recently, REM sleep was thought to be the only sleep stage in which mentation during sleep could occur but recent findings, notably by D. Oudiette et al., “Dreaming without REM sleep” Consciousness and Cognition, vol. 21,3 (2012): 1129-1140, has shown that generation of mental activity during sleep may be independent of sleep stage. Accordingly, as defined herein, the sleep state of the individual may be any one of the four sleep stages, e.g., Nl, N2, N3, REM sleep, or an awake state, as well as “sleep with mentation state” and “sleep without mentation state.”

[0023] Referring now to FIG. 2, an exemplary system for determining an individual’s sleep state is provided. System 100 may include wearable 200, e.g., an earplug sized and shaped to be disposed adjacent an individual’s ear canal, having one or more sensors 202 configured tomeasure physiological information indicative of activations of the individual’s middle ear muscles, device 120, e.g., a mobile device such as a smart phone, tablet, a smart watch, and / or a computer, having a graphical user interface for displaying information and / or receiving user input, and sleep state determination platform 300. In addition, system 100 optionally may include one or more additional sensors 210 configured to measure additional physiological information, as described in further detail below, e.g., during the same sleep session of the individual sensed by sensors 202.

[0024] Wearable 200 may be worn by an individual such that at least a portion of wearable 200 is disposed adjacent the individual’s ear canal, e.g., at least partially within and / or around the ear canal. For example, in some embodiments, wearable 200 may comprise a viscoelastic material, e.g., foam, configured to transition between a compressed state and an expanded state where wearable 200 conforms to the anatomy of the individual’s ear where wearable 200 is located, such that wearable 200 may be comfortably worn by the individual. Additionally, wearable 200 may include a power source, e.g., a rechargeable battery, configured to supply power to sensor 202, and communication circuitry for transmitting and receiving data between sensor 202 and the other electronic components of system 100, e.g., sleep state determination platform 300. Wearable 200 may include memory for storing data generated by sensor 202, and further for transmission to the other electronic components of system 100.

[0025] In a preferred embodiment, sensor 202 may include one or more electrodes configured to measure an electric and / or magnetic field propagated by muscle contractions of middle ear muscles, and to generate signals indicative of the electric and / or magnetic field such that MEMA events, if any, are observable from readings of the signals. The electrode may be a piece of conductor material (e.g., metal or conductive polymer) sized and shaped to non- invasively contact the individual’s skin, e.g., within the ear canal, to measure the electrical potential of a biological activity, e.g., micro-movements of the skin due to muscle contractions of middle ear muscles. For example, the electrode may measure EMG from middle ear muscles by filtering frequencies of the electrical signal above, e.g., 40 Hz, between 40-80 Hz, or preferably above 65 Hz. As will be understood by a person having ordinary skill in the art, EMG requires a higher sample rate compared to, for example, EEG. Accordingly, sensor 202 may be configured to generate signals comprising a high sample rate sufficient for EMG measurement.

[0026] Additionally or alternatively, sensor 202 may include one or more pressure transducers configured to measure movement of an air pocket trapped within the ear canal between an eardrum of the individual and the pressure transducer, and to generate signals indicative of the measured air pocket movement such that MEMA events, if any, are observable from readings of the signals. For example, the pressure transducer may be constructed similar to that described in D. Siegel et al. “An Inexpensive Alternative for Recording Middle Ear Muscle Activity (MEMA) During Sleep” Sleep vol. 15,6 (1992): 567-70. doi: 10.1093 / sleep / 15.6.567. In some embodiments, sensor 202 may include one or more impedancemetry sensors configured to measure reverberation of sound by a tympanic membrane of the individual, and to generate signals indicative of the measured sound reverberation such that MEMA events, if any, are observable from readings of the signals, as described in the Pessah article. Additionally, or alternatively, sensor 202 may include one or more otoscope cameras configured to generate video data of an eardrum of the individual, such that MEMA events, if any, are observable from readings of the video data.

