Method, apparatus, and computer programs for sleep enhancement and diagnostics using electroencephalogram monitoring

Abstract

A method includes identifying electroencephalogram (EEG) sensor data obtained by a wearable device associated with a user's head and providing the EEG sensor data as a first data input to a first trained machine learning model. A first output associated with sleep quality data associated with the user is obtained from the first trained machine learning model. A second input including the sleep quality data and user input data is provided to a second trained machine learning model. A second output associated with sleep therapy data is obtained from the second trained machine learning model. Subsequent to obtaining the second output, subsequent EEG sensor data and / or subsequent user input data is identified, and the first and / or second trained machine learning model is re-trained to generate updated sleep therapy data based on the subsequent EEG sensor data and / or subsequent user input data meeting a first threshold value.

Description

Attorney Docket No.: 46528.6 (L0002PCT)METHOD, APPARATUS, AND COMPUTER PROGRAMS FOR SLEEP ENHANCEMENT AND DIAGNOSTICS USING ELECTROENCEPHALOGRAM MONITORINGTECHNICAL FIELD

[0001] The present disclosure pertains to the field of computer software and technologies for healthcare applications, in particular, for sleep and brain health.BACKGROUND

[0002] Sleep is a state of reduced mental and physical activity in which consciousness is altered and certain sensory activity is inhibited. During sleep, most of the body’s systems are in an anabolic state, helping restore the body’s systems which helps maintain mood, memory, cognitive function, etc. Sleep disorders may affect a body’s systems.BRIEF DESCRIPTION OF DRAWINGS

[0003] The present disclosure is illustrated by way of example, and not by way of limitation, in the figures of the accompanying drawings in which like references indicate similar elements. It should be noted that different references to “an” or “one” embodiment in this disclosure are not necessarily to the same embodiment, and such references mean at least one.

[0004] FIGS. 1 A-B are block diagrams illustrating the overall components and architecture of a sleep monitoring and assessment system, according to certain embodiments.

[0005] FIG. 2A is a block diagram illustrating a wearable device, according to certain embodiments.

[0006] FIGS. 2B-E illustrate various perspectives of the wearable device, according to certain embodiments.

[0007] FIG. 3 is a block diagram illustrating a methodology of a first trained machine learning model (e.g., a sensor monitoring algorithm), according to certain embodiments.

[0008] FIG. 4 is a block diagram illustrating a methodology of a second trained machine learning model (e.g., a therapy algorithm), according to certain embodiments.

[0009] FIGS. 5 A-B are block diagrams illustrating the details of a processing device executing control logic (e.g., a patient facing computer program, a clinician facing computer program), according to certain embodiments.

[0010] FIGS. 6A-B illustrate graphical user interfaces (GUIs) associated with receiving data input from diaries and / or tests, according to certain embodiments.Attorney Docket No.: 46528.6 (L0002PCT)

[0011] FIG. 7 illustrates a GUI associated with establishing a connection between the client device and a biometric sensor (e.g., a wearable communication system), according to certain embodiments.

[0012] FIGS. 8A-B illustrate GUIs associated with providing an output (e.g., sleep therapy data, data output), according to certain embodiments.

[0013] FIG. 9 illustrates a GUI associated with providing output (e.g., reports, graphs) to the second processing device, according to certain embodiments.

[0014] FIGS. 10A-D are flow diagrams of methods associated with sleep enhancement and diagnostics, according to certain embodiments.

[0015] FIG. 11 is a block diagram illustrating a computer system, according to certain embodiments.DETAILED DESCRIPTION

[0016] The present disclosure is directed to electroencephalogram sleep monitoring and sleep improvement systems (e.g., method, apparatus, and computer programs for electroencephalogram sleep monitoring and sleep improvement made thereof; method, apparatus, and computer programs for sleep enhancement and diagnostics using electroencephalogram monitoring).

[0017] Sleep is a state of reduced mental and physical activity in which consciousness is altered and certain sensory activity is inhibited. Sleep occurs in repeating periods, during which the body alternates between different modes (e.g., rapid eye movement (REM) and non-REM (NREM) sleep). During sleep, most of the body’s systems are in an anabolic state, helping restore the body’s systems (e.g., immune, nervous, skeletal, muscular, etc. systems) which helps maintain mood, memory, cognitive function, etc. Sleep disorders may include dyssomnias (e.g., insomnia, hypersomnia, narcolepsy, sleep apnea, etc.), parasomnias (e.g., sleepwalking, rapid eye movement sleep behavior disorder), bruxism, circadian rhythm sleep disorders, etc., and may affect a body’s systems. Because many sleep disorders stem from disrupted sleep patterns that are difficult to detect subjectively, tracking sleep offers a way to objectively monitor and analyze sleep behavior, enabling a more accurate diagnosis and personalized treatment strategies.

[0018] Sleep tracking is the process of monitoring and recording physiological and / or behavioral data (e.g., inactivity, movement, heart rate, brain activity, etc.) during sleep to evaluate sleep patterns, quality, and stages over time. Such sleep tracking may be used to diagnose sleep disorders, mental disorders, psychotic disorders, etc. Sleep disorders may alsoAttorney Docket No.: 46528.6 (L0002PCT) be diagnosed through sleep studies. A sleep study may be a test that records the activity of the body during sleep to test for different sleep characteristics and disorders. Sleep studies may include simple sleep studies, polysomnography, multiple sleep latency tests (MSLTs), maintenance of wakefulness tests (MWTs), and home sleep tests (HSTs). In medicine, sleep studies can be useful to identify and rule out various sleep disorders. Sleep studies can also be valuable to psychology, to provide insight into brain activity and other physiological factors of both sleep disorders and normal sleep.

[0019] Recent studies indicate that insomnia affects about 30% of adults worldwide, with approximately 10-15% experiencing chronic insomnia. A global survey showed that 35% of respondents reported sleep disturbances, particularly heightened during the COVID-19 pandemic. Another sleep disorder is obstructive sleep apnea (OSA), which has a reported prevalence as high as 34% in men and 17% in women. Recent estimates suggest that around 1 billion people worldwide may be affected by OSA, with many cases remaining undiagnosed. When left undiagnosed, OSA can lead to serious health consequences (e.g., chronic fatigue, impaired cognitive function, cardiovascular disease, etc.) due to repeated disruptions to breathing and sleep without the individual being aware such disruptions are occurring.

[0020] Sleep disorders and associated complaints have increased in the population and are often a consequence of modern lifestyles. Along with modern lifestyles comes the desire to find answers within technology. As technology advances, the search for wearable devices and mobile apps that monitor sleep has grown about 18.5%. However, there is a general lack of scientific evidence indicating the validity and / or accuracy of conventional wearable devices and mobile apps when compared with “gold standard” methods (e.g., polysomnography, electroencephalogram readings, etc.). These conventional devices and apps also lack certifications by regulatory entities (e.g., the Food and Drug Administration (FDA), etc.), so the safety and efficacy of conventional devices and apps is not well established.

[0021] Conventionally, sleep monitoring devices such as smartwatches, rings, and fitness bands do not have clinical validation for accurate monitoring of sleep staging (e.g., the classification of sleep into distinct stages), sleep fragmentation (e.g., the interruption of continuous sleep by frequent arousals and / or awakenings), and sleep parameters (e.g., quantifiable characteristics of sleep used to assess sleep quality and architecture) like a conventional polysomnography sleep exam, as they do not use electroencephalogram (EEG) sensors for obtaining signals from the brain during sleep.Attorney Docket No.: 46528.6 (L0002PCT)

[0022] On the other hand, the EEG systems currently available for sleep exams are cumbersome and uncomfortable, as they are not wearable, not portable, not lightweight, not integrated with digital therapy, and cannot monitor sleep continuously. Conventional EEG systems are also not intended for in-home care settings. Exams involving EEG sensors and readings are conventionally done at hospitals and clinics, where the exams are administered by trained staff utilizing dozens of sensors, making such exams expensive and intimidating to patients.

[0023] Furthermore, there is currently no way of monitoring EEG readings continuously and accurately for in-home care treatments. There is also no current method for combining readings from an EEG sensor with biometric data readings from other wearables such as an oximeter ring, an accelerometer for conducting an actigraphy, continuous positive airway pressure (CPAP) masks, smartwatches, etc. for medical applications. The combinations of various biometric data provide clinical information for in-home care diagnostics and treatment, which is not available with current technologies.

[0024] In addition, it is currently not possible to monitor EEG readings continuously in combination with medication treatment, which provides information for patients and clinicians on the safety and efficacy of medications and / or treatments. Currently, the only way for measuring brain signals via EEG readings is with a conventional polysomnography exam, which is conventionally conducted before the start of the treatment. However, this exam is rarely done during the course of the treatment due to its high cost, lack of portability, and low accessibility, which negatively impacts the success of the treatment. Furthermore, the polysomnography exam is not readily accessible in most countries, with some patients waiting on waiting lists months to years long to receive access to a polysomnography exam.

[0025] Furthermore, there is currently no digital behavioral treatment integrated with EEG monitoring. Conventional digital behavioral treatments such as Cognitive Behavioral Therapy for Insomnia (CBT-I) do not utilize objective measurements from EEG readings.Additionally, the conventional EEG diagnostic devices available at the consumer and / or patient level are not digitally integrated with treatment.

[0026] Conventional solutions suffer from practical limitations such as bulky, uncomfortable equipment that is not suitable for continuous or in-home use. Data collected by traditional devices often lack integration between EEG signals, biometric sensors, and user input, resulting in incomplete or inaccurate diagnostics. Furthermore, these systems typically cannot update or personalize treatment recommendations in real-time, impeding effective and individualized sleep therapy. Conventional solutions are also hindered by technicalAttorney Docket No.: 46528.6 (L0002PCT) limitations including inefficient data processing and transmission protocols, which lead to increased processing overhead, higher power consumption, and greater latency in analysis and feedback. Conventional systems often rely on subjective user input or indirect biometric proxies rather than direct EEG measurements, resulting in lower data integrity and reduced diagnostic accuracy for sleep staging and disorder detection. The lack of continuous, in-home monitoring restricts data collection to episodic, clinic-based assessments, missing longitudinal trends and real-world variability in sleep patterns. Additionally, these systems are unable to provide real-time updates to machine learning models, resulting in static, nonpersonalized therapy recommendations that fail to adapt to individual user needs or changes in sleep behavior. Collectively, these disadvantages limit the effectiveness, convenience, and clinical utility of conventional sleep monitoring and therapy solutions.

[0027] The devices, systems, and methods disclosed herein provide solutions to these and other shortcomings of conventional systems.

[0028] A processing device identifies electroencephalogram (EEG) sensor data obtained by a wearable device associated with a head of a user (e.g., patient, client, a third-party, a clinician, etc.). In some embodiments, the wearable device is configured to store the EEG sensor data, where the EEG sensor data is obtained by the processing device from the wearable device responsive to a wireless connection (e.g., wireless personal area network (WPAN, Bluetooth®, etc.), wireless local area network (WLAN, Wi-Fi®, etc.), etc.) being established with the wearable device. In some embodiments, the wearable device includes a memory configured to store between about 20 hours and about 80 hours of EEG sensor data. In some embodiments, the wearable device is configured to sample the EEG sensor data at about 128 Hz to about 512 Hz, store the sampled data locally (optionally subsequently aggregating it into predetermined time intervals or epochs (e.g., 30 seconds) via the processing device), and transmit the stored EEG sensor data in discrete packets to the processing device after a monitoring period has ended. The stored EEG sensor data may be partitioned into multiple discrete packets formatted for low-power wireless transmission (e.g, via Bluetooth® or Wi-Fi®) after the monitoring period. Each packet may include metadata such as timestamps, sequence identifiers, and integrity checks (e.g., checksums) and may encapsulate the entire about 128 Hz to about 512 Hz sample of the one or more predetermined epochs or segments of EEG data to support validation ordered by the processing device. In some embodiments, packets are encrypted consistent with the firmware’s hardware-based security features.Attorney Docket No.: 46528.6 (L0002PCT)

[0029] The processing device further provides the EEG sensor data as a first data input to a first trained machine learning model. The processing device further obtains a first output from the first trained machine learning model, where the first output is associated with sleep quality data associated with the user. In some embodiments, the first trained machine learning model is trained by identifying historical EEG sensor data, identifying historical sleep quality data, and training a first machine learning model using data input including the historical EEG sensor data and target output including the historical sleep quality data to generate the first trained machine learning model.

[0030] The processing device further provides a second input to a second trained machine learning model, where the second input includes the sleep quality data (e.g., the first output from the first machine learning model), and user input data. In some embodiments, the user input data includes user input associated with the user, where the user input is received via a user interface. In some embodiments, the user input data includes biometric data associated with the user. The biometric data may be different from the EEG sensor data and may be obtained by at least one additional sensor device and / or API connection.

[0031] The processing device further obtains a second output from the second trained machine learning model, where the second output is associated with sleep therapy data. In some embodiments, the second trained machine learning model is trained by identifying historical sleep quality data, historical user input, and historical biometric data; identifying historical sleep therapy data; and training a second machine learning model using data input including the historical sleep quality data, the historical user input, and the historical biometric data and target output including the historical sleep therapy data to generate the second trained machine learning model.

[0032] Subsequent to obtaining the second output associated with the sleep therapy data, the processing device further identifies subsequent EEG sensor data and / or subsequent user input data. Responsive to determining that the subsequent EEG sensor data and / or the subsequent user input data meets a first threshold value, the processing device further retrains the first trained machine learning model and / or the second trained machine learning model to generate updated sleep therapy data. In some embodiments, the re-training is based on the subsequent EEG sensor data and / or the subsequent user input data. In some embodiments, the first threshold value includes a predefined variance between a baseline sleep-quality metric included in the sleep quality data and a corresponding metric derived from the subsequent EEG sensor data and / or the subsequent user input data. In some embodiments, the baseline sleep-quality metric includes a sleep efficiency value, a sleepAttorney Docket No.: 46528.6 (L0002PCT) onset latency value, a wake-after-sleep-onset value, a total rapid eye movement (REM) sleep duration value, a total non-rapid eye movement (NREM) sleep duration value, and / or a total sleep time value.

[0033] In some embodiments, the processing device further receives additional EEG sensor data associated with the user, where the additional EEG sensor data is received after providing the sleep therapy data. In some embodiments, the processing device further determines, based on the additional EEG sensor data, whether the sleep therapy data is producing a targeted effect. The targeted effect may include improved EEG sensor data, improved sleep quality data, or similar data.

[0034] In some embodiments, the devices, systems, and methods disclosed provide accurate sleep monitoring and digital personalized treatment for diagnostics and treatment of sleep disorders, using objective measurements (e.g., EEG sensor data, biometric data, etc.) and subjective measurements (e.g., user input) from a user interface (e.g., a patient-facing mobile application).

[0035] In some embodiments, the present disclosure includes systems, methods, and / or devices to improve sleep quality and support diagnostic decisions using EEG sleep monitoring. The present disclosure includes a system (e.g., a wearable sensor system) that includes: awearable device (e.g., a wearable EEG electrode), a processing device executing instructions (e.g., a patient-facing computer program), a first machine learning model (e.g., a sensor monitoring algorithm), and a second machine learning model (e.g., a therapy algorithm). In some embodiments, the system further includes a server (e.g., a backend system and data server) and / or a second processing device executing instructions (e.g., a clinician -facing computer program). Inputs from the system and / or processing device may be transformed into outputs by the first machine learning model and the second machine learning model to be used for sleep enhancement and diagnostics of sleep disorders.

[0036] The present disclosure addresses the lack of accessible, effective, and affordable sleep disorder treatments and diagnostics, the lack of sleep psychologists offering effective insomnia treatments, and the lack of in-home EEG monitoring for more personalized, accurate and effective treatment and diagnostics.

[0037] The present disclosure is widely applicable and can assist with in-home monitoring, treatment, and diagnosis of sleep and / or brain disorders. The present disclosure enables a treatment plan that may be accessed remotely, digitally, effectively, and affordably, thus closing a gap in primary and chronic care, mainly for the aging population. The present disclosure provides connected methodologies and systems containing processing devices,Attorney Docket No.: 46528.6 (L0002PCT) executable instructions (e.g., software, computer programs, etc.), wearable sensor systems, and / or algorithms to generate accurate, effective, and accessible information for diagnostics and treatment of sleep disorders. The integrated system of the present disclosure may combine objective monitoring from wearable sensors (e.g., the wearable device including at least one EEG electrode, additional sensors, etc.) with subjective monitoring from patientfacing computer programs in order to provide accurate and personalized information for diagnostics and treatments, which is not possible with conventional solutions.

[0038] The systems, devices, and methods disclosed herein have advantages over conventional solutions. The integration of a wearable device (e.g., EEG headband) with digital behavioral therapy allows for a more effective, personalized, complete, and seamless diagnostics and treatment combination. The personalization of the cognitive behavioral therapy for insomnia, which combines objective measures (e.g., from EEG readings) together with subjective measure (e.g., from clinical diaries, anamneses, questionnaires, therapy, etc.). Combining EEG sensor data with other objective non-EEG measures from additional sensors such as oximeters, smartwatches, fitness bands, rings, accelerometers for actigraphy, CPAP masks, etc. These combinations are not obtainable with conventional technologies and provide clinical information for diagnostics and treatment, such as better detection of awakenings, greater accuracy in cases of sleep fragmentation, improved detection of hypopneas not associated with desaturation, distinction of desaturation events during wakefulness, correction of sleep apnea metrics based on total sleep time, assistance in the diagnosis of sleep disorders (e.g., insomnia, sleep apnea, co-morbid insomnia and sleep apnea (COMISA), narcolepsy, hypersomnia, etc.).

[0039] Furthermore, the use of EEG sensor data with user input (e.g., a medication management diary) may allow for a substantially continuous assessment of efficacy and safety of medication, as well as provide support for withdrawal of medication, as the clinician is able to see the objective sleep data (e.g., EEG sensor data, sleep quality data, etc.) in order to adjust the medication accordingly via a second processing device (e.g., the clinician-facing computer program). Some insomnia medications may have side effects, dependency and tolerance, and alter the sleep stages by potentially increasing deep sleep and reducing REM sleep, which affect cognition such as memory. Therefore, the present disclosure may provide periodic (e.g., real-time, daily, weekly, etc.) assessment and continuous objective sleep monitoring using EEG sensor data, which may increase the safety and efficacy of prescribed medications.Attorney Docket No.: 46528.6 (L0002PCT)

[0040] In some embodiments, the wearable device of the present disclosure is lightweight and flexible and storesEEG sensor readings, so there is no need for real-time wireless (e.g., Bluetooth®) connection during use (e.g., during sleep). In some embodiments, the wearable device of the present disclosure has a longbattery life (e.g., at least 8 hour battery life, at least 12 hour battery life, at least 16 hour battery life, at least 20-hour battery life, at least 24-hour battery life, at least 50 -hour battery life, at least 80-hour battery life, etc.) in order to collect data during the day and night for a more comprehensive analysis of sleep disorders (e.g., narcolepsy) where sleep is to be assessed during the day. The wearable device (e.g., headband) of the present disclosure may also be customizable for children and adults by changing colors, size, closure, etc. The EEG sensor(s) (e.g., the headband electrode(s)) of the present disclosure can be optionally disposable and easily replaced with a new one using a detachable connection.

[0041] The present disclosure overcomes limitations of conventional solutions and results in technical advantages by providing a lightweight, comfortable, washable, and wearable EEG system specifically designed for continuous, in-home use. The system integrates objective EEG measurements with additional biometric sensor data and user input, enabling comprehensive and accurate sleep diagnostics. By employing optimized data collection and transmission protocols, the present disclosure reduces processing overhead, conserves device power, and minimizes latency in data analysis and feedback. Unlike conventional systems that rely on subjective or indirect measures, the present disclosure’s direct EEG monitoring ensures high data integrity and precise sleep staging, supporting robust detection of sleep disorders. Continuous, real-world monitoring allows for the capture of longitudinal sleep trends and natural variability, providing a more complete picture of the user’s sleep health. Furthermore, the system enables continuous (e.g., real-time, daily, periodic, etc.) updates to machine learning models, allowing therapy recommendations to be dynamically personalized and adapted to the user’s evolving needs and behaviors. These technical advantages collectively enhance the effectiveness, convenience, and clinical value of sleep monitoring and therapy, addressing the shortcomings of prior solutions.

[0042] FIGS. 1 A-B are block diagrams illustrating the overall components and architecture of a sleep monitoring and assessment system, according to certain embodiments. FIG. 1 A is a block diagram illustrating the overall components of the sleep monitoring and assessment system 100A (e.g., the system), accordingto some embodiments. FIG. IB is a block diagram illustrating a system architecture 100B, according to some embodiments. Components andAttorney Docket No.: 46528.6 (L0002PCT) features of FIG. 1 A and / or IB that have the same or similar name and / or reference number may have the same or similar functionality, material, components, etc.

[0043] Referringto FIG. 1 A, the sleep monitoring and assessment system 100A includes a wearable device 104 (e.g., a wearable sensor system) and a client device 120 (e.g., a patient device, a clinician device, a user device, etc.). In some embodiments, the wearable device 104 is associated with a head of a user (e.g., is worn on a user’s head). The wearable device 104 may be a headband that includes one or more EEG sensors. Further aspects of the wearable device 104 will be described in reference to FIGS. 2A-E.

[0044] In some embodiments, the wearable device 104 is configured to obtain (e.g., collect) electroencephalogram (EEG) sensor data (e.g., input, biometric EEG data, etc.). EEG sensor data 102 may refer to recordings of a user’s brain activity that is measured via an EEG sensor (e.g., electrode). EEG sensor data 102 may include signals that reflect the synchronized firing of neurons that are captured in the form of various waveforms that depend on the user’s brain state. EEG sensor data 102 may be recorded during the user’s sleep period (and, in some embodiments, during wakefulness), stored, and subsequently processed and interpreted to provide insight into a user’s brain activity while the user is sleeping, enabling objective assessment of the user’s various sleep stages, sleep disturbances, and overall sleep quality.

[0045] This collected EEG sensor data 102 is provided (e.g., sent) to a client device 120. The client device 120 may include control logic 122 (e.g., executed by a processing device, a patient-facing computer program). The client device 120 (e.g., via control logic 122, via processing device) may be configured to execute instructions or otherwise interact with the wearable device 104, as will be describedin detail. In some embodiments, the system 100A includes memory coupled to a processing device (e.g., that executesthe control logic 122). In some embodiments, the memory is part of the wearable device 104. In some embodiments, the memory is part of the client device 120. In some embodiments, the memory is separate from the wearable device 104 and / or the client device 120. In some embodiments, the memory is part of both the wearable device 104 (e.g., a wearable device memory) and the client device 120 (e.g., a client device memory). The memory can include volatile memory and optionally also non-volatile memory, such as a storage device. For example, the memory can be a non-transitory machine-readable storage medium storing instructions which, when executed by the processing device (e.g., of the wearable device 104, of the client device 120), causes the processing device to perform particular operations that will be discussed herein.

[0046] The control logic 122 may provide the EEG sensor data 102 to a first trained machine learning model 190 as a first data input. The first trained machine learning modelAttorney Docket No.: 46528.6 (L0002PCT)190 and the associated methodology will be described in greater detail with reference to FIG. 3.

[0047] In some embodiments, the first trained machine learning model 190 may provide sleep quality data associated with the user as a first output 192. The sleep quality data may be objective data that is based on the EEG sensor data 102 collected by the wearable device 104 associated with the head of the user. The sleep quality data output by the first trained machine learning model 190 may include a range of quantitative and qualitative metrics that characterize the user’s sleep. For example, the sleep quality data may comprise sleep stage percentages (e.g., the proportion of time spent in REM, Nl, N2, N3, and wake stages), total sleep time, sleep efficiency, sleep onset latency, sleep stage transitions, REM latency, wake after sleep onset (WASO), and overall sleep duration. These metrics may be derived from the analysis of the EEG sensor data 102, which is processed and segmented into analysis windows (e.g., 30-second epochs) and classified according to established sleep staging criteria.

[0048] In some embodiments, the first trained machine learning model 190 may be trained using labeled data from gold-standard polysomnography exams, enabling the model to achieve high accuracy in sleep staging and event detection. The model may be validated to achieve accuracy levels exceeding 60%, 70%, 80%, 90%, or even 95% in comparison to full polysomnography (type 1) exams. The control logic 122 may execute the first trained machine learning model 190 locally on a mobile device, desktop computer, or over a cloud system, providing flexibility in deployment and scalability for remote or home-based sleep monitoring.

[0049] The control logic 122 may provide the first output 192 from the first trained machine learning model 190 as a second input to a second trained machine learning model 195 (e.g., a set of therapy algorithms). The second trained machine learning model 195 and the associated methodology will be described in greater detail with reference to FIG. 4.

[0050] The control logic 122 may obtain a second output 197 (e.g., sleep therapy data) from the second trained machine learning model 195. The sleep therapy data output by the second trained machine learning model 195 may include quantitative and qualitative recommendations, feedback, or interventions designed to improve sleep quality and address specific sleep disorders of the user. For example, the sleep therapy data may include individualized sleep schedules, behavioral modification strategies, cognitive behavioral therapy (CBT) modules, reminders, notifications, or other guidance related to sleep hygiene, medication management, or device usage (e.g., CPAP or intraoral devices).Attorney Docket No.: 46528.6 (L0002PCT)

[0051] In some embodiments, the second trained machine learning model 195 may utilize advanced machine learning techniques, such as supervised learning, non-supervised learning, deep learning, reinforcement learning, Large Language Models (LLM), or digital twins to adapt therapy recommendations overtime based on ongoing user data and therapy outcomes. The sleep therapy data may be presented to the user through the client device 120 (e.g., executing patient-facing computer program) and / or to a clinician via a second client device 125 executing a second control logic 127 (e.g., a clinician-facing computer program, a second processing device, etc.), enablingboth self-guided and clinician-supported therapy pathways. This adaptive, data-driven approach supports continuous improvement in sleep health and therapy efficacy, leveraging both periodic sensor data and user engagement.

[0052] In some embodiments, the sleep therapy data generated by the second trained machine learning model 195 may be delivered to the user and / or clinician in a variety of formats, including text, graphs, videos, notifications, and reminders. The output may include actionable information related to sleep schedules, daily routines, habits, and behaviors, as well as personalized cognitive techniques from cognitive behavioral therapy for insomnia, supporting both education, intervention and engagement in the therapy process. The control logic 122 may present these outputs through the user interface, enabling users to track progress, receive personalized feedback, and access multimedia resources that reinforce healthy sleep practices.

[0053] In some embodiments, the control logic 122 (e.g., via the wearable device 104 and / or additional biometric sensors) may monitor the user over a period of time (e.g., about 30 days, at least 7 days, or at least 3 days, etc.) to assess therapy effectiveness and adapt recommendations accordingly. This longitudinal monitoring enables the system to detect trends, evaluate the impact of interventions, and refine therapy protocols for optimal outcomes. The sleep therapy data may also be used to support diagnostic decisions or treatment for a range of sleep disorders, including insomnia, obstructive sleep apnea, narcolepsy, REM latency disorder, co-morbid insomnia and sleep apnea (COMISA), as well as to guide CPAP therapy, intraoral device use, medication management, and cognitive behavioral therapy.

[0054] In some embodiments, the client device 120 sends the EEG sensor data 102 to a server machine 170 (e.g., a backend system, a data server, etc.). The first trained machine learning model 190 and / or the second trained machine learning model 195 may be implemented on the server machine 170, and the EEG sensor data 102 and other data (e.g., the user input data) may be processed by the trained machine learning models via the serverAttorney Docket No.: 46528.6 (L0002PCT) machine 170. The second output 197 (e.g., the sleep therapy data) may then be sent to the second client device 125 (e.g., executing second control logic 127, executing the clinicianfacing computer program), as well as to the client device 120 (e.g., executing control logic 122, executing the patient-facing computer program), where the second output 197 is presented to the user (e.g., the patient) and / or the clinician. During this process, some data can also move back and forth, such as from the client device 120 to the server machine 170, in the case of non-EEG biometric data or subjective data collected from the user interface of the client device 120.

[0055] Referring to FIG. IB, the system architecture 100B (e.g., system) includes the wearable device 104, the client device 120 configured to execute the control logic 122, a predictive server 112, and a data store 140. In some embodiments, the predictive server 112 is part of a predictive system 110. In some embodiments, the predictive system 110 further includes server machine 170 and / or server machine 180. In some embodiments, the client device 120 includes a recommendation component 123 (e.g., the control logic 122 of FIG. 1A).

[0056] In some embodiments, one or more of the wearable device 104, additional biometric sensor(s) 126, client device 120, second client device 125, predictive server 112, data store 140, server machine 170, and / or server machine 180 are coupled to each other via a network 130 for generating predictive data 116 for the sleep therapy data (e.g., to perform a corrective action). In some embodiments, network 130 is a public network that provides the client device 120 and / or the second client device 125 with access to the predictive server 112, data store 140, and other publicly available computing devices. In some embodiments, network 130 is a private network that provides the client device 120 and / or the second client device 125 access to the wearable device 104, the additional biometric sensors 126, the data store 140, and other privately available computing devices. In some embodiments, network 130 includes one or more Wide Area Networks (WANs), Local Area Networks (LANs), wired networks (e.g., Ethernet network), wireless networks (e.g., an 802. 11 network or a Wi-Fi® network), WPAN (e.g., Bluetooth®), cellular networks (e.g., a Long Term Evolution (LTE) network), routers, hubs, switches, server computers, cloud computing networks, and / or a combination thereof.

[0057] In some embodiments, the client device 120 and / or the second client device 125 includes a computing device such as a Personal Computer (PC), laptop, mobile phone, smart phone, tablet computer, netbook computer, etc. In some embodiments, the client device 120 includes a recommendation component 123. In some embodiments, the recommendationAttorney Docket No.: 46528.6 (L0002PCT) component 123 may also be included in the predictive system 110 (e.g., machine learning processing system). In some embodiments, the recommendation component 123 is alternatively included in the predictive system 110 (e.g., instead of being included in client device 120). The client device 120 and / or the second client device 125 includes an operating system that allows users to one or more of consolidate, generate, view, or edit data, provide directives to the predictive system 110 (e.g., machine learning processing system), etc.

[0058] In some embodiments, the recommendation component 123 receives user input (e.g., via a user interface such as a Graphical User Interface (GUI) displayed via the client device 120), receives EEG sensor data 142 (e.g., the EEG sensor data 102 of FIG. 1A from the wearable device 104, from the client device 120, from the data store 140), receives sleep quality data 152 (e.g., from the client device 120, from the data store 140), receives user input data 162 (e.g., from the additional biometric sensor(s) 126, from the client device 120, from the data store 140), etc. In some embodiments, the recommendation component 123 one or more of transmits the data (e.g., user input, EEG sensor data 142, sleep quality data 152, user input data 162, etc.) to the predictive system 110, receives predictive data 116 from the predictive system 110, generates updated sleep therapy data (e.g., performance of a corrective action, re-training the first and / or second trained machine learning models) based on the predictive data 116, etc. In some embodiments, the recommendation component 123 stores data (e.g., user input, EEG sensor data 142, sleep quality data 152, user input data 162, etc.) in the data store 140 and the predictive server 112 retrieves data from the data store 140. In some embodiments, the predictive server 112 stores output (e.g., predictive data 116) of the first trained machine learning model 190 and / or the second trained machine learning model 195 in the data store 140 and the client device 120 and / or the second client device 125 retrieves the output from the data store 140. In some embodiments, the recommendation component 123 receives an indication of subsequent EEG sensor data 148 and / or subsequent user input data 168 meeting a threshold value (e.g., a first threshold value) from the predictive system 110 and causes the first and / or second trained machine learning models to be re-trained to generate updated sleep therapy data based on the determination that subsequent EEG sensor data 148 and / or subsequent user input data 168 meets the threshold value.

[0059] In some embodiments, the predictive data 160 is associated with predicted performance data (e.g., sleep therapy data, updated sleep therapy data, etc.). In some embodiments, a corrective action is performed based on the predictive data 160. In some embodiments, the corrective action includes one or more of providing an alert (e.g., aAttorney Docket No.: 46528.6 (L0002PCT) notification and / or a reminder containing information about sleep schedules, routines, habits, behaviors, CBTi therapy techniques, etc. based on the sleep therapy data), generating sleep therapy data based on EEG sensor data 142, sleep quality data 152, and / or user input data 162, generating updated sleep therapy data based on subsequent EEG sensor data 148 and / or subsequent user input data 168, etc. In some embodiments, the corrective action includes providing feedback control (e.g., determining whether the sleep therapy data is producing a targeted effect, such as improved EEG sensor data and / or improved sleep quality data).