[0027] As described above, in some embodiments, system 100 may include one or more additional sensors 210 configured to measure additional physiological information and generate one or more signals indicative of the measured additional physiological information, e.g., during the same sleep session of the individual sensed by sensors 202, to facilitate removal of artifacts to isolate combined and / or pure MEMA events from the signals generated by sensor 202, as described in further detail below. An artifact is a non-MEMA event that may be a signature on the sensor signals that looks like a MEMA event, but that is actually random noise, e.g., non- negligible movement of the individual during sleep. As defined herein, MEMA events may include combined MEMA events, e.g., activations of the middle ear muscles with simultaneous activations of the individual’s jaw muscles (e.g., masseter, suprahyoid, pterygoidian, etc.) as measured by sensor 210, as well as pure MEMA events, e.g., only activations of the middle ear muscles without activations of the individual’s jaw muscles. In accordance with the principles of the present disclosure, system 100 may be used to determine an individual’s sleep state by observing combined and / or pure MEMA events.

[0028] Sensors 210 may be disposed on wearable 200 and / or on other parts of the individual’s body, and may be configured to measure at least one of acceleration, audio, airflow,muscle atony, rapid eye movement, cardiac activity, or blood oxygen concentration, which may be indicative of, e.g., activation of a jaw of the patient, movement of an eye of the patient, movement of the entire body, a bruxism event, etc. For example, sensor 210 may be an accelerometer configured to measure acceleration and generate accelerometer data. The accelerometer may be disposed on wearable 200 to measure acceleration at the level of the individual’s ear canal, and / or on, e.g., the individual’s jaw to measure acceleration at the level of the individual’s jaw, such that the norm of the 3D acceleration vector may be used to quantify quantity of movement of the individual (e.g., noise).

[0029] Moreover, sensors 210 may be one or more electrodes configured to measure at least one of EEG, ECG, EMG, or EOG, depending on the location of the individual’s body the respective electrode is placed, as well as the filtering settings applied to the electrode signals for measurement. For example, an electrode placed on the individual’s scalp and filtered between, e.g., 0.1 to 70 Hz, may measure the individual’s brain activity (e.g., an EEG electrode), and an electrode placed on the individual’s jaw, e.g., in front of the masseter muscle, and filtered to only keep frequencies above, e.g., 40 to 80 Hz, or preferably above 65 Hz, may measure EMG from the masseter muscles of the individual’s jaw. In addition, sensor 210 may be a movementsensitive strain gauge having a tube, e.g., a Tygon tube, that is inserted into an ear mold configured to be at least partially disposed in the individual’s ear canal, and which may be slightly displaced in the ear canal by a variety of actions (e.g. facial twitches, movement of the jaw, movement of the Tygon tube). Movement of the ear mold may cause a change in pressure, electric field, and / or image which, without the strain gauge activation, may be mistaken for an endogenous MEMA event, as described in further detail below.

[0030] Device 120 may have an application installed thereon for interfacing with wearable 200 and / or sleep state determination platform 300. For example, the application may permit device 120 to transmit user input to wearable 200 and / or sleep state determination platform 300, e.g., start and stop commands, commands to pair device 120 with wearable 200, etc., and to receive data from sleep state determination platform 300 for display to a user, e.g., a hypnogram graphically illustrating the individual’s sleep states during a sleep session, and / or for generating an alarm, e.g., an audible alarm, if the user is determined to be in a sleep stage close to the awake state. Sleep state determination platform 300 may be located on one or more servers, e.g., storedon device 120, on cloud 160, or on wearable 200, and may communicate with sensor 202 and device 120 via network 150.

[0031] Network 150 may include any one, or a combination of networks, such as a local area network (LAN), a wide area network (WAN), a telephone network, a cellular network, a cable network, a wireless network, and / or private / public networks, such as the Internet. For example, network 150 may support communication technologies, such as TCP / IP, Bluetooth, cellular, near-field communication (NFC), Wi-Fi, Wi-Fi direct, machine-to-machine communication, and / or man-to-machine communication. Information shared between sleep state determination platform 300, wearable 200, and / or device 120 may be stored on cloud storage 160 and may be bi-directional in nature. For example, in one case, information may be transferred from sleep state determination platform 300 to cloud storage 160. Such information stored on cloud storage 160 may be accessed and downloaded by device 120, or other devices, e.g., a remote computing device.

[0032] Referring now to FIG. 3, components that may be included in a sleep state determination platform 300 are described in further detail. Sleep state determination platform 300 may include one or more processors 302, communication system 304, and memory 306. Communication system 304 may include a wireless transceiver that allows sleep state determination platform 300 to communicate with wearable 200, cloud storage 160, and / or device 120. The wireless transceiver may use any of various communication formats, such as, for example, an Internet communications format, or a cellular communications format.

[0033] Memory 306, which is one example of a non-transitory computer- readable medium, may be used to store operating system (OS) 320, device interface module 308, sensor interface module 310, sensor data processing module 312, sleep state determination module 314, hypnogram generation module 316, and / or sleep pathology determination module 318. The modules are provided in the form of computer-executable instructions that may be executed by processor 302 for performing various operations (inline or offline) in accordance with the disclosure. For example, processor 302 may wait until it receives the entire signal from the one or more sensors after a sleep sensing session (e.g., an entire night) prior to processing, oralternatively, processor 302 may process the signal in real-time as the signal is received during the sleep sensing session.

[0034] Device interface module 310 may be executed by processor 302 for interfacing with an application installed on device 120, e.g., a mobile application installed on a smart phone. For example, device interface module 308 may determine if device 120 is paired with sleep state determination platform 300, e.g., via communication system 304, and further may transmit data to device 120 when paired. Moreover, device interface module 308 may pause the transmission of data to device 120 when it is determined that device 120 is outside a predetermined range from wearable 200, and may cause the remaining data to be stored, e.g., either on memory 306 or on cloud storage 160, until device 120 is determined to be within the predetermined range. In some embodiments, device interface module 308 may automatically transmit data to device 120 when paired, or alternatively, device interface module 308 may transmit a push notification to device 120 to download the data. For example, device interface module 308 may provide feedback information to device 120 for display to the user including, for example, information regarding the user’s quality of sleep as well as recommendations to, e.g., sleep earlier, sleep more, increase or decrease activity during the day, etc. In addition, device interface module 308 may interface with device 120 to receive the one or more signals from device 120 indicative of a command by the user to, e.g., start or stop a sleep sensing session, and / or pair device 120 with sleep state determination platform 300, and further to transmit data to device 120 including, for example, a hypnogram as described above, an alert when the power source level of wearable 200 is low and / or depleted, and / or if the user is determined to be in a sleep stage close to awake, etc.

[0035] Sensor interface module 310 may be executed by processor 302 for interfacing with sensors 202 of wearable 200. For example, sensor interface module 310 may cause wearable 200 to initiate a sleep sensing session responsive to a start command signal received from device 120, such that sensors 202 begins to sense measurements and generate signals indicative of the sensed measurements, and / or terminate a sleep sensing session responsive to a stop command signal received from device 120, such that sensors 202 stop sensing measurements. In some embodiments, sensors 202 may continuously sense measurements and generate data as long as the power source of wearable 200 is not depleted and wearable 200 is powered on. Additionally, sensor interface module 310 may receive data generated by sensors 202, which may be stored onmemory 306. Moreover, sensor interface module 310 may interface with additional sensors 210, and receive data generated by additional sensors 210. As described above, sensor interface module 310 may receive data generated by sensors 202, 210 for inline and / or online processing.

[0036] Sensor data processing module 312 may be executed by processor 302 for processing data received by sensor interface module 310 from sensors 202 to automatically identify MEMA events in the signals generated by sensors 202, and optionally, for processing data received by sensor interface module 310 from additional sensors 210 to remove artifacts and isolate MEMA events in the signals received from sensors 202. Particularly, physiological events (e.g., activation of the jaws, bruxism, movements, etc.) observed in the signals from sensors 202 may be mistaken as MEMA events. To automatically identify MEMA events in the signals from sensors 202, sensor data processing module 312 may filter the signals based on one or more specific signal characteristics, e.g., signal strength on a predefined frequency band, temporal pattern, phase portrait, event amplitude, power, or energy, and perform temporal pattern recognition. For example, sensor interface module 310 may divide the signal generated by sensors 202 during a sleep sensing session into 1 second bits, e.g., an epoch, and calculate an indicator such as the maximum amplitude of the signal or the power of the epoch, which may be calculated by summing the squares of each of the values acquired during a 1 second epoch.