[0060] In some embodiments, the predictive server 112, server machine 170, and server machine 180 each include one or more computing devices such as a rackmount server, a router computer, a server computer, a personal computer, a mainframe computer, a laptop computer, a tablet computer, a desktop computer, Graphics Processing Unit (GPU), accelerator Application-Specific Integrated Circuit (ASIC) (e.g., Tensor Processing Unit (TPU)), etc.

[0061] The predictive server 112 includes a predictive component 114. In some embodiments, the predictive component 114 receives the EEG sensor data 142 (e.g., receives from the client device 120, retrieves from the data store 140) and generates sleep quality data 152 associated with objective measurements of a user’s sleep quality. The predictive component 114 may further receive the sleep quality data 152 and the user input data 162 (e.g., receives from the client device 120, retrieves from the data store 140) and generates predictive data 116 (e.g., sleep therapy data). In some embodiments, the predictive component 114 uses one or more trained machine learning models (e.g., the first trained machine learning model 190, the second trained machine learning model 195) to determine the predictive data 116 associated with a user’s sleep quality and / or user sleep therapy data. In some embodiments, the first trained machine learning model 190 is trained using historical sensor data (e.g., historical EEG sensor 144) and historical performance data (e.g., historical sleep quality data 154). In some embodiments, the second trained machine learning model 195 is trained using historical sleep quality data 154, historical user input data 164 (e.g., historical user input, historical biometric data), and historical performance data (e.g., historical sleep therapy data).

[0062] In some embodiments, the predictive system 110 (e.g., the predictive server 112, the predictive component 114) generates predictive data 116 using supervised machine learning (e.g., supervised data set, historical EEG sensor data 144 labeled with historical sleep quality data 154, etc.). In some embodiments, the predictive system 110 generates predictive data 116 using semi-supervised learning (e.g., historical EEG sensor data 144 is of normal sleepAttorney Docket No.: 46528.6 (L0002PCT) patterns and quality, semi-supervised data set, sleep quality data 152 is a predictive percentage, etc.). In some embodiments, the predictive system 110 generates predictive data 116 using unsupervised machine learning (e.g., unsupervised data set, clustering, clustering based on historical EEG sensor data 144, etc.).

[0063] In some embodiments, the EEG sensor data 142 (e.g., historical EEG sensor data 144, currentEEG sensor data 146, subsequent EEG sensor data 148, etc.) is processed by the client device 120 and / or by the predictive server 112. In some embodiments, processing of the EEG sensor data 142 includes generating features. In some embodiments, the features are a portion of the EEG sensor data, processed EEG sensor data, patterns in the EEG sensor data 142, or a combination of values from the EEG sensor data 142 (e.g., ratio of sensor values, etc.). In some embodiments, the EEG sensor data 142 includes features that are used by the predictive component 114 for obtaining the sleep quality data 152 and / or the predictive data 116.

[0064] In some embodiments, the sleep quality data 152 (e.g., historical sleep quality data 154, current sleep quality data 156, etc.) is processed by the client device 120 and / or by the predictive server 112. In some embodiments, processing of the sleep quality data 152 includes generating features. In some embodiments, the features are a portion of the sleep quality data, processed sleep quality data, patterns in the sleep quality data 152, or a combination of values from the sleep quality data 152. In some embodiments, the sleep quality data 152 includes features that are used by the predictive component 114 for obtaining the predictive data 116 (e.g., the sleep therapy data).

[0065] In some embodiments, the user input data 162 (e.g., historical user input data 164, current user input data 166, subsequent user input data 168, etc.) is processed by the client device 120 and / or by the predictive server 112. In some embodiments, processing of the user input data 162 includes generating features. In some embodiments, the features are a portion of the user input data, processed user input data, patterns in the user input data 162, or a combination of values from the user input data 162. In some embodiments, the user input data 162 includes features that are used by the predictive component 114 for obtaining the predictive data 116 (e.g., the sleep therapy data). In some embodiments, the features of the sleep quality data 152 and the features of the user input data 162 are combined and used by the predictive component 114 for obtaining the predictive data 116 (e.g., the sleep therapy data).

[0066] In some embodiments, the data store 140 is memory (e.g., random access memory), a drive (e.g., a hard drive, a flash drive), a database system, or another type of component orAttorney Docket No.: 46528.6 (L0002PCT) device capable of storing data. In some embodiments, the data store 140 includes multiple storage components (e.g., multiple drives or multiple databases) that span multiple computing devices (e.g., multiple server computers). In some embodiments, the data store 140 stores one or more of EEG sensor data 142, sleep quality data 152, user input data 162, and / or predictive data 116 (e.g., the sleep therapy data).

[0067] EEG sensor data 142 includes historical EEG sensor data 144, current EEG sensor data 146, and subsequentEEG sensor data 148. In some embodiments, EEG sensor data 142 may be obtained by the wearable device 104 and may be associated with a user’s objective sleep data (e.g., brain wave patterns associated with sleep). In some embodiments, at least a portion of the EEG sensor data 142 is from client device 120, data store 140, and / or the wearable device 104.

[0068] Sleep quality data 152 (e.g., performance data, the first output 192 of FIG. 1A) includes historical sleep quality data 154 (e.g., historical performance data) and current sleep quality data 156 (e.g., current performance data). The sleep quality data 152 may include a baseline sleep-quality metric derived from the EEG sensor data 142. In some embodiments, the baseline sleep-quality metric includes at least one of a sleep efficiency value, a sleep onset latency value, a wake-after-sleep-onset value, a total rapid eye movement (REM) sleep duration value, a total non-rapid eye movement (NREM) sleep duration value (e.g., sleep stages N1 N2 N3, NREM), a total sleep time value, or other sleep-quality metrics. The baseline sleep-quality metrics may further include sleep fragmentation, microarousal, desaturation index, heart rate, movement, sleep schedules, time to bed, time to sleep, wake-up time, go-to-bed time, sleep duration, sleep latency, REM sleep latency, REM sleep, non-rapid eye movement (NREM) sleep, percentage and duration of sleep stages, wake after sleep onset (WASO), sleep transitions, sleep efficiency, etc. In some examples, the sleep quality data 152 is indicative of a user’s sleep quality. In some embodiments, at least a portion of the sleep quality data 152 is associated with a quality of EEG sensor data 142 obtained by the wearable device 104. In some embodiments, at least a portion of the sleep quality data 152 is based on EEG sensor data from the wearable device 104 (e.g., historical sleep quality data 154 includes EEG sensor data indicating objective sleep-related metrics, etc.).

[0069] User input data 162 includes historical user input data 164, current user input data 166, and subsequentuser input data 168. In some embodiments, at least a portion of the user input data 162 is associated with subjective sleep data (e.g., user input) that is provided by a user via a user interface of the client device 120. The subjective sleep data may be associated with user input associated with diaries, anamneses, clinical assessment questionnaires, andAttorney Docket No.: 46528.6 (L0002PCT)CBTi etc. In some embodiments, at least a portion of the user input data 162 is associated with biometric data (e.g., biometric data receives from additional biometric sensor(s) 126). The biometric data may be objective data associated with data received from an EEG sensor, a photoplethysmogram (PPG) sensor, an electrocardiogram (ECG) sensor, an accelerometer sensor, etc. In some examples, the user input data 162 is indicative of whether a user is experiencing efficient sleep and / or whether the sleep therapy data (e.g., predictive data 116) is producing a targeted effect.

[0070] In some embodiments, the client device 120 provides sleep quality data 152 (e.g., performance data). In some examples, the client device 120 provides (e.g., based on user input) sleep quality data 152 that indicates an abnormality in user brain waves associated with sleep (e.g., disordered sleep).

[0071] In some embodiments, historical data includes one or more of historical EEG sensor data 144, historical sleep quality data 154, and / or historical user input data 164 (e.g., at least a portion for training the first trained machine learning model 190 and / or the second trained machine learning model 195). Current data includes one or more of current EEG sensor data 146, current sleep quality data 156, and / or current user input data 166 (e.g., at least a portion to be input into the first trained machine learning model 190 and / or second trained machine learning model 195 subsequentto training the models using the historical data). Subsequent data includes one or more of subsequent EEG sensor data 148 and / or subsequent user input data 168. In some embodiments, the subsequent data is used for retraining the first trained machine learning model 190 and / or the second trained machine learning model 195.

[0072] In some embodiments, the predictive data 116 is to be used for sleep quality monitoring (e.g., to produce a targeted effect). The targeted effect may include at least one of improved EEG sensor data 142 and / or improved sleep quality data 152.

[0073] In some embodiments, predictive system 110 further includes server machine 170 and server machine 180. Server machine 170 includes a data set generator 172 that is capable of generating data sets (e.g., a set of data inputs and a set of target outputs) to train, validate, and / or test the first machine learning model 190 and / or the second machine learning model 195. The data set generator 172 has functions of data gathering, compilation, reduction, and / or partitioning to put the data in a form formachine learning. In some embodiments (e.g, for small datasets), partitioning (e.g., explicit partitioning) for post-training validation is not used. Repeated cross-validation (e.g., 5-fold cross-validation, leave-one-out-cross-validation) may be used during training where a given dataset is in-effect repeatedly partitioned into different training and validation sets during training. A model (e.g., the best model, the modelAttorney Docket No.: 46528.6 (L0002PCT) with the highest accuracy, etc.) is chosen from vectors of models over automatically- separated combinatoric subsets. In some embodiments, the data set generator 172 may explicitly partition the historical data (e.g., historical EEG sensor data 144 and corresponding historical sleep quality data 154, historical user input data 164) into a training set, a validating set, and / or a testing set. In some embodiments, the predictive system 110 (e.g., via predictive component 114) generates multiple sets of features (e.g., training features). In some examples, a first set of features corresponds to a first set of types of EEG sensor data 142 (e.g., first types of EEG signals, first combination of values, first patterns in the values) that correspond to each of the data sets (e.g., training set, validation set, and testing set) and a second set of features correspond to a second set of types of EEG sensor data 142 (e.g., second types of EEG signals different from the first types of EEG signals, second combination of values different from the first combination, second patterns different from the first patterns) that correspond to each of the data sets.

[0074] Server machine 180 includes a training engine 182, a validation engine 184, selection engine 185, and / or a testing engine 186. In some embodiments, an engine (e.g., training engine 182, a validation engine 184, selection engine 185, and a testing engine 186) refers to hardware (e.g., circuitry, dedicated logic, programmable logic, microcode, processing device, etc.), software (such as instructions run on a processing device, a general purpose computer system, or a dedicated machine), firmware, microcode, or a combination thereof. The training engine 182 is capable of training a first machine learning model 190 and / or a second machine learning model 195 using one or more sets of features associated with the training set from data set generator 172. In some embodiments, the training engine 182 generates multiple trained machine learning models 190, where each trained machine learning model 190 corresponds to a distinct set of parameters of the training set (e.g., EEG sensor data 142, user input data 162) and corresponding responses (e.g., sleep quality data 152, predictive data 116). In some embodiments, multiple models are trained on the same parameters with distinct targets for the purpose of modeling multiple effects.

[0075] The validation engine 184 is capable of validating a trained machine learning model (e.g., the first trained machine learning model 190, the second trained machine learning model 195) using a corresponding set of features of the validation set from data set generator 172. For example, a first trained machine learning model 190 that was trained using a first set of features of the training set is validated using the first set of features of the validation set. The validation engine 184 determines an accuracy of each of the trained machine learning models 190 based on the corresponding sets of features of the validation set. The validationAttorney Docket No.: 46528.6 (L0002PCT) engine 184 evaluates and flags (e.g., to be discarded) trained machine learning models (e.g., the first trained machine learning model 190, the second trained machine learning model 195) that have an accuracy that does not meet a threshold accuracy. In some embodiments, the selection engine 185 is capable of selecting one or more trained machine learning models (e.g., the first trained machine learning model 190, the second trained machine learning model 195) that have an accuracy that meets a threshold accuracy. In some embodiments, the selection engine 185 is capable of selecting the trained machine learning model (e.g., the first trained machine learning model 190, the second trained machine learning model 195) that has the highest accuracy of the trained machine learning models.

[0076] The testing engine 186 is capable of testing a trained machine learning model (e.g., the first trained machine learning model 190, the second trained machine learning model 195) using a corresponding set of features of a testing set from data set generator 172. For example, a first trained machine learning model 190 that was trained using a first set of features of the training set is tested using the first set of features of the testing set. The testing engine 186 determines a trained machine learning model (e.g., the first trained machine learning model 190, the second trained machine learning model 195) that has the highest accuracy of all of the trained machine learning models based on the testing sets.

[0077] In some embodiments, the first trained machine learning model 190 and / or the second trained machine learning model 195 (e.g., used for classification) refers to a model artifact that is created by the training engine 182 using a training set that includes data inputs and corresponding target outputs (e.g., correctly classifies a condition or ordinal level for respective training inputs). Patterns in the data sets can be found that map the data input to the target output (the correct classification or level), and the first trained machine learning model 190 and / or the second trained machine learning model 195 is provided mappings that captures these patterns. In some embodiments, the first trained machine learning model 190 and / or the second trained machine learning model 195 uses one or more of Gaussian Process Regression (GPR), Gaussian Process Classification (GPC), Bayesian Neural Networks, Neural Network Gaussian Processes, Deep Belief Network, Gaussian Mixture Model, or other Probabilistic Learning methods. Non probabilistic methods may also be used including one or more of Support Vector Machine (SVM), Radial Basis Function (RBF), clustering, Nearest Neighbor algorithm (k-NN), linear regression, random forest, neural network (e.g., artificial neural network), etc. In some embodiments, the first trained machine learning model 190 and / or the second trained machine learning model 195 is a multi-variate analysis (MV A) regression model.Attorney Docket No.: 46528.6 (L0002PCT)

[0078] The predictive component 114 provides current EEG sensor data 146 (e.g., as input) to the first trained machine learning model 190 and runs the first trained machine learning model 190 (e.g., on the input to obtain one or more outputs). The predictive component 1 Mis capable of determining (e.g., extracting) predictive data 116 from the first trained machine learning model 190 and determines (e.g., extracts) uncertainty data that indicates a level of credibility that the predictive data 116 corresponds to current sleep quality data 156. In some embodiments, the predictive component 114 or recommendation component 123 use the uncertainty data (e.g., uncertainty function or acquisition function derived from uncertainty function) to decide whether to use the predictive data 116 to provide sleep therapy data or whether to further train the first trained machine learning model 190.

[0079] The predictive component 114 further provides current sleep quality data 156 and current user input data 166 (e.g., as input) to the second trained machine learning model 195 and runs the second trained machine learning model 195 (e.g., on the input to obtain one or more outputs). The predictive component 114 is capable of determining (e.g., extracting) predictive data 116 from the second trained machine learning model 195 and determines (e.g., extracts) uncertainty data that indicates a level of credibility that the predictive data 116 corresponds to current sleep therapy data. In some embodiments, the predictive component 114 or recommendation component 123 use the uncertainty data (e.g., uncertainty function or acquisition function derived from uncertainty function) to decide whether to use the predictive data 116 to provide sleep therapy data or whether to further train the second trained machine laeming model 195.

[0080] For purpose of illustration, rather than limitation, aspects of the disclosure describe the training of one or more machine learning models (e.g., the first trained machine learning model 190, the second trained machine learning model 195) using historical data (i.e., prior data, historical EEG sensor data 144, historical sleep quality data 154, historical user input data 164) and providing current data (e.g., currentEEG sensor data 146, current sleep quality data 156, current user input data 166) into the one or more trained probabilistic machine learning models (e.g., the first trained machine learning model 190, the second trained machine learning model 195) to determine predictive data 116. In other implementations, a heuristic model or rule-based model is used to determine predictive data 116 (e.g., without using a trained machine learning model). In other implementations non-probabilistic machine learning models may be used. The predictive component 114 monitors historical EEG sensor data 144, historical sleep quality data 154, and historical user input data 164.Attorney Docket No.: 46528.6 (L0002PCT)

[0081] In some embodiments, the functions of client device 120, predictive server 112, server machine 170, and server machine 180 are provided by a fewer number of machines. For example, in some embodiments, server machines 170 and 180 are integrated into a single machine, while in some other embodiments, server machine 170, server machine 180, and predictive server 112 are integrated into a single machine. In some embodiments, client device 120 and predictive server 112 are integrated into a single machine.

[0082] In general, functions described in one embodiment as being performed by client device 120, predictive server 112, server machine 170, and server machine 180 can also be performed on predictive server 112 in other embodiments, if appropriate. In addition, the functionality attributed to a particular component can be performed by different or multiple components operating together. For example, in some embodiments, the predictive server 112 determines whether the sleep thereapy data is producing a targeted effect based on the predictive data 116. In another example, client device 120 determines the predictive data 116 based on data received from the trained machine learning model (e.g., the first trained machine learning model 190, the second trained machine learning model 195).

[0083] In addition, the functions of a particular component can be performed by different or multiple components operating together. In some embodiments, one or more of the predictive server 112, server machine 170, or server machine 180 are accessed as a service provided to other systems or devices through appropriate application programming interfaces (APIs).

[0084] In some embodiments, a “user” is represented as a single individual. However, other embodiments of the disclosure encompass a “user” being an entity controlled by a plurality of users and / or an automated source. In some examples, a set of individual users federated as a group of administrators is considered a “user.”

[0085] Although embodiments of the disclosure are discussed in terms of determining predictive data 116 for sleep monitoring of disordered sleep (e.g., to determine if the sleep therapy data is producing a targeted effect, re-training the first and / or second trained machine learning models, etc.), in some embodiments, the disclosure can also be generally applied to monitoring of brain waves or EEG signals of a user.

[0086] FIGS. 2A-E illustrate various aspects of the wearable device 104 (e.g., the wearable sensor system), according to certain embodiments. FIG. 2A is a block diagram 200A illustrating various components of the wearable device 104, according to some embodiments. FIGS. 2B-C illustrate views of the wearable device 104 positioned on a user’s head, according to some embodiments. FIGS. 2D-E illustrate perspective views of the wearableAttorney Docket No.: 46528.6 (L0002PCT) device 104, accordingto some embodiments. Components and features of FIG. 1 A, IB, 2A, 2B, 2C, 2D, and / or 2E that have the same or similar name and / or reference number may have the same or similar functionality, material, components, etc.

[0087] Referring to FIG. 2A, the wearable device 104 includes various components. All the disclosed components may be used together to capture, process, and / or share the data (e.g., the EEG sensor data 102 of FIG. 1 A, etc.) to the client device 120 (e.g., executing a patientfacing computer program). The wearable device 104 includes, in some embodiments, one or more wearable EEG electrodes 202, a wearable processor unit 204, a wearable communication system 206, a wearable storage device 208, a wearable signal quality system 210, a wearable battery device 212, firmware 214, a printed circuit board (PCB) 216, an on / off switch 218 (e.g., a power switch, power control, etc.), and / or additional biometric sensor(s) 220 (e.g., the additional biometric sensor(s) 126 of FIG. IB).

[0088] The wearable device 104 includes one or more wearable EEG electrodes 202 configured to detect electrical activity from the user’s scalp. In some embodiments, the EEG electrodes 202 may be single-channel or multi-channel, and may be constructed from flexible or rigid conductive materials, such as conductive fabric, conductive printing, or metallic elements (e.g., silver, carbon, copper, or gold). The electrodes may be embedded in a headband, hat, headset, patch, or other head-worn device, and may be designed as disposable, permanent, or detachable components. In some embodiments, the electrodes are dry and made of silver-based conductive fabric to maximize comfort and signal quality during extended wear. The placement and design of the EEG electrodes 202 may enable continuous, unobtrusive monitoring of brain activity in home and clinical environments.

[0089] The wearable device 104 further includes a wearable processor unit 204, which is a low-power or ultra-low-power microcontroller responsible for real-time processing of EEG sensor data from the wearable device 104. The wearable processor unit 204 may be configured for signal processing, filtering, wireless communication, power management, and integration with on-chip memory. In some embodiments, the wearable processor unit 204 supports Bluetooth® Low Energy (BLE) or other wireless protocols, and may include hardware-based security features to increase the integrity and privacy of the EEG sensor data and other biometric data of the user. The wearable processor unit’s 204 ultra -low-power consumption enables prolonged use on battery power, supporting long-term and continuous monitoring applications.

[0090] The wearable device 104 further includes a wearable communication system 206, which facilitates data transfer between the wearable device 104 and external devices, such asAttorney Docket No.: 46528.6 (L0002PCT) the client device 120 of FIG. 1 A. This system may utilize low-power, short-range wireless protocols, including Bluetooth®, Wi-Fi®, Zigbee, or near-filed communication (NFC), to transmit collected EEG sensor data and / or other biometric data. In some embodiments, the wearable communication system 206 is configured to transfer data after monitoring sessions, which enables the wearable device 104 to operate autonomously during sleep. The wearable communication system 206 may also support over-the-air firmware updates and secure data transmission to protect user privacy.

[0091] The wearable device 104 further includes a wearable storage device 208, which provides onboard memory for storing EEG sensor data collected during monitoring sessions. Storage options may include internal or external flash memory, ferroelectric random access memory (FRAM), EEPROM, Secure Digital (SD) cards, or embedded Multi Media Card (eMMC) modules. In some embodiments, the wearable storage device 208 is an internal flash memory with a long life cycle and low power consumption, capable of storing several nights of data (e.g., at least 4-7 nights). This local storage capability allows the wearable device 104 to function independently of real-time wireless connections, enhancing usability for users who may not have immediate access to a paired client device.

[0092] The wearable device 104 further includes a wearable signal quality system 210, which is designed to monitor and increase the integrity of the EEG signals collected by the wearable EEG electrode 202. The wearable signal quality system 210 may assess amplitude and frequency characteristics, such as maintaining amplitude below 100 mV and focusing on the expected sleep frequency range (e.g., 0.01-35 Hz). The wearable signal quality system 210 may automatically detect and flag invalid epochs or segments contaminated by artifacts, which are then excluded from downstream analysis. By maintaining high signal quality, the wearable signal quality system 210 supports accurate sleep staging and reliable diagnostic outputs.

[0093] The wearable device 104 further includes a wearable battery device 212, which supplies power to all components of the wearable device 104. Battery options may include prismatic lithium batteries, rechargeable Li-Polymer or Li-Ion batteries, Nickel-Metal Hydride (NiMH) batteries, Lithium-Iron Phosphate (LiFePO4) batteries, or Zinc-Air batteries. In some embodiments, the wearable battery device 212 is a lightweight, high-energy-density coin lithium battery with a long cycle life, providing at least 8-24 hours of continuous monitoring. In some embodiments, the wearable battery device 212 provides at least 36 hours of continuous monitoring. In some embodiments, the wearable battery device 212 provides at least 48 hours of continuous monitoring. In some embodiments, the wearable battery deviceAttorney Docket No.: 46528.6 (L0002PCT)212 provides at least 72 hours of continuous monitoring. In some embodiments, a supercapacitor may be included to enable rapid charging and handle high power bursts, further extending battery life and device autonomy.

[0094] The wearable device 104 may further include firmware 214, which is embedded software that controls the operation of the wearable device 104, including hardware control, data processing, communication, energy management, user interface, updates, and security. The firmware 214 may support real-time sensor data processing and wireless communication on the same chip, with ultra-low-power consumption and hardware-based security features such as encryption and secure boot. Updates for the firmware 214 may be delivered over-the- air (OTA) to ensure the device remains up-to-date with the latest features and security enhancements.

[0095] The wearable device 104 may further include a printed circuit board (PCB) 216, which integrates the electronic components of the wearable device 104, including the wearable processor unit 204, the wearable communication system 206, the wearable storage device 208, and connectors forthe wearable EEG electrode 202. The PCB 216 may be rigid or flexible, and may be protected by a case made of plastic or other suitable materials. In some embodiments, the PCB 216 is designed with multiple layers (e.g., more than four) to optimize space and signal routing, and maybe connected to the wearable EEG electrode 202 via conductive wires, pogo-pins, threads, adhesives, or magnetic connectors.

[0096] The wearable device 104 may further include an on / off switch 218, which provides user control over the power state of the wearable device 104. The on / off switch 218 may be implemented as a slide switch, tactile push button, rocker switch, toggle switch, or capacitive touch switch, and may include visual indicators such as LEDs to display power and Bluetooth® status. The on / off switch 218 may enable a user to activate or deactivate the device, conserving battery life when notin use. The on / off switch 218 may enable a user to activate the sleep monitoring independently from the client device 120.

[0097] The wearable device 104 may optionally include or interface with additional biometric sensor(s) 220, such as photoplethy smogram (PPG) sensors, accelerometers, electrocardiogram (ECG) sensors, oximeters, or other wearable devices (e.g., rings, smartwatches, CPAP masks). The additional biometric sensor(s) 220 may collect biometric data, including heart rate, oxygen saturation, movement, temperature, or respiratory effort, which can be integrated with EEG sensor data to enhance sleep diagnostics and therapy personalization. In some embodiments, the additional biometric sensor(s) 220 are external toAttorney Docket No.: 46528.6 (L0002PCT) the wearable device 104 and communicate via wireless protocols or application programming interfaces (APIs).

[0098] The wearable device 104 may optionally include a user feedback system 222 integrated into the headband 232 to deliver periodic cues using one or more modalities such as LEDs, haptics, and audio. LEDs may be embedded along the forehead or temple regions of the headband and driven by the firmware 214 to display status indicators (e.g., power, recording, signal quality) and event-specific patterns or colors (e.g., a gentle pulsing light to prompt repositioning when poor electrode contact is detected by the wearable signal quality system 210). A compact vibration motor may be housed within the headband casing to deliver silent haptic alerts, such as escalating vibration patterns used as a smart alarm or to rouse the user when disordered sleep patterns or events are detected by the first trained machine learning model 190 (e.g., excessive fragmentation, prolonged WASO) or when therapy prompts from the second trained machine learning model 195 are issued. In some embodiments, a low-profile piezoelectric buzzer or micro-speaker may produce soft audio tones for wake cues or adherence prompts, with intensity and schedules configurable through the patient-facing computer program executed by control logic 122. The feedback modalities may be user-selectable, time-windowed (e.g., quiet hours), and mapped to distinct alert profiles so that notifications and reminders generated by the therapy algorithm are delivered in a manner suited to the user’s preferences and clinical protocol.