[0037] MEMA events are generally wide in amplitude and short in duration, and further may have a specific time-frequency pattern; whereas, EMG activity from the jaw, which may have a signature in the same frequency band as the MEMA events, are generally longer in duration and have a more “phasic” temporal pattern, e.g., several bursts in a row. Moreover, EMG signals are generally less elevated during REM sleep than during other sleep stages (e.g., N2, N3, awake). Accordingly, when the amplitude or the power is above a predetermined value threshold, sensor interface module 310 may classify that event as a MEMA event or a candidate MEMA event, as described in further detail below. In some embodiments, the predetermined value threshold may be adaptive on a case-by-case basis depending on the surrounding basal activity, e.g., the average density of MEMA event over the entire sleep sensing session, such that the predetermined value threshold may be patient-specific. For example, the predetermined value threshold may be set to a higher value for individual’s with a condition such as bruxism where the surrounding basal activity is generally higher. In some embodiments, sensor data processing module 312 mayexecute a machine learning algorithm to identify the MEMA events in the signal from sensors 202 via temporal pattern recognition, the machine learning algorithm trained with a dataset of measured physiological information comprising one or more MEMA events.

[0038] Moreover, to remove artifacts and isolate MEMA events in the signals received from sensors 202, sensor data processing module 312 may process the signals from sensors 202 in conjunction with the signals from additional sensors 210, so as to not confuse MEMA events with other physiological events (e.g., activation of the jaws, bruxism, movements, etc.). By taking into account additional physiological information measured by additional sensors 210, actual contractions of the middle ear muscles may be identified with higher reliability. As described above, sensors 210 may be activated to measure additional physiological data during the same sleep sensing session that is sensed by sensors 202, such that both sensors 202 and sensors 210 generate data for the same sleeping sensing session. For example, sensors 210 may be activated to measure physiological information of the individual synchronously with sensors 202. Additionally, sensor data processing module 312 may temporally align the signals generated by sensors 210 with the signals generated by sensors 202, e.g., if sensors 210 is not activated synchronously with sensors 202.

[0039] By comparing temporally aligned signals from sensors 202 and additional sensors 210, sensor data processing module 312 may determine if a signature of the signal from sensors 202, which may have been identified as a candidate MEMA event as described above, is an artifact or not. For example, if signals received from sensors 210, e.g., an accelerometer and / or a strain gauge, indicate the synchronous occurrence of non-negligible activity with a candidate MEMA event in the signal received from sensors 202, sensor data processing module 312 may determine that the candidate MEMA event is an artifact, and accordingly, remove / disregard the artifact from the signal received from sensors 202. Conversely, if the signals received from sensors 210 do not indicate non-negligible activity at the same time as the occurrence of a candidate MEMA event in the signal received from sensors 202, sensor data processing module 312 may determine that the candidate MEMA event is a MEMA event, e.g., an actual contraction of the individual’s middle ear muscles.

[0040] As another example, if the EMG signal received from sensor 210, e.g., a movementsensitive strain gauge having a tubing and ear mold at least partially disposed in the individual’s ear canal, indicates the concomitant occurrence of movement of the ear mold (e.g., due to facial twitches, movement of the jaw, movement of the Tygon tube, etc.) with the identified MEMA event, sensor data processing module 312 may determine that the candidate MEMA event is an artifact. Conversely, if the EMG signal received from sensor 210 does not indicate that a movement of the ear mold occurred at the same time as the candidate MEMA event, sensor data processing module 312 may determine that the candidate MEMA event is a MEMA event. In accordance with the principles of the present disclosure, additional physiological information measured by one or more additional sensors, e.g., muscle atony measured by very low-amplitude EMG readings at the chin / jaw level, rapid eye movements measured by EOG, in the EEG signals having different frequency distributions (fewer low frequencies, more high frequencies), patterns in the EEG signal such as “saw tooth” theta waves, and / or amplified cardiac, respiratory, and oxygenation variability, may be used to identify artifacts and facilitate isolation of MEMA events in the signals from sensors 202. As these markers are not specific to REM sleep per se, when used in conjunction with physiological information measured by sensors 202, they may facilitate accurate classification of REM sleep so as to not be confused with, e.g., light or deep sleep. For example, muscle atony also occurs during deep sleep, and rapid eye movements only occur for 50% of the time during REM sleep.

[0041] As described above, a combined MEMA event includes an activation of the individual’s middle ear muscles with simultaneous activation of the individual’s jaw muscles. Accordingly, to determine whether an identified MEMA event is a combined MEMA event or a pure MEMA event, if the EMG signal received from sensor 210, e.g., an electrode disposed on the individual’s jaw, indicates the synchronous occurrence of activation of the individual’s jaw muscles with the identified MEMA event, sensor data processing module 312 may determine that the MEMA event is a combined MEMA event. Conversely, if the EMG signal received from sensor 210 does not indicate that an activation of the jaw muscles occurred at the same time as the identified MEMA event (or if the activation is negligible, e.g., falls below a predetermined threshold), sensor data processing module 312 may determine that the identified MEMA event is a pure MEMA event. As described above, system 100 may determine an individual’s sleep statebased on MEMA events identified in the signals received from sensors 202 without distinguishing between combined MEMA events and pure MEMA events.