[0099] FIGS. 2B-C illustrate views of the wearable device 104 positioned on a user’s head 230, according to some embodiments. FIG. 2B is a front perspective view of the wearable device 104 secured to the user’s head 230 via a headband 232. FIG. 2C is a side perspective view of the wearable device 104 secured to the user’s head 230 via the headband 232. In some embodiments, In some embodiments, the wearable device 104 is secured to the user’s head 230 using alternative configurations beyond a headband 232. For example, the wearable device 104 may be integrated into a hat, cap, or visor that comfortably fits over the scalp, or may be incorporated into a headset or helmet structure for more stable placement during movement. In other embodiments, the device may be attached using adhesive patches or flexible strips that adhere directly to the skin, or may be embedded in a soft, adjustable fabric wrap or bandana. The design may also include adjustable straps, clips, or magnetic fasteners to accommodate different head sizes and shapes, ensuring secure and reliable electrode contact for accurate EEG signal acquisition during sleep or daily activities.[000100]FIGS. 2D-E illustrate perspective views of the wearable device 104, according to some embodiments. FIG. 2D is a front-side perspective view of the wearable device 104Attorney Docket No.: 46528.6 (L0002PCT) attached to the headband 232. FIG. 2E is a read-side perspective view of the wearable device 104 attached to the headband 232. In some embodiments, as seen in FIG. 2D, the wearable device 104 includes one or more LEDs 234 that indicate a status of the wearable battery device and / or Bluetooth® connection 212. As seen in FIG. 2E, the wearable device 104 may include one or more wearable EEG electrodes 202. In some embodiments, the wearable device 104 includes a single wearable EEG electrode 202. In some embodiments, the wearable device 104 includes two or more wearable EEG electrodes 202, a signal electrode 242, a ground electrode 246, and reference electrode 244.[OOOlOlJThe wearable EEG electrode 202 may be placed around the head and may be connected to a detachable printed circuit board (e.g., the PCB 216) on the user’s head 230 (e.g., at the forehead). Additional biometric sensor(s) 220 may also be worn and / or used by the user, such as an oximetry ring worn on the finger. In some embodiments, the wearable device 104 is connected to the client device (e.g., the client device 120 of FIG. 1A, executing a patient-facing computer program), by the wearable communication system 206. The wearable device 104 may collect the EEG sensor data and send it to the client device 120 and / or a server (e.g., the server machine 170, the backend system) to be treated by the algorithms (e.g., the first trained machine leamingmodel 190 ofFIG. 1 A, the second trained machine leamingmodel 195 ofFIG. IB) and stored in a data store (e.g., the data store 140 of FIG. IB, a data server). The server may send the treated data (e.g., the first output 192 of the first trained machine learning model 190, the second output 197 of the second trained machine leamingmodel 195) backto the client device 120 and / or a second client device (e.g, the second client device 125 of FIG. IB executing a clinician -facing computer program), where the output is shown to the user (e.g., a patient) and / or a clinician via a user interface in the form of text, graphs, downloadable reports, personalized orientations for treatment and diagnostics of sleep disorders.[000102]In some embodiments, the headband 232 of the wearable device 104 is made of a flexible fabric-based single-channel headband, with a detachable PCB 216 placed on the head 230 of the user (e.g., on the forehead of the user). The fabric-based headband 232 may be disposable and connected via a magnet connector or pogo-pins to the PCB 216. The PCB 216 may be placed in a rigid case. The fabric of the headband 232 may be elastic and may have different colors and patterns.[000103]The wearable device 104, in some embodiments, stores several nights of readings and does not need to be connected to the client device 120 in real-time and / or during sleep monitoring. In some embodiments, the wearable device 104 is used as a standalone sensor forAttorney Docket No.: 46528.6 (L0002PCT) continuous home monitoring of EEG sensor data in combination with a medication diary from the patient-facing computer program executed on the client device 120 in orderto detect efficacy and safety of medication.[000104]In some embodiments, the wearable device 104 is made of non-disposable materials. The PCB 216 may be made of a flexible material, which reduces the size of the protective case used to protect the PCB 216 and enables the wearable device 104 to be lighter, flexible, and more comfortable.[000105]In some embodiments, the wearable device 104 is used in combination with additional biometric sensor(s) 220 such as a CPAP mask to capture objective data associated with obstructive sleep apnea, such as sleep fragmentation, hypopnea index, etc.[000106]In some embodiments, the wearable device 104 is used to collect data (e.g., EEG sensor data) in order to personalize features such as the Cognitive Behavioral Therapy protocol (e.g., sleep therapy data) of the patient-facing computer program executed on the client device 120, such as providing the exact sleep / awake time for personalized “Sleep Restriction module” guidelines to improve sleep efficiency, providing the best time for going to bed in the “Stimulus Control module”, providing accurate information about causes of awakenings during the night in order to personalize the orientation from “Stimulus Control Module”, or monitoring electrocardiogram (ECG) and stress in order to adjust the techniques associated with anxiety and personalize the “Paradoxical Intention Module” and “Cognitive Restructuring Module”, among others.[000107JFIG. 3 is a block diagram illustrating a methodology 300 of the first trained machine learning model 190 (e.g., the sensor monitoring algorithm), according to certain embodiments. Components and features of FIG. 1 A, IB, 2A, 2B, 2C, 2D, 2E, and / or 3 that have the same or similar name and / or reference number may have the same or similar functionality, material, components, etc.[000108]The wearable EEG electrode 202 may collect the EEG sensor data 102, and the control logic 122 may provide this EEG sensor data 102 to the first trained machine learning model 190 as the first data input. This EEG sensor data 102 may be processed in several phases, including a data pre-processing phase310, a data segmentation phase 320, an extraction of features phase 330, a classification model phase 340 (where the first trained machine learning model 190 is developed and stored in the server machine 170), and an extraction of output parameters phase 350.[000109]The first trained machine learning model 190 may be implemented, during the classification model phase 340, as a data classification model that leverages a range ofAttomey Docket No.: 46528.6 (L0002PCT) machine learning and / or deep learning techniques to analyze and interpret the EEG sensor data 102. In some embodiments, the model utilizes supervised learning algorithms, such as regression or classification methods, including linear regression, logistic regression, support vector machines, decision trees, or ensemble approaches like random forests and gradient boosting, to classify sleep stages or detect relevant events. Additionally, the control logic 122 may employ deep learning architectures, such as convolutional neural networks (CNNs) for extracting spatial features or image-based analysis, and recurrent neural networks (RNNs) for modeling sequential or time-series aspects of the EEG data. The system may further incorporate advanced techniques such as transfer learning to leverage pre-trained models for improved performance, and reinforcement learning to adaptively optimize classification strategies based on feedback to increase the adaptability and efficiency of the data classification model.[OOOllOJIn some embodiments, the first trained machine learning model 190 may execute one or more operations on the EEG sensor data 102 including the data pre-processing phase 310 and the data segmentation phase 320 to prepare the EEG sensor data 102 for analysis. The EEG sensor data 102 may be prepared in the data pre-processing phase 310 using a variety of techniques, including cleaning, filtering, normalizing, and / or standardizing the EEG sensor data 102. Cleaning may include handling missing values by identifying, interpolating, and / or removing segments of the EEG sensor data 102 where signal dropout or loss has occurred. Cleaning may further include correcting inconsistencies of the EEG sensor data 102 by detecting and / or mitigating artifacts such as abrupt signal spikes, baseline drifts, and / or noise introduced by user movement, poor electrode contact, or other such factors. [OOOlllJIn some embodiments, the EEG sensor data 102 is further prepared the data preprocessingphase 310 by filtering, which is configured to preserve sleep-relevant frequency content while suppressing noise and motion artifacts. Filtering may include band-pass filtering (e.g., 0.3-35 Hz) using finite impulse response (FIR) or infinite impulse response (IIR) implementations such as Butterworth or Chebyshev filters. Gaussian or median filters may be applied in the time domain to reduce high-amplitude spikes without distorting phase. In some embodiments, a Kalman filter may be used to perform adaptive noise suppression where noise statistics vary over time, improving downstream spectral feature reliability. [000112]In some embodiments, the EEG sensor data 102 is further prepared the data preprocessing phase 310 by normalizing and / or standardizing to provide consistent scaling across channels and nights. Normalizing may include encoding categorical annotations or labels associated with the EEG sensor data 102 (e.g., sleep stage labels, artifact flags, deviceAttorney Docket No.: 46528.6 (L0002PCT) configuration states, etc.) using methods to convert categorical data into a numerical format (e.g., one-hot or label encoding) so that non-numeric metadata can be interpreted alongside numeric features. Standardizing may include rescaling continuous features to zero mean and unit variance (e.g., z-score per-channel or per-night) or applying min-max scalingto bounded physiological features to improve numerical stability and convergence of the first trained machine learning model 190.[000113]In some embodiments, the control logic 122 applies dimensionality reduction to simplify datasets while retaining clinically informative variance. Dimensionality reduction may include principal component analysis (PCA) to project feature vectors onto a lowerdimensional sub space that captures dominant variance, and / or feature selection (e.g., mutual information, recursive feature elimination, or model-based importance) to remove redundant orweakly informative features. In some embodiments, transformations such as logarithmic scaling are applied to skewed features (e.g., band power, spectral entropy) to stabilize variance and approximate normality to improve classifier calibration and robustness. [000114]In some embodiments, the processing device addresses dataset imbalance using class balancing techniques. Class balancing may include oversampling minority classes (e.g., REM or N1 epochs) using random duplication or synthetic methods (e.g., SMOTE or time- series-aware variants) to reduce bias toward majority classes. Class balancing may additionally or alternatively include under sampling majority classes (e.g., N2 epoch) so that misclassification penalties reflect clinical priorities for sleep staging and arousal detection. [000115]In some embodiments, the first trained machine learning model 190 may be trained by comparing the EEG sensor data 102 collected by the wearable EEG electrode 202 to one or more exams (e.g., a polysomnography exam, the Golden Standard Polysomnography exam), where the labeled signal from the polysomnography exam was used to teach the EEG sensor data (e.g., the simultaneous signal) from the wearable device 104. The signal from the wearable EEG electrode 202 may be prepared, cleaned and filtered using IIR (Infinite Impulse Response) filters, such as Butterworth. The signal may also normalized by going through under sampling, and one or more (e.g., 17) statistical and / or amplitude-based features were extracted. The first trained machine learning model 190 may implement a Bagged Decision Trees technique as a classification model. In some embodiments, an overall accuracy of about 80% was achieved for sleep staging classification, such as REM, Nl, N2, N3 , and Awake stages. In some embodiments, the output of the first trained machine learning model 190 (e.g., the sleep quality data 152 of FIG. IB) may be displayed via the user interface of the client device 120 as a graph.Attorney Docket No.: 46528.6 (L0002PCT)[000116]Following the data pre-processing phase 310, the EEG sensor data 102 may enter the data segmentation phase 320. During the data segmentation phase 320, the EEG sensor data 102 may be divided into discrete, manageable segments or epochs to facilitate downstream analysis and classification. In some embodiments, the control logic 122 segments the continuous EEG signal into fixed-length windows, such as 30-second epochs, which are commonly used in clinical sleep staging protocols. This segmentation enables the system to align with established standards for sleep analysis and to associate each segment with specific sleep stages or events. The segmentation process may also include overlap between windows to capture transitional features and improve temporal resolution. By structuring the data into consistent segments, the system ensures that subsequent feature extraction and classification steps are both efficient and clinically meaningful.[000117]Followingthe data segmentation phase 320, the EEG sensor data 102 may enter the extraction of featuresphase 330. In the extraction of features phase 330, the segmented EEG data may be transformed into a set of quantitative descriptors that capture relevant physiological and / or statistical characteristics. The control logic 122 may extract a variety of features from each segment, including statistical features (e.g., mean, variance, skewness), amplitude-based features (e.g., peak-to-peak amplitude, signal energy), and frequencydomain features derived from Fourier or wavelet transforms (e.g., band power in delta, theta, alpha, and beta ranges). Additional features may include entropy-based measures to assess signal complexity, as well as time-dependent characteristics such as trends, seasonality, or lag features. In some embodiments, principal component analysis (PCA) or other dimensionality reduction techniques are applied to the extracted features to reduce redundancy and focus on the most informative components. This comprehensive feature set provides the foundation for accurate and robust classification of sleep stages and events.[000118]Followingthe extraction of features phase 330, the EEG sensor data 102 may enter the classification model phase 340. The classification model phase 340 involves applying a trained machine learning or deep learning model (e.g., the first trained machine learning model 190, the sensor monitoring algorithm) to the extracted features in order to assign each data segment to a specific sleep stage or to detect relevant sleep-related events. The first trained machine learning model 190 may utilize supervised learning algorithms, such as decision trees, support vector machines, logistic regression, or ensemble methods like random forests and gradient boosting, as well as deep learning architectures including convolutional neural networks (CNNs) and recurrent neural networks (RNNs). In some embodiments, the model is trained using labeled data from gold-standard polysomnography exams, enablingAttorney Docket No.: 46528.6 (L0002PCT) high accuracy in sleep staging (e.g., exceeding 80%, 90%, or even 95% as compared to type 1 polysomnography). The classification model phase 340 may also incorporate transfer learning or reinforcement learning to adapt to new data and improve performance over time. The resulting sleep stage labels or event detections are then used to generate output parameters for sleep quality assessment and therapy personalization, as will be described in further detail.[000119] After processing and classification, the first trained machine learning model 190 is configured to extract a range of output parameters in the extraction of output parameters phase 350 from the EEG sensor data 102 to provide both qualitative and quantitative information relevant to sleep improvement and diagnostics. These output parameters may include, but are not limited to, sleep stage percentages and total time spent in each stage, sleep schedules, sleep efficiency, sleep onset latency, sleep stage transitions, REM latency, wake after sleep onset (WASO), and overall sleep durations. By quantifying these metrics, the system enables comprehensive assessment of sleep architecture and patterns, supporting both clinical and self-guided evaluation of sleep quality.[000120]Following the processing and analysis of the EEG sensor data 102 by the first trained machine learning model 190, the system utilizes the resulting sleep quality data 152, in combination with additional user input, to generate personalized sleep therapy recommendations (e.g., sleep therapy data) through the second trained machine learning model 195, as depicted in FIG. 4.[000121JFIG. 4 is a block diagram illustrating a methodology 400 of the second trained machine learning model 195 (e.g., a therapy algorithm), according to certain embodiments. Components and features of FIG. 1A, IB, 2A, 2B, 2C, 2D, 2E, 3, and / or 4 that have the same or similar name and / or reference number may have the same or similar functionality, material, components, etc.[000122]The subjective and objective data (e.g., the sleep quality data 152 of FIG. IB, the user input data 162 of FIG. IB) collected from a client device (e.g., the client device 120 of FIG. 1 A, including a processing device to execute control logic 122, to execute a patientfacing computer program) and the wearable EEG electrode 202 (e.g., a wearable sensor electrode) are selected according to combinations, then the limits, values and calculations are performed, which then goes to a classification model to generate the outcomes. In some embodiments, the second trained machine learning model 195 is configured to generate sleep therapy data by processing a combination of user input data 162 and sleep quality data 152. The second trained machine learning model 195 may include various modules, including aAttorney Docket No.: 46528.6 (L0002PCT) definition of variable combinations module 420, a definition of variables limits and calculations modules 430, a classification model module 440, and a definitions of outcomes module 450. The second trained machine learning model 195 may enable the system to deliver adaptive, data-driven therapy recommendations tailored to the unique targets and circumstances of each user.[000123]In some embodiments, the user input data 162 provided to the second trained machine learning model 195 includes user input 410 and / or biometric data 412. User input 410 may include subjective information collected via the client device 120 (e.g., executing control logic, executing a patient-facing computer program), such as responses to sleep diaries, behavioral questionnaires, self-reported symptoms, or feedback on therapy adherence and perceived sleep quality. The user input 410 may include user input associated with the user, where the user input is received via a user interface (e.g., of the client device 120). The user interface may be an application accessed via the client device 120. Biometric data 412 may include objective physiological measurements obtained from the wearable device 104 and / or additional biometric sensors (e.g., the additional biometric sensor(s) 220 of FIG. 2A), such as EEG-derived sleep metrics, heart rate, oxygen saturation, movement, or other relevant health indicators. The biometric data 412 may be associated with the user and may be different from the EEG sensor data 102. In some embodiments, the biometric data 412 includes photoplethy smogram (PPG) sensor data, electrocardiogram (ECG) sensor data, electromyography (EMG) sensor data, electrooculogram (EOG) sensor data, accelerometer sensor data, etc. By integrating both subjective and objective data, the system enables a comprehensive and personalized approach to sleep therapy, supporting both self-guided and clinician-directed care.[000124]The definition of variable combinations module 420 involves selecting and combining relevant input variables from the user input data 162 and sleep quality data 152 to inform the second trained machine learning model 195 (e.g., the therapy algorithm). These variables may include, but are not limited to, sleep efficiency, sleep onset latency, wake after sleep onset (WASO), responses to sleep questionnaires, oxygen desaturation index, sleep duration, REM sleep latency, sleep medication use, sleep stage transitions, and sleep staging percentages. The system may use domain knowledge, clinical guidelines, or data-driven insights to determine which combinations of variables are most predictive or actionable for therapy personalization. In some embodiments, the system dynamically adjusts the selected variable combinations based on ongoing monitoring and user feedback to optimize therapy outcomes.Attorney Docket No.: 46528.6 (L0002PCT)[000125]The definition of variables limits and calculations module 430 refers to establishing threshold values, ranges, or calculation rules for each selected variable or combination of variables. For example, the system may define clinically relevant cutoffs for sleep efficiency (e.g., below 85% indicating poor sleep), or set limits for acceptable sleep onset latency or WASO. Calculations may include aggregating nightly metrics over a monitoring period (e.g., 3 days, 7 days, 30 days, etc.), computing averages, trends, or deviations, and normalizing or standardizing variables to establish comparability across users and sessions. These limits and calculations provide the basis for interpreting input data and triggering specific therapy recommendations to accommodate both short-term changes and long-term trends in the user’s sleep health.