[0042] FIG. 4 is a graph illustrating a 30 second portion of sensor signals measured via various sensors disposed on an individual overnight during a sleep sensing session. El, E2, C3, and C4 are channels of EEG activity of the individual, R Suprahyiod and R Masseter are channels of EMG activity of the individual’s jaw muscles, R Middle Ear is a channel of a signal generated by an in-ear sensor, e.g., a pressure transducer configured to measure movement of an air pocket trapped within the ear canal between an eardrum of the individual and the pressure transducer, Thorax and Abdomen are channels of biosignals indicative of the individual’s breathing activity, ECG is a channel of ECG activity of the individual’s heart, and Volume Audio is a channel of sound in the room where the individual was sleeping. For example, the pressure transducer was constructed similar to that described in the 1991 Siegel article.

[0043] As shown in FIG. 4, the signal of the R Middle Ear channel appears to include some kind of weak basal activity; however, by comparing the R Middle Ear channel signal with the ECG channel signal, the basal activity is observed to be synchronous with the heart ECG, and therefore may be disregarded as a non-MEMA event. In contrast, the power of the signal of the R Middle Ear channel within box 400 exceeds a predetermined threshold, and is thereby classified as a MEMA event. Moreover, the MEMA event further may be classified as a pure MEMA event as neither the R Suprahyiod nor the R Masseter EMG channels indicate activation of the jaw muscles at the same time as the MEMA event. In addition, upon identification of one or more MEMA events in the signal from sensors 202, sensor data processing module 312 further may generate a plot of the signal as a function of density of the one or more MEMA events over time, e.g., the duration of the sleep sensing session, as shown in FIG. 5 described in further detail below. For example, the density function of MEMA events may be plotted as the number of MEMA events per minute. Moreover, sensor data processing module 312 further may distinguish whether the detected MEMA events are unilateral (e.g., occurring in only one ear of the individual, which only occurs about 15% of the time) or bilateral (e.g., occurring in both ears of the individual), such as when the individual uses two wearables 200 (one in each ear).

[0044] Sleep state determination module 314 may be executed by processor 302 for automatically predicting the individual’s sleep state, e.g., Nl, N2, N3, or REM sleep stages, an awake state, or sleep with or without mentation states, based on the MEMA events identified in the signal from sensors 202 by sensor data processing module 312, e.g., based on the density plot of MEMA events generated by sensor data processing module 312. For example, sleep state determination module 314 may determine that the individual is in REM sleep if the frequency of the occurrence of MEMA events per unit of time exceeds a predetermined frequency threshold, as shown in FIG. 5. FIG. 5 is a graph illustrating a density function of MEMA events over time, e.g., number of MEMA events in a predefined unit of time, overlaid with time spent in REM sleep. For example, the graph may be a plot of the moving average of the number of MEMA events over an average of, e.g., ten seconds, one minute, etc. As shown in FIG. 5, the time during which the individual was in REM sleep is accurately predicted according to precise signal matching with the identified MEMA events.

[0045] Hypnogram generation module 316 may be executed by processor 302 for generating a hypnogram illustrating the time spent by the individual in one or more sleep states, e.g., Nl, N2, N3, REM, and / or awake, for the duration of the sleep sensing session. Hypnogram generation module 316 may transmit the generated hypnogram to device 120 for display to a user.

[0046] Sleep pathology determination module 318 may be executed by processor 302 for determining if the individual has one or more sleep pathologies based on the characteristics of the occurrence of MEMA events in the signal from sensors 202. In individuals with a sleep pathology, such as apnea or bruxism (e.g., phasic jaw clenching), the characteristics of the recording of middle ear activations during the sleep sensing session are likely to be different than in individuals without the sleep pathology, e.g., more MEMA events throughout the sleep sensing session. For example, phasic bruxism, tonic bruxism, and mixt bruxism events are all longer in duration than MEMA events. Accordingly, by distinguishing between combined and pure MEMA events, sleep pathology determination module 318 may determine that the individual has bruxism if the ratio of the number of combined MEMA events over the number of total MEMA events (e.g., combined and pure MEMA events) exceeds a predetermined threshold,in addition to the total number of MEMA events being abnormally elevated, e.g., above a predetermined total event threshold.

[0047] While various illustrative embodiments of the invention are described above, it will be apparent to one skilled in the art that various changes and modifications may be made therein without departing from the invention. The appended claims are intended to cover all such changes and modifications that fall within the true scope of the invention.

Claims

WHAT IS CLAIMED:

1. An apparatus for identifying a sleep state of a patient, the apparatus comprising: an earpiece sized and shaped to be disposed adjacent an ear canal of the patient, the earpiece comprising one or more sensors configured to measure physiological information indicative of middle ear muscle activation (MEMA) events of the patient and generate one or more signals indicative of the measured physiological information; and a controller operatively coupled to the one or more sensors, the controller having instructions that, when executed by a processor of the controller, cause the controller to: detect an occurrence of one or more MEMA events in the one or more signals received from the one or more sensors; and identify the sleep state of the patient based on the occurrence of one or more MEMA events.

2. The apparatus of claim 1 , wherein the one or more MEMA events comprise muscle contractions of at least one of a tensor tympani muscle, a tensor veli palatine muscle, or a stapedian muscle.

3. The apparatus of claim 1, wherein the controller is programmed to: detect a frequency of the occurrence of one or more MEMA events; and identify the sleep state of the patient if the frequency exceeds a predetermined threshold.

4. The apparatus of claim 3, wherein the controller is programmed to identify the sleep state as an REM sleep stage of the patient if the frequency exceeds a predetermined REM threshold.

5. The apparatus of claim 3, wherein the predetermined threshold is adaptive and selected based on a basal activity of the patient.

6. The apparatus of claim 3, wherein the controller is programmed to identify the sleep state as a sleep with mentation state if the frequency exceeds a predetermined sleep with mentation state threshold7. The apparatus of claim 1 , wherein the one or more sensors comprise an electrode configured to measure an electric field propagated by muscle contractions of middle ear muscles, the electric field indicative of MEMA events of the patient.

8. The apparatus of claim 1, wherein the one or more sensors comprise a pressure transducer configured to measure movement of an air pocket trapped within the ear canal between an eardrum of the patient and the pressure transducer, the movement of the air pocket indicative of MEMA events of the patient.

9. The apparatus of claim 1 , wherein the one or more sensors comprise an impedancemetry sensor configured to measure reverberation of sound by a tympanic membrane of the patient, the reverberation of sound by the tympanic membrane indicative of MEMA events of the patient.

10. The apparatus of claim 1, wherein the one or more sensors comprise an otoscope camera configured to generate video data of an eardrum of the patient, the video data indicative of MEMA events of the patient.

11. The apparatus of claim 1 , wherein the controller is programmed to: preprocess the one or more signals to isolate the one or more MEMA events in the one or more signals; generate one or more preprocessed signals comprising the one or more isolated MEMA events; and detect the occurrence of one or more MEMA events in the one or more preprocessed signals.

12. The apparatus of claim 11, wherein the controller is programmed to filter the one or more signals based on one or more specific signal characteristics of the one or more signals to preprocess the one or more signals to isolate the one or more MEMA events in the one or more signals.

13. The apparatus of claim 12, wherein the one or more specific signal characteristics comprise at least one of power on a predefined frequency band, temporal pattern, phase portrait, event amplitude, overall power, energy, root mean square, or entropy.

14. The apparatus of claim 11, further comprising: one or more additional sensors configured to measure additional physiological information and generate one or more additional signals indicative of the measured additional physiological information, wherein the controller is programmed to: temporally align the one or more additional signals with the one or more signals; and use the one or more additional signals to preprocess the one or more signals and distinguish non-MEMA events to facilitate isolation of the one or more MEMA events in the one or more signals.

15. The apparatus of claim 14, wherein the one or more additional sensors are configured to measure at least one of acceleration, audio, airflow, muscle atony, rapid eye movement, cardiac activity, EEG brain activity, or blood oxygen concentration.