[000126]The classification model module 440 within the second trained machine learning model 195 applies machine learning or deep learning techniques to the processed input data in order to classify the user’s sleep health status and determine corresponding sleep therapy data (e.g., therapy interventions). The classification model module 440 may utilize supervised learning algorithms, such as decision trees, logistic regression, support vector machines, or ensemble methods, as well as deep learning architectures like neural networks. In some embodiments, the model is trained on historical data to recognize patterns associated with different sleep disorders or therapy responses, and may incorporate transfer learning or reinforcement learning to adapt recommendations overtime. The classification model module 440 may also be periodically retrained or updated based on subsequent EEG sensor data and / or subsequent user input data to enable sleep therapy data (e.g., sleep therapy recommendations) that are current and effective (e.g., produce a targeted effect).[000127] The definition of outcomes module 450 involves mapping the classified user state and calculated variables to specific therapy outputs (e.g., sleep therapy data). Outcomes may include quantitative and qualitative feedback, such as personalized sleep schedules, behavioral modification strategies, cognitive behavioral therapy (CBT) modules, reminders, notifications, or multimedia educational content. The system may also generate reports or recommendations for clinicians, supporting both self-guided and clinician-directed therapy pathways. In some embodiments, the outcomes are presented to the user via the patientfacing computer program as text, graphs, videos, notifications, or reminders, and may be further customized based on user preferences or clinical input. By continuously updating sleep therapy data based on subsequent EEG sensor data and / or subsequent user input data, the second trained machine learning model 195 (e.g., therapy algorithm) provides adaptive, data-driven improvement in sleep health and therapy efficacy.Attorney Docket No.: 46528.6 (L0002PCT)[000128] The second trained machine learning model 195 may generate sleep therapy data by converting input data (e.g., the sleep quality data and user input data) into output recommendations through a process that involves defining variable combinations and limits, classifying data using machine learning classification models, and determining appropriate outcomes. The model may consider a combination of variables, including sleep efficiency, sleep onset latency, wake after sleep onset (WASO), responses to sleep questionnaires, oxygen desaturation index, sleep duration, REM sleep latency, sleep medication use, sleep stage transitions, sleep staging percentages, etc. By analyzing these variables, the system can tailor the sleep therapy data (e.g., sleep therapy recommendations) to the user’s unique sleep profile.[000129]In some embodiments, the sleep therapy data (e.g., output of the second trained machine learning model 195) can be used for providing information to diagnostic decisions or treatment for sleep disorders, including but not limited to insomnia, obstructive sleep apnea, narcolepsy, REM latency disorder, COMISA diagnostic, intraoral device treatment, and medication treatment. In some embodiments, the sleep therapy data includes personalized recommendations. For example, if a user’s percentage of N1 is greater than 5% (obtained from the wearable EEG electrode 202), and a user’s Oxygen Desaturation Index is greater than 15 (obtained from an additional biometric sensor 220, such as an oximetry sensor), the following message is generated and presented to the user via the user interface:The percentage of N1 sleep is increased. N1 stage is a lighter and transitional phase of sleep; when it is elevated, it may indicate that the sleep is lighter, more superficial, or fragmented. There was a high desaturation index, requiring further evaluation for sleep apnea.[000130]Following the generation of sleep therapy data by the second trained machine learning model 195 as illustrated in FIG. 4, the system architecture depicted in FIG. 5 demonstrates the flow of data and control among various components responsible for collecting, processing, storing, and delivering personalized sleep therapy recommendations to the user and / or clinician.[000131JFIGS. 5 A-B are block diagrams illustrating the details of a processing device executing control logic 122, according to certain embodiments. FIG. 5 A is a block diagram 500A illustrating the processing device (e.g., the client device 120 of FIG. 1A) executing the control logic 122 (e.g., a patient facing computer program) and providing the output 197 toAttorney Docket No.: 46528.6 (L0002PCT) the client device 120, according to some embodiments. FIG. 5B is a block diagram 500B illustrating the processing device executing control logic 122 and providing the output 197 to the second client device 125 (e.g., executing a clinician facing computer program), according to some embodiments. Components and features of FIG. 1A, IB, 2A, 2B, 2C, 2D, 2E, 3, 4, 5 A, and / or 5B that have the same or similar name and / or reference number may have the same or similar functionality, material, components, etc.[000132]In some embodiments, input (e.g., user input 410, biometric data 412, input from third-party API 510) is obtained from the processing device. The input may be transmitted to the server machine 170 (e.g., a backend system), where it is treated by the algorithms (e.g., the first trained machine learning model 190, the second trained machine learning model 195) and stored in the data store 140 (e.g., a data server) as data. The data may then be transferred to the processing device as an output 197 to the user.[000133]In some embodiments, the control logic 122, which may be implemented on the client device 120 to execute a patient-facing computer program, orchestrates the collection and processing of user data, the interaction with machine learning models, and the presentation of outputs to the user and / or clinician. The control logic 122 is may receive user input 410, which may include subjective data such as responses to sleep diaries, behavioral questionnaires, and self-reported symptoms, as well as objective data from the wearable EEG sensor 202 and / or the additional biometric sensors 220 of FIG. 2A.[000134]In addition to direct user input (e.g., user input 410 and / or biometric data 412), the system may receive input from one or more third-party application programming interface(s) (API(s)) 510. This input may include data from external health or wellness platforms, additional wearable devices, or cloud-based services that provide complementary biometric orbehavioral information. The integration of third-party data sources enables the system to provide analysis and recommendations with a broader context of the user’s health and lifestyle, supporting more comprehensive and personalized sleep therapy data.[000135]The server machine 170, which may function as a backend system and data server, is responsible for executing the core machine learning algorithms and managing data storage and retrieval. The server machine 170 receives processed data from the control logic 122 and coordinates the operation of the first trained machine learning model 190 and the second trained machine learning model 195. The first trained machine learning model 190 analyzes EEG sensor data to generate sleep quality data, while the second trained machine learning model 195 processes this sleep quality data in combination with user input and third-party data to generate sleep therapy data.Attorney Docket No.: 46528.6 (L0002PCT)[000136]The data store 140 may serve as a centralized repository for all relevant data, including raw and processed sensor data, user input, third-party API data, intermediate outputs from the machine learning models, and final therapy recommendations. The data store 140 may provide data integrity, support longitudinal analysis, and enable the system to track user progress and adapt sleep therapy data over time.[000137]The second output 197, which represents the sleep therapy data generated by the second trained machine learning model 195, is delivered back to the control logic 122 for presentation to the user or clinician. In some embodiments, the second output 197 is presented to a uservia the client device 120, as seen in FIG. 5A. In some embodiments, the second output 197 is presented to a clinician via the second client device 125, as seen in FIG. 5B. The second output 197 (e.g., sleep therapy data) may include personalized recommendations, feedback, therapy modules, notifications, and reports, which can be accessed via the patient-facing or clinician-facing computer programs (e.g., the control logic 122, executed by a processing device). The system’s architecture, as illustrated in FIG. 5, enables integration of user data, analytics, and adaptive therapy delivery, supporting both self-guided and clinician-supported pathways for sleep improvement and disorder management.[000138JFIGS. 6A-B illustrate graphical user interfaces 600A-B associated with receiving data inputfor diaries and / or tests, according to certain embodiments. FIG. 6 A illustrates the GUI 600 A configured to receive input associated with diaries 612, according to some embodiments. FIG. 6B illustrates the GUI 600B configured to receive input associated with tests 622, according to some embodiments. In some embodiments, the GUIs 600 A-B include a graphical user element (e.g., button) for accessing diaries, tests, questionnaires, and / or anamneses. In some embodiments, the GUIs 600A-B are configured to present sleep-related information, collect user input, support engagement with sleep therapy data, etc. Components and features of FIG. 1 A, IB, 2 A, 2B, 2C, 2D, 2E, 3, 4, 5 A, 5B, 6A, and / or 6B that have the same or similar name and / or reference number may have the same or similar functionality, material, components, etc.[000139]The diary interface 610 provides users with access to various sleep diaries 612, which are digital forms or modules for recording daily sleep patterns, habits, and experiences. Sleep diaries 612 may include fields for bedtime, wake time, perceived sleep quality, nighttime awakenings, and other relevant metrics. The diary description 614 may offer contextual information or instructions to guide users in completing their entries, such as the importance of consistent tracking or tips for accurate reporting.Attorney Docket No.: 46528.6 (L0002PCT)[000140]A calendar 616 is integrated into the GUI 600 A to help users visualize and navigate their diary entries over time. The calendar 616 may highlight days with completed entries, missed entries, or upcoming prompts, supporting longitudinal tracking of sleep behavior. Diary status indicators 618 may provide visual cues regarding the completion and / or update status of each diary entry, such as icons or color codes indicating whether a diary is filled, pending, or overdue.[000141]The test interface 620 represented in GUI 600B may enable users to access and complete various sleep tests 622, which may include standardized assessments, questionnaires, and / or clinical scales relevant to sleep health. Sleep tests 622 may evaluate factors such as sleep quality, insomnia severity, dysfunctional beliefs about sleep, or risk of sleep apnea. The test interface 620 may present instructions, progress indicators, and links to results or feedback, supporting both initial assessment and ongoing monitoring.[000142]Collectively, the GUIs 600A-B illustrated in FIGS. 6A-B support user engagement with the control logic 122 (e.g., patient-facing computer program) by providing intuitive access to diaries, tests, and sleep-related information. These interfaces facilitate the collection of subjective data, support therapy adherence, and enable users to monitor their progress over time.[000143JFIG 7 illustrates a GUI 700 associated with establishing a connection between the client device 120 and a biometric sensor (e.g., a wearable communication system), according to certain embodiments. In some embodiments, the GUI 700 is an interface screen (e.g., a user interface) from the patient-facing computer program that connects with the biometric sensor (e.g., wirelessly, such as by Bluetooth®). Components and features of FIG. 1A, IB, 2 A, 2B, 2C, 2D, 2E, 3, 4, 5 A, 5B, 6 A, 6B, and / or 7 that have the same or similar name and / or reference number may have the same or similar functionality, material, components, etc. [000144]In some embodiments, the GUI 700 is executed on the client device 120, which may be a smartphone, tablet, or other computing device running control logic 122 (e.g., the patient-facing computer program). The GUI 700 may be configured to guide the user through a process of connecting the client device 120 and the control logic 122 with the biometric sensor. The biometric sensor may include a wearable EEG sensor (e.g., the wearable EEG sensor 202 of FIG. 2A) of wearable device (e.g., the wearable device 104 of FIG. 1A) or one or more additional biometric sensors (e.g., the additional biometric sensor(s) 220 of FIG. 2A), such as an oximeter, heart rate monitor, or other physiological monitoring device.Attorney Docket No.: 46528.6 (L0002PCT)[000145]A back button 710 is provided within the GUI 700, allowing the user to navigate to a previous screen or menu. This feature may provide intuitive navigation and user control throughout the setup process.[000146]The instructions for connecting to the biometric sensor 720 may be displayed within the GUI 700. These instructions may include step-by-step guidance, such as turning on the biometric sensor, activating Bluetooth® and location services on the client device 120, and keeping the device in close proximity to the sensor to facilitate pairing and data transfer. The instructions may enable the connection between the biometric sensor and the client device 120, such that the client device 120 may receive data(e.g., EEG sensor data 102 of FIG. 1A) from the biometric sensor. The data received by the client device may be utilized by various algorithms (e.g., the first trained machine learning model 190, the second trained machine learning model 195) and integrated with the sleep monitoring and therapy system to provide sleep therapy data to the user. The GUI 700 supports the integration of biometric data from a variety of sensors, enabling the system to collect physiological information for use in sleep analysis, therapy personalization, and ongoing monitoring.[000147JFIGS. 8A-B illustrate GUIs 800 A-B associated with providing an output (e.g., sleep therapy data 197 of FIG. 1A, data output), according to certain embodiments. FIG. 8A illustrates the GUI 800A associated with providing an instruction pane 840, according to some embodiments. FIG. 8B illustrates the GUI 800B associated with providing a results pane 850, according to some embodiments. Components and features of FIG. 1A, IB, 2A, 2B, 2C, 2D, 2E, 3, 4, 5A, 5B, 6A, 6B, 7, 8A, and / or 8B that have the same or similar name and / or reference number may have the same or similar functionality, material, components, etc.[000148] A trash button 810 is provided within the GUIs 800 A-B, enabling the user to delete a particular reading or session (e.g., from the client device 120, from the control logic 122, from the patient facing computer program, etc.). This feature may enable user control over stored data and may facilitate management of historical records.[000149]The reading details section 820 displays information about a specific sleep monitoring session or test, such as the date and time of the reading, duration, and reference number. This section provides context for the user to identify and review individual sessions. [000150]Various reading sections 830 may be included to organize and present different aspects of the recorded data. These sections can display metrics such as sleep stage distribution, sleep efficiency, or other relevant parameters derived from the analysis of EEG and biometric sensor data.Attorney Docket No.: 46528.6 (L0002PCT)[000151] An instructions pane 840 is included in the GUIs 800A-B, which includes detailed guidelines 842. The instructions pane 840 may provide the user with personalized recommendations, explanations of results, or next stepsbased on the data collected during the session. Detailed guidelines 842 may be generated by a therapy algorithm (e.g., the second trained machine learning model 195) and tailored to the user’s specific sleep profile and therapy plan. In some embodiments, the detailed guidelines 842 are in the form of text, and correspond to an output of algorithms (e.g., the first trained machine learning model 190, the second trained machine learning model 195). The detailed guidelines 842 may be presented as personalized orientations that consider the quantitative and qualitative input data from diaries, questionnaires, anamneses, and wearable sensor systems.[000152]The results pane 850 may present the outcome of the sleep monitoring session or test. This pane may include test results 852, such as scores from sleep quality assessments, apnea indices, or other diagnostic metrics. The results pane 850 may also display detailed guidelines 854, offering further interpretation of the results and actionable advice for the user. In some embodiments, the detailed guidelines 854 are in the form of text, and correspond to an output of algorithms (e.g., the first trained machine learning model 190, the second trained machine learning model 195). The detailed guidelines 854 maybe presented as personalized orientations that consider the quantitative and qualitative input data from diaries, questionnaires, anamneses, and wearable sensor systems.[000153]In some embodiments, the results pane 850 includes historical reports that include data and / or metrics for a previous three-day period, a previous seven-day period, a previous 14-day period, a pervious 30-day period, a previous 60-day period, a previous 90-day period, a previous 180-day period, a previous 365-day period, or any other subset of these periods. Collectively, the GUIs 800A-B illustrated in FIGS. 8A-B support the delivery of comprehensive sleep therapy output and data output, enabling users to review their progress, understand their sleep health, and access personalized recommendations.[000154JFIG 9 illustrates a GUI 900 associated with providing output (e.g., reports, graphs) to the second processing device 125 (e.g., executing control logic 122, executing a clinician facing computer program), according to certain embodiments. Reports 912A-Z may be associated with outputs of the algorithms (e.g., the first trained machine learning model 190, the second trained machine learning model 195), which are based on the quantitative and qualitative input data from diaries, questionnaires, anamneses, and biometric sensors (e.g., the wearable device 104, additional biometric sensor(s) 220, the wearable sensor systems). Components and features of FIG. 1A, IB, 2 A, 2B, 2C, 2D, 2E, 3, 4, 5 A, 5B, 6 A, 6B, 7, 8 A,Attorney Docket No.: 46528.6 (L0002PCT)8B, and / or 9 that have the same or similar name and / or reference number may have the same or similar functionality, material, components, etc.[000155]The GUI 900 may be executed on the second client device 125, which may be a desktop computer, laptop, tablet, or mobile device running a clinician-facing computer program or second control logic (e.g., the second control logic 127 of FIG. 1 A), according to certain embodiments. The GUI 900 may present outputs such as reports and graphs to clinicians or other authorized users, supporting the review and management of sleep monitoring and therapy data.[000156]The GUI 900 may include a reports pane 910, which serves as the primary area for displaying various reports 912A-Z. Each report 912A-Z may correspond to a specific patient, monitoring session, or analysis period, and may include detailed information such as sleep stage distributions, sleep efficiency, event indices, therapy adherence, and other clinically relevant metrics. Reports 912A-Z may be presented in both textual and graphical formats, enabling clinicians to quickly interpret trends, compare sessions, and make informed decisions regarding diagnosis or therapy adjustments. In some embodiments, reports 912A-Z may be customizable or downloadable for record-keeping or sharing with other healthcare professionals. In some embodiments, the reports 912A-Z are historical reports that include data and / or metrics for a previous three-day period, a previous seven-day period, a previous 14-day period, a pervious 30-day period, a previous 60-day period, a previous 90-day period, a previous 180-day period, a previous 365-day period, or any other subset of these periods. [000157]The GUI 900 may include a menu bar 920 to facilitate navigation and access to additional features within the clinician -facing computer program. The menu bar 920 may provide options for switching between different patients, accessing historical data, generating new reports, exporting data, or adjusting user preferences. This component supports efficient workflow and enhancesthe usability of the system for clinicians managing multiple users or large datasets.[000158JFIGS. 10A-D are flow diagrams of methods 1000A-D associated with sleep enhancement and diagnostics, according to certain embodiments. In some embodiments, methods 1000A-D are performed by processing logic (e.g., control logic) that includes hardware (e.g., circuitry, dedicated logic, programmable logic, microcode, processing device, etc.), software (such as instructions run on a processing device, a general purpose computer system, or a dedicated machine), firmware, microcode, or a combination thereof. In some embodiments, methods 1000A-D are performed, at least in part, by the predictive system 110 and / or the client device 120 of FIG. IB. In some embodiments, method 1000A is performed,Attorney Docket No.: 46528.6 (L0002PCT) at least in part, by predictive system 110 (e.g., the server machine 170 and the data set generator 172 of FIG. IB). In some embodiments, the predictive system 110 uses method 1000 A to generate a data set to at least one of train, validate, or test a machine learning model. In some embodiments, method 1000B is performed by the client device 120 (e.g., the recommendation component 123). In some embodiments, method 1000C is performed by the server machine 180 (e.g., the training engine 182, etc.). In some embodiments, method OOD is performed by the predictive server 112 (e.g., the predictive component 114). In some embodiments, a non-transitory machine-readable storage medium stores instructions that when executed by a processing device (e.g., of predictive system 110, of server machine 180, of predictive server 112, etc.), cause the processing device to perform one or more of methods 1000A-D.[000159]For simplicity of explanation, methods 1000A-D are depicted and described as a series of operations. However, operations in accordance with this disclosure can occur in various orders and / or concurrently and with other operations not presented and described herein. Furthermore, in some embodiments, not all illustrated operations are performed to implement methods 1000A-D in accordance with the disclosed subject matter. In addition, those skilled in the art will understand and appreciate that methods 1000A-D could alternatively be represented as a series of interrelated states via a state diagram or events. [000160JFIG. 10A is a flow diagram of a method 1000 A for generating a data set for a machine learning model for generating predictive data (e.g., the predictive data 116 of FIG. IB), according to some embodiments.[000161]Referring to FIG. 10A, in some embodiments, at block 1002 the processing logic (e.g., predictive component 114 of predictive server 112 of FIG. IB, the processing device or control logic 122 of FIG. 1 A) implementing method 1000 A initializes a training set T to an empty set.[000162] At block 1004, processing logic generates first data input (e.g., first training input, first validating input) that includes historical EEG sensor data (e.g., the historical EEG sensor data 144 of FIG. IB). In some embodiments, the first data input includes a first set of features for types of EEG sensor data and a second data input includes a second set of features for types of EEG sensor data. In some embodiments, the first data input includes historical polysomnography data. The historical polysomnography data may be the same as the historical EEG sensor data. The historical polysomnography data may be different from the historical EEG sensor data. For example, in some embodiments, the historical polysomnography data is associated with a “gold-standard” polysomnography used to train aAttorney Docket No.: 46528.6 (L0002PCT) machine learning model. In some embodiments, the first data input includes a first set of features for types of polysomnography data and a second data input includes a second set of features for types of polysomnography data.[000163]In some embodiments, at block 1006, processing logic generates a first target output for one or more of the data inputs (e.g., first data input). In some embodiments, the first target output is historical sleep quality data (e.g., the historical sleep quality data 154 of FIG. IB). [000164]In some embodiments, at block 1008, processing logic optionally generates mapping data that is indicative of an input / output mapping. The input / output mapping (or mapping data) refers to the data input (e.g., one or more of the data inputs described herein), the target output for the data input (e.g., where the target output identifies the historical sleep quality data 154), and an association between the data input(s) and the target output.[000165] At block 1010, processing logic adds the data (e.g., the historical EEG sensor data from block 1004, mapping data generated at block 1008) to data set T.[000166] At block 1012, processing logic branches based on whether data set T is sufficient for at least one of training, validating, and / or testing the first machine learning model 190 and / or the second machine learning model 195 (e.g., uncertainty of the trained machine learning model meets a threshold uncertainty). If so, execution proceeds to block 1014, otherwise, execution continues back to block 1004. It should be noted that in some embodiments, the sufficiency of data set T is determined based simply on the number of input / output mappings in the data set, while in some other implementations, the sufficiency of data set T is determined based on one or more other criteria (e.g., a measure of diversity of the data examples, accuracy, etc.) in addition to, or instead of, the number of input / output mappings.[000167] At block 1014, processing logic provides data set T (e.g., to the server machine 180) to train, validate, and / or test the first machine learning model 190 and / or the second machine learning model 195. In some embodiments, data set T is a training set and is provided to the training engine 182 of the server machine 180 to perform the training. In some embodiments, data set T is a validation set and is provided to the validation engine 184 of server machine 180 to perform the validating. In some embodiments, data set T is a testing set and is provided to the testing engine 186 of the server machine 180 to perform the testing. In the case of a neural network, for example, input values of a given input / output mapping (e.g., numerical values associated with data inputs) are input to the neural network, and output values (e.g., numerical values associated with target outputs) of the input / output mapping are stored in the output nodes of the neural network. The connection weights in the neuralAttorney Docket No.: 46528.6 (L0002PCT) network are then adjusted in accordance with a learning algorithm (e.g., back propagation, etc.), and the procedure is repeated for the other input / output mappings in data set T.[000168] After block 1014, a machine learning model (e.g., the first machine learning model 190, the second machine learning model 195) can be at least one of trained using the training engine 182 of the server machine 180, validated using the validation engine 184 of the server machine 180, and / or tested using the testing engine 186 of the server machine 180. The trained machine learning model is implemented by the predictive component 114 (of the predictive server 112) to generate predictive data (e.g., the predictive data 116) for sleep enhancement and diagnostics. In some embodiments, a data set is generated to train a supervised machine learning model (e.g., data set of block 1014 includes data input of block 1004 and target output of block 1006). In some embodiments, a data set is generated to train an unsupervised machine learning model (e.g., clustering, data set of block 1014 includes data input of block 1004 but does not include target output of block 1006).[000169JFIG. 10B is a flow diagram of a method 1000B associated with sleep enhancement and diagnostics, according to some embodiments.[000170]In some embodiments, at block 1020, processing logic identifies EEG sensor data obtained by a wearable device (e.g., the wearable device 104 of FIG. 1A). The wearable device may be associated with a head of a user. In some embodiments, the wearable device includes a wearable EEG sensor (e.g., the wearable EEG sensor 202 of FIG. 2 A) that is held proximate to a forehead of the user via a headband (e.g., the headband 232 of FIG. 2B). [000171]In some embodiments, the wearable device is configured to store the EEG sensor data, and the EEG sensor data is obtained from the wearable device responsive to a connection (e.g., a wireless connection) being establishing with the wearable device. In some embodiments, the wearable device includes a memory configured to store between about 20 hours and about 80 hours of EEG sensor data. In some embodiments, the wearable device is configured to store the EEG sensor data at predetermined time intervals and to transmit the stored EEG sensor data in multiple discrete packets to the processing device (e.g., the client device 120) after a monitoring period has ended.[000172]At block 1022, the processing logic may provide the EEG sensor data as a first data inputto a first trained machine leamingmodel (e.g., the first trained machine learning model 190 of FIG. 1 A, a first set of sensor monitoring algorithms, etc.).[000173] At block 1024, the processing logic may obtain a first output (e.g., the first output 192 of FIG. 1 A) from the first trained machine learning model. The first output may beAttorney Docket No.: 46528.6 (L0002PCT) associated with sleep quality data (e.g., the sleep quality data 152 of FIG. IB). In some embodiments, the sleep quality data is objective EEG data associated with the user.[000174] At block 1026, the processing logic may provide a second input to a second trained machine learning model (e.g., the second trained machine learning model 195 of FIG. 1 A, a second set of therapy algorithms, etc.). The second input may include the sleep quality data of block 1024. The second input may further include user input data (e.g., the user input data 162 of FIG. IB). In some embodiments, the user input data includes user input and / or biometric data, each associated with the user. The user input may be received via a user interface (e.g., via the client device 120) and may include subjective input data from the user related to sleep diaries, anamneses, clinical assessment questionnaires, etc. The biometric data may be different from the EEG sensor data of block 1020 and may be obtained by at least one additional sensor device (e.g., the additional biometric sensor(s) 126 of FIG. IB). [000175] At block 1028, the processing logic may obtain a second output from the second trained machine learning model. The second output may be associated with sleep therapy data (e.g., sleep therapy protocol, sleep therapy recommendations, etc.).[000176]At block 1030, the processing logic may identify subsequent EEG sensor data and / or sub sequent user input data. In some embodiments, this data is obtained subsequent to (e.g., after) obtainingthe second output associated with the sleep therapy data in block 1028. [000177] At block 1032, responsive to determining that the subsequent EEG sensor data and / or the sub sequent user input data meets a first threshold value, the first trained machine learning model and / or the second trained machine learning model may be re-trained based on the subsequentEEG sensor dataand / orthe subsequent user input data. This re-training may generate updated sleep therapy data (e.g., an updated sleep therapy protocol, an updated sleep therapy recommendation, an updated sleep therapy treatment plan, etc.).[000178]In some embodiments, the first threshold value includes a predefined variance between a baseline sleep-quality metric included in the sleep quality data and a corresponding metric derived from the subsequent EEG sensor data and / or the subsequent user input data. The baseline sleep-quality metric may include a sleep efficiency value, a sleep onset latency value, a wake-after-sleep-onset value, a total rapid eye movement (REM) sleep duration value, a total non-rapid eye movement (NREM) sleep duration value, and / or a total sleep time value. In some embodiments, the baseline sleep-quality metric may include other objective sleep measurements.Attorney Docket No.: 46528.6 (L0002PCT)[000179]In some embodiments, at block 1034, the processing logic may receive the subsequent EEG sensor data associated with the user. The subsequent EEG sensor data may be received after providing the sleep therapy data.[000180]In some embodiments, at block 1036, the processing logic may determine, based on the subsequent EEG sensor data, whether the sleep therapy data is producing a targeted effect. In some embodiments, the targeted effect includes improved EEG sensor data and / or improved sleep quality data.[000181JFIG. 10C is a flow diagram of a method 1000C for training a machine learning model (e.g., the first machine learning model 190 of FIG. IB) for determining predictive data (e.g., the predictive data 116 of FIG. IB) associated with sleep quality data, according to some embodiments.[000182] At block 1040 of method 1000C, the processing logic identifies historical EEG sensor data (e.g., the historical EEG sensor data 144 of FIG. IB).[000183] At block 1042, the processing logic identifies historical sleep quality data (e.g., the historical sleep quality data 154 of FIG. IB). In some embodiments, the historical sleep quality data is associated with objective sleep data (e.g., sleep stages, etc.) using the historical EEG sensor data. In some embodiments, the historical sleep quality data is indicative of whether sleep is disordered.[000184] At block 1044, the processing logic trains a classifier machine learning model using data input including historical EEG sensor data and target output including the historical sleep quality data to generate a trained classifier machine learning model (e.g., configured to provide output associated with sleep enhancement and diagnostics). The sleep enhancement and diagnostics may be by using the trained classifier machine learning model of FIG. 10C. In some embodiments, the trained machine learning model is a neural network. In some embodiments, only historical EEG sensor data is identified (e.g., historical sleep quality data is not identified) and the classifier machine learning model is trained using data input including the historical EEG sensor data (e.g., is an unsupervised model, clustering, is not trained using target output).[000185JFIG. 10D is a flow diagram of a method OOD for using a trained machine learning model (e.g., the second machine learning model 195 of FIG. IB) associated with sleep therapy data (e.g., sleep enhancement and diagnostics), according to some embodiments. FIG. 10D may be used for block 1026 of FIG. 10B.[000186]Referring to FIG. 10D, at block 1060 of method WOOD, the processing logic identifies historical sleep quality data, historical user input, and historical biometric data.Attorney Docket No.: 46528.6 (L0002PCT)[000187] At block 1062, the processing logic identifies historical sleep therapy data. In some embodiments, the historical sleep therapy data is associated with subjective and objective sleep data (e.g., sleep data, responses entered via the user interface, etc.) using the historical sleep quality data, historical user input, and historical biometric data. In some embodiments, the historical sleep therapy data is indicative of whether the sleep therapy data is producing a targeted effect (e.g., improved EEG sensor data, improved sleep quality data).[000188] At block 1064, the processing logic trains a classifier machine learning model using data input including historical sleep quality data, historical user input, and historical biometric data and target output including the historical sleep therapy data to generate a trained classifier machine learning model (e.g., configured to provide output associated with sleep enhancement and diagnostics). The sleep enhancement and diagnostics may be by using the trained classifier machine learning model of FIG. 10D. In some embodiments, the trained machine learning model is a neural network. In some embodiments, only historical EEG sleep quality data, historical user input, and historical biometric data is identified (e.g., historical sleep therapy data is not identified) and the classifier machine learning model is trained using data input including the historical sleep quality data, historical user input, and historical biometric data (e.g., is an unsupervised model, clustering, is nottrained usingtarget output).[000189]The present disclosure may include a system, device, method, and / or non-transitory computer-readable program to execute operations that includes any combination of the following features (e.g., features 1-14.6) and / or any combination of the claims.[000190]The system, methods and devices that provide electroencephalogram sleep monitoring and sleep improvement are described.[000191] 1) The System comprises at least one Wearable Sensor System, containing at least one Wearable EEG electrode; at least one Patient-facing computer program; at least one Sensor Monitoring Algorithm; at least one Therapy Algorithm; at least one Backend System and Data Server; and at least one clinician -facing computer program.[000192] 1.1) wherein said wearable EEG electrode can be single-channel electrode, [000193] 1.2) wherein said wearable EEG electrode can be multi-channel electrodes, [000194] 1.3) wherein said wearable EEG electrode can be made of flexible or rigid electrode conductive materials, including but not limited to material such as conductive fabric, conductive printing or coating, conductive adhesives, conductive patches, or any materials made with silver, carbon, copper, iron, gold, or aluminum. Most preferably silver-based conductive fabric.Attorney Docket No.: 46528.6 (L0002PCT)[000195] 1.4) wherein said wearable EEG electrodes can be made of disposable, permanent or detachable EEG sensor electrodes. Preferably disposable.[000196] 1.5) wherein said wearable EEG electrodes can be made of dry or wet electrodes materials. Preferably dry electrodes.[000197] 1.6) wherein said wearable EEG electrode is embedded in devices such as headband, hat, headset, patch or any head-worn devices made of fabric, leather, plastic, rubber, etc. Preferably flexible headband made of fabric.[000198] 1.7) wherein said wearable EEG electrodes are the only sensor of the Wearable Sensor System.[000199] 1.8) wherein said wearable EEG electrode is combined with at least a non-EEG electrode sensor to collect biomedical data from electrocardiogram (ECG), electromyography (EMG), electrooculogram (EOG), photoplethysmogram (PPG), Accelerometer, and any other sensor from the body, Preferably PPG, accelerometer, and ECG.[000200] 1.8.1) wherein said wearable non-EEG electrode is obtained from devices such as ring, smartwatch, smart phones, oximeter, smart band, earrings, ear bugs, CPAP masks, intraoral devices, or any other wearable devices.[000201] 1.8.2) wherein said wearable non-EEG electrode devices described above collects data including but not limited to oxygen desaturation index, heart rate, heart rate variability, steps, temperature, sleep stages, sleep fragmentation index, blood pressure, respiration rate sensor, glucose level, galvanic skin response, electrodermal activity, heart muscle function, pulse wave velocity, apnea hypopnea index, CPAP machine pressure, CPAP machine usage time, respiratory effort-related sleep arousal, among others.[000202] 1.8.3) wherein said wearable non-EEG electrode devices described above transfer information to the Patient-facing computer program via communication integrations such as Bluetooth®, Wi-Fi®, application programming interface (API), quick-response (QR) codes and other means.[000203] 1.9) wherein said wearable EEG electrode is used to capture EEG data from individuals wearing the electrodes during the day or night and to transfer to the patient-facing computer program.[000204J2.) The system comprises at least a Wearable Sensor System, containing at least a wearable processor / controller unit[000205J2.1) Wherein said wearable processor / controller unit is a low-power or ultra-low power microcontroller. Preferably optimized for signal processing, filtering, wireless communication, power management unit, and integrated memory.Attorney Docket No.: 46528.6 (L0002PCT)[000206J2.2) Wherein said wearable processor / controllerunitis used for real-time processing of sensor data from wearable devices. Preferably the controller integrates Bluetooth Low Energy (BLE) for wireless communication, on-chip memory to store sensor data temporarily, providing flexibility for offline data processing when needed. Most preferably the controller features hardware-based security to ensure the integrity and privacy of biometric data, and an ultra-low-power consumption mode, enabling prolonged use on battery-powered devices, extending battery life for long-term wearable applications. Peripheral interfaces, to forward EEG data to storage and preferably with availability of two or more wireless interfaces. Most preferably with large internal memory.[000207J3) The system comprises at least a wearable sensor system, containing at least a wearable storage device[000208] 3. 1) Wherein said wearable storage device is a ferroelectric random access memory (FRAM) memory, Electrically-Erasable Programmable ROM (EEPROM) memory, Secure Digital (SD) Card or embedded MultiMediaCard (eMMC) module, magnetoresistive random access memory (MRAM) memory, or FLASH memory. Preferably internal or external flash memory. Most Preferably internal flash memory with a long life cycle and low power consumption, such as NAND (“NOT AND”) Flash memory and NOR (“NOT-OR”) Flash memory.[000209J3.2) Wherein said wearable storage device is used to store 7 days of data. Preferably 4 nights of data.[000210J4) The system comprises at least a wearable sensor system, containing at least a wearable battery device[000211J4.1) Wherein said wearable battery device is a prismatic lithium battery, a rechargeable Li-Polymer battery, rechargeable Lithium-Ion (Li-Ion) battery, Nickel-Metal Hydride (NiMH) battery, Li-Ion or Li-Polymer batteries, Lithium-Iron Phosphate (LiFePO4) battery, Zinc-Air battery. Preferably long cycle life, lightweight and high-energy density small coin lithium battery.[000212J4.2) Wherein said wearable battery device is used for providing an autonomy of24h of monitoring. Preferably at least lOh.