16. The apparatus of claim 15, wherein the one or more additional sensors comprise an accelerometer configured to measure acceleration and generate accelerometer data, and wherein the controller is programmed to use the accelerometer data to preprocess the one or more signals and distinguish non-MEMA events to facilitate isolation of the one or more MEMA events in the one or more signals.

17. The apparatus of claim 14, wherein the measured additional physiological information is indicative of at least one of activation of a jaw of the patient, movement of an eye of the patient, or a bruxism event.

18. The apparatus of claim 17, wherein the controller is programmed to use the physiological information indicative of activation of the jaw of the patient to preprocess the oneor more signals and distinguish combined MEMA events from pure MEMA events to facilitate isolation of the one or more MEMA events in the one or more signals.

19. The apparatus of claim 1, wherein the controller is programmed to execute a machine learning algorithm to detect the occurrence of one or more of MEMA events in the one or more signals and to identify the sleep state of the patient based on the occurrence of one or more MEMA events, the machine learning algorithm trained on a temporal record of MEMA events.

20. The apparatus of claim 1 , wherein the controller is programmed to generate a hypnogram illustrating a plurality of sleep states of the patient over time.

21. The apparatus of claim 1, wherein the controller is programmed to identify a sleep pathology of the patient based on the occurrence of one or more of MEMA events.

22. The apparatus of claim 21, wherein the sleep pathology comprises sleep apnea or bruxism.

23. A method for identifying a sleep state of a patient, the method comprising: measuring, via one or more sensors of an earpiece disposed adjacent an ear canal of the patient, physiological information indicative of middle ear muscle activation (MEMA) events of the patient, and generating one or more signals indicative of the measured physiological information; detecting, via a controller operatively coupled to the one or more sensors, an occurrence of one or more MEMA events in the one or more signals received from the one or more sensors; and identifying, via the controller, the sleep state of the patient based on the occurrence of one or more MEMA events.

24. The method of claim 23, wherein detecting the occurrence of one or more MEMA events in the one or more signals comprises detecting, via the controller, a frequency of the occurrence of one or more MEMA events, and wherein identifying the sleep state of the patient based on the occurrence of one or more MEMA events comprises identifying, via the controller, the sleep state of the patient if the frequency exceeds a predetermined threshold.

25. The method of claim 24, further comprising selecting the predetermined threshold based on a basal activity of the patient.

26. The method of claim 24, wherein identifying the sleep state of the patient based on the occurrence of one or more MEMA events comprises identifying, via the controller, the sleep state as an REM sleep stage of the patient if the frequency exceeds the predetermined threshold.

27. The method of claim 23, wherein measuring physiological information indicative of MEMA events comprises measuring, via an electrode, an electric field propagated by muscle contractions of middle ear muscles, the electric field indicative of MEMA events of the patient.

28. The method of claim 23, further comprising: preprocessing, via the controller, the one or more signals to isolate the one or more MEMA events in the one or more signals; and generating, via the controller, one or more preprocessed signals comprising the one or more isolated MEMA events, wherein detecting the occurrence of one or more MEMA events in the one or more signals comprising detecting, via the controller, the occurrence of one or more MEMA events in the one or more preprocessed signals.

29. The method of claim 28, wherein preprocessing the one or more signals comprises filtering, via the controller, the one or more signals based on one or more specificsignal characteristics of the one or more signals to isolate the one or more MEMA events in the one or more signals.

30. The method of claim 28, further comprising: measuring, via one or more additional sensors operatively coupled to the controller, additional physiological information and generating one or more additional signals indicative of the measured additional physiological information; and temporally aligning, via the controller, the one or more additional signals with the one or more signals, wherein preprocessing the one or more signals comprises using, via the controller, the one or more additional signals to distinguish non-MEMA events to facilitate isolation of the one or more MEMA events in the one or more signals.

31. The method of claim 23, wherein detecting the occurrence of one or more MEMA events in the one or more signals and identifying the sleep state of the patient based on the occurrence of one or more MEMA events comprise executing, via the controller, a machine learning algorithm to detect the occurrence of one or more MEMA events in the one or more signals and to identify the sleep state of the patient based on the occurrence of one or more MEMA events, the machine learning algorithm trained on a temporal record of MEMA events.

32. The method of claim 23, further comprising generating, via the controller, a hypnogram illustrating a plurality of sleep states of the patient over time.

33. The method of claim 23, further comprising identifying, via the controller, a sleep pathology of the patient based on the occurrence of one or more MEMA events.

Images