[000213] 5) The system comprises at least a wearable sensor system, containing at least a wearable capacitor device[000214J5.1) Wherein said wearable capacitor device is a supercapacitor, used to provide rapid charging and high power output to handle high bursts of energy while preserving battery life.Attorney Docket No.: 46528.6 (L0002PCT)[000215J6) The system comprises at least a wearable sensor system, containing at least a wearable communication system.[000216J6.1) Wherein said wearable communication system is a low power consumption and short-range communication system, ideal for continuous data transmission in wearable devices.[000217J6.2) Wherein said wearable communication system is chosen from Bluetooth, WiFi, Zigbee, NFC (Near Field Communication), LoRa (Long Range), and 4G / 5G Cellular.Preferably Bluetooth® and Wi-Fi®.[000218J6.3) Wherein said wearable communication system is used for transferring the collected EEG data to the Patient-facing computer program.[000219J6.4) Wherein said wearable communication system is connected to the Patient- Facing Computer Program before, during and after the use of the wearable sensor system. Preferably after the use of the wearable sensor system.[000220J7) The system comprises at least a wearable sensor system, containing at least a firmware system.[000221J7.1) Wherein said firmware system is chosen to include real-time sensor data processing, wireless communication on the same chip, ultra-low-power consumption, and hardware-based security features such as encryption and secure boot.[000222J7.2) Wherein said firmware system is updated over-the-air (OTA) or not over-the- air. Preferably over-the-air.[000223J7.3) Wherein said firmware system is used for functions including but not limited to Hardware Control, Data Processing, Communication, Energy Management, User Interface, Updates and Security.[000224J8) The system comprises at least a wearable sensor system, containing at least a wearable signal quality system.[000225] 8.1) Wherein said wearable signal quality system is chosen from levels of amplitude and frequency range according to the expected. Preferably amplitude lower than lOOmV and containing low frequencies that are expected in sleep (0.01-35Hz).[000226] 8.2) Wherein said wearable signal quality system is used for removing sleep epochs that are not valid for the algorithm classification.[000227J9) The system comprises at least a wearable sensor system, wherein the components are integrated in a printed circuit board.[000228J9.1) Wherein the printed circuit board can be rigid or flexible.Attorney Docket No.: 46528.6 (L0002PCT)[000229J9.2) Wherein the printed circuit board can be covered and protected by a case made of any type of plastic, or non-plastic material.[000230J9.3) wherein the printed circuit board can be permanently connected to the wearable EEG electrodes or detachable from the wearable EEG electrodes.[000231J9.4) wherein the printed circuit board can be connected to the wearable EEG electrodes by means of any conductive material such as wire, thread, coating, printing plug, pin, rivet, adhesive, magnet, which can be made of any conductive element such as iron, silver, copper, aluminum.[000232J9.5) wherein the printed circuit board can be made of several layers. Most preferably higher than 4 layers.[000233J9.6) Wherein said PCB material is chosen from FR4 (Flame Retardant 4), GEM-1 (Composite Epoxy Material), Polyimide, Rogers 4000 series, and Aluminum-based PCBs. Preferably FR4 and Polyimide.[000234] 10) The system comprises at least a wearable sensor system, containing at least a On / Off switch system.[000235] 10.1) Wherein said On / Off switch system is chosen from slide Switches SPDT (Single Pole, Double Throw) On-On, Tactile Push Buttons, Momentary Push Buttons, Rocker Switches, Toggle Switches, and Capacitive Touch Switches. Preferably a Slide Switch SPDT On-On and a Capacitive Touch Switch resistance.[000236] 10.2) Wherein said On / Off switch system is used for turning the wearable sensor system on and off.[000237] 10.3) Wherein said On / Off switch system, contains visual power indication to the user, preferably containing LEDs.[000238] 11) The system comprises at least a Sensor Monitoring Algorithm.[000239] 11.1) Wherein the sensor monitoring algorithm uses input signal from Wearable Sensor System, and perform the following steps[000240] 11.1.1) Data preprocessing and segmentation encompasses various techniques to prepare data for analysis. This includes cleaning (handling missing values and correcting inconsistencies), filtering (using FIR, IIR, Butterworth, Chebyshev, Gaussian, Kalman, and median filters), normalizing (through one-hot or label encoding), and standardizing the data. Dimensionality reduction methods like principal component analysis (PCA) and feature selection simplify datasets, while transformations such as logarithmic scaling adjust distributions. Additionally, class balancing techniques, including oversampling and under sampling, address dataset imbalances.Attorney Docket No.: 46528.6 (L0002PCT)[000241] 11.1.2) Feature extraction involves identifying key patterns to improve analysis and modeling. Techniques include extracting statistical features (mean, variance), amplitudebased features (peak values, signal energy), and using Fourier and wavelet transforms to analyze frequency and time information. Entropy -based features measure signal complexity, while Principal Component Analysis (PCA) reduces dimensionality by identifying components that explain the most variance. Additionally, trends, seasonality, and lag features capture time-dependent characteristics.[000242] 11.1.3) Data Classification Model is a method for classifying data using machine learning and deep learning techniques. Machine learning includes supervised methods such as regression and classification algorithms (e.g., linear regression, logistic regression, support vector machines, decision trees, and ensemble methods like random forests and gradient boosting). Deep learning employs neural networks, including convolutional neural networks (CNNs) for image analysis and recurrent neural networks (RNNs) for sequential data. Techniques such as transfer learning and reinforcement learning enhance the model's adaptability and efficiency.[000243] 11.1.4) Extraction of output parameters is a stage where the desired parameters are extracted from the data in order to provide information for sleep improvement and diagnostics, including but not limited to qualitative and quantitative data such as sleep stage percentages and total time, sleep schedules, sleep efficiency, sleep onset latency, sleep stage transitions, REM latency, wake after sleep onset (WASO), sleep durations, and so forth. [000244] 11.2) Wherein the output can be used for providing information to diagnostic decisions or treatment for sleep disorders, including but not limited to insomnia, obstructive sleep apnea, narcolepsy, REM latency disorder, COMISA diagnostic, CPAP treatment, intraoral device treatment, medication treatment, cognitive behavioral therapy, among others. [000245] 11.3) Wherein the output is used to personalize the clinical protocol of cognitive behavioral therapy and other features present in the patient facing computer program.[000246] 11.4) Wherein the sensor monitoring algorithm has an accuracy of higher than 80%. Preferably higher than 90%. Most preferably higher than 95%, in comparison to the full polysomnography exam type 1.[000247] 11.5) Wherein the sensor monitoring algorithm is executed over a cloud system, or locally in the mobile phone device, or local in desktop over the internet. Preferably over a cloud system.[000248] 11.6) Wherein the Data Classification Model used in the system was developed by machine learning techniques, comparing the Wearable EEG System to the Golden StandardAttorney Docket No.: 46528.6 (L0002PCT)Polysomnography exam, where the labeled sleep stages from the polysomnography exam signal was used to teach the simultaneous signal from the Wearable EEG System, using the same methodology described in item 11.1.[000249] 12) The system comprises at least a Therapy Algorithm.[000250] 12.1) Wherein the Therapy Algorithm uses qualitative and quantitative, and objective and subjective inputs from Wearable Sensor System and Patient-facing computer program, including but not limited to EEG sensor, PPG sensor, ECG sensor, accelerometer sensor, diaries, anamneses, clinical assessment questionnaires, among others.[000251] 12.2) Wherein the Therapy Algorithm generates output as quantitative and qualitative feedback, including but not limited to text, graphs, videos, notifications and reminders containing information about sleep schedules, routines, habits and behaviors. [000252] 12.3) Wherein the output can be used for improving sleep quality.[000253] 12.4) Wherein the method for converting the input into output involves defining the variables combinations and limits, classifying the data using Machine Learning classification models according to item 11.1.3, and defining the outcomes.[000254] 12.5) Wherein the method for converting the input into output involves a combination of variables such as sleep efficiency, sleep onset latency, WASO, sleep questionnaires, oxygen desaturation index, sleep duration, REM sleep latency, sleep medication use, sleep stage transition, sleep staging percentage, among others.[000255] 12.6) Wherein the method for converting the input into output involves monitoring the patient for 30 days, preferably at least 7 days, most preferably at least 3 days.[000256] 12.7) Wherein the output is shown in the Patient-facing computer program, including user interfaces including but not limited to text, graphs, videos, notifications and reminders.[000257] 12.8) Wherein the output can be used for providing information to diagnostic decisions or treatment for sleep disorders, including but not limited to insomnia, obstructive sleep apnea, narcolepsy, REM latency disorder, COMISA diagnostic, CPAP treatment, intraoral device treatment, medication treatment, cognitive behavioral therapy, among others. [000258] 12.9) Wherein the output is used to personalize the clinical protocol of cognitive behavioral therapy and other features present in the patient facing computer program.[000259] 13) The system comprises at least a Patient-facing computer program.[000260] 13.1) Wherein the patient facing computer program is an application downloadable from public app stores.Attorney Docket No.: 46528.6 (L0002PCT)[000261] 13.2) Wherein the patient-facing application program is integrated to the Wearable Sensor System through the wearable communication system in order to collect wearable sensor outputs.[000262] 13.3) Wherein the patient-facing application program collects qualitative and quantitative inputs from users through features such as diaries, questionnaires, anamneses, and integrations from third-parties APIs, including non-EEG wearable sensors.[000263] 13.4) Wherein the patient-facing application program sends the information to the cloud backend system for storage and therapy algorithm calculations.[000264] 13.5) Wherein the patient-facing application program presents the output to the patient using reports, graphs, text, notifications and reminders, including downloadable reports[000265] 14) The system comprises at least a clinician -facing computer program.[000266] 14.1) Wherein the clinician -facing computer program is a website, hosted over internet to be used in desktop, laptops, tablets or mobile phones.[000267] 14.2) Wherein the clinician -facing computer program collects data from the Backend system to present to the clinician using visual interface such as text, graphs, reports, lists, including customizable reports, and downloadable reports.[000268] 14.3) Wherein the clinician -facing computer program report generated contains information about qualitative, quantitative, objective and subjective parameters of sleep, including but not limited to sleep staging, microarousal, desaturation index, heart rate, movement, sleep schedules, patient clinical profile, time to bed, time to sleep, wake-up time, go-to-bed time, sleep duration, sleep latency, REM sleep latency, REM sleep, non-rapid eye movement (NREM) sleep, percentage and duration of sleep stages, WASO, sleep transitions, sleep efficiency, among others.[000269] 14.4) Wherein the clinician-facing computer program report generated contains input from clinician including but not limited to clinician signature, brand identity and personalized subjective orientations.[000270J 14.5) Wherein the clinician-facing computer program is used for monitoring, treatment, assessing and diagnosing patients with professionals including but not limited to doctors of any type, dentists, physiotherapists, psychologists, among others.[000271] 14.6) Wherein the clinician-facing computer program sends information to the cloud backend system for storage and processing.[000272JFIG. 11 is a block diagram illustrating a computer system 1100, according to certain embodiments.Attorney Docket No.: 46528.6 (L0002PCT)[000273]In some embodiments, computer system 1100 is connected (e.g., via a network, such as a Local Area Network (LAN), an intranet, an extranet, or the Internet) to other computer systems. In some embodiments, computer system 1100 operates in the capacity of a server or a client computer in a client-server environment, or as a peer computer in a peer-to- peer or distributed network environment. In some embodiments, computer system 1100 is provided by a personal computer (PC), a tablet PC, a Set-Top Box (STB), a Personal Digital Assistant (PDA), a cellular telephone, a web appliance, a server, a network router, switch or bridge, or any device capable of executing a set of instructions (sequential or otherwise) that specify actions to be taken by that device. Further, the term “computer” shall include any collection of computers that individually or jointly execute a set (or multiple sets) of instructions to perform any one or more of the methods described herein.[000274]In a further aspect, the computer system 1100 includes a processing device 1102, a volatile memory 1104 (e.g., Random Access Memory (RAM)), a non-volatile memory 1106 (e.g., Read-Only Memory (ROM) or Electrically -Erasable Programmable ROM (EEPROM)), and a data storage device 1118, which communicate with each other via a bus 1130. [000275]In some embodiments, processing device 1102 is provided by one or more processors such as a general purpose processor (such as, for example, a Complex Instruction Set Computing (CISC) microprocessor, a Reduced Instruction Set Computing (RISC) microprocessor, a Very Long Instruction Word (VLIW) microprocessor, a microprocessor implementing other types of instruction sets, or a microprocessor implementing a combination of types of instruction sets) or a specialized processor (such as, for example, an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), a Digital Signal Processor (DSP), or a network processor).[000276]In some embodiments, computer system 1100 further includes a network interface device 1108 (e.g., coupled to network 1120). In some embodiments, computer system 1100 also includes a video display unit 1110 (e.g., a liquid crystal display (LCD)), an alphanumeric input device 1112 (e.g., a keyboard), a cursor control device 1114 (e.g., a mouse), and a signal generation device 1116.[000277]In some implementations, data storage device 1118 includes a non-transitory computer-readable storage medium 1124 on which store instructions 1126 encoding any one or more of the methods or functions described herein, including a recommendation component 123, predictive component 114, and / or instructions for implementing methods described herein.Attorney Docket No.: 46528.6 (L0002PCT)[000278]In some embodiments, instructions 1126 also reside, completely or partially, within volatile memory 1104 and / or within processing device 1102 during execution thereof by computer system 1100, hence, in some embodiments, volatile memory 1104 and processing device 1102 also constitute machine-readable storage media.[000279]While computer-readable storage medium 1124 is shown in the illustrative examples as a single medium, the term “computer-readable storage medium” shall include a single medium or multiple media (e.g., a centralized or distributed database, and / or associated caches and servers) that store the one or more sets of executable instructions. The term “computer-readable storage medium” shall also include any tangible medium that is capable of storing or encoding a set of instructions for execution by a computer that cause the computer to perform any one or more of the methods described herein. The term “computer- readable storage medium” shall include, but not be limited to, solid-state memories, optical media, and magnetic media.[000280]In some embodiments, the methods, components, and features described herein are implemented by discrete hardware components or are integrated in the functionality of other hardware components such as ASICs, FPGAs, DSPs or similar devices. In some embodiments, the methods, components, and features are implemented by firmware modules or functional circuitry within hardware devices. In some embodiments, the methods, components, and features are implemented in any combination of hardware devices and computer program components, or in computer programs.[000281]Unless specifically stated otherwise, terms such as “identifying,” “receiving,” “causing,” “training,” “generating,” “providing,” “obtaining,” “interrupting,” “determining,” “transmitting,” or the like, refer to actions and processes performed or implemented by computer systems that manipulates and transforms data represented as physical (electronic) quantities within the computer system registers and memories into other data similarly represented as physical quantities within the computer system memories or registers or other such information storage, transmission or display devices. In some embodiments, the terms “first,” “second,” “third,” “fourth,” etc. as used herein are meant as labels to distinguish among different elements and do not have an ordinal meaning according to their numerical designation.[000282]Examples described herein also relate to an apparatus for performing the methods described herein. In some embodiments, this apparatus is specially constructed for performing the methods described herein, or includes a general-purpose computer systemAttorney Docket No.: 46528.6 (L0002PCT) selectively programmed by a computer program stored in the computer system. Such a computer program is stored in a computer-readable tangible storage medium.[000283] Some of the methods and illustrative examples described herein are not inherently related to any particular computer or other apparatus. In some embodiments, various general- purpose systems are used in accordance with the teachings described herein. In some embodiments, a more specialized apparatus is constructed to perform methods described herein and / or each of their individual functions, routines, subroutines, or operations. Examples of the structure for a variety of these systems are set forth in the description above. [000284]The above description is intended to be illustrative, and not restrictive. Although the present disclosure has been described with references to specific illustrative examples and implementations, it will be recognized that the present disclosure is not limited to the examples and implementations described. The scope of the disclosure should be determined with reference to the following claims, along with the full scope of equivalents to which the claims are entitled.[000285]The preceding description sets forth numerous specific details such as examples of specific systems, components, methods, and so forth in order to provide a good understanding of several embodiments of the present disclosure. It will be apparent to one skilled in the art, however, that at least some embodiments of the present disclosure may be practiced without these specific details. In other instances, well-known components or methods are not described in detail or are presented in simple block diagram format in order to avoid unnecessarily obscuring the present disclosure. Thus, the specific details set forth are merely exemplary. Particular implementations may vary from these exemplary details and still be contemplated to be within the scope of the present disclosure.[000286]The terms “over,” “under,” “between,” “disposed on,” and “on” as used herein refer to a relative position of one material layer or component with respect to other layers or components. For example, one layer disposed on, over, or under another layer may be directly in contact with the other layer or may have one or more intervening layers. Moreover, one layer disposed between two layers may be directly in contact with the two layers or may have one or more intervening layers. Similarly, unless explicitly stated otherwise, one feature disposed between two features may be in direct contact with the adjacent features or may have one or more intervening layers.[000287]The words “example” or “exemplary” are used herein to mean serving as an example, instance or illustration. Any aspect or design described herein as “example’ or “exemplary” is not necessarily to be construed as preferred or advantageous over otherAttorney Docket No.: 46528.6 (L0002PCT) aspects or designs. Rather, use of the words “example” or “exemplary” is intended to present concepts in a concrete fashion.[000288]Reference throughout this specification to “one embodiment,” “an embodiment,” or “some embodiments” means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment. Thus, the appearances of the phrase “in one embodiment,” “in an embodiment,” or “in some embodiments” in various places throughout this specification are not necessarily all referring to the same embodiment. In addition, the term “or” is intended to mean an inclusive “or” rather than an exclusive “or.” That is, unless specified otherwise, or clear from context, “X includes A or B” is intended to mean any of the natural inclusive permutations. That is, if X includes A; X includes B; or X includes both A and B, then “X includes A or B” is satisfied under any of the foregoing instances. In addition, the articles “a” and “an” as used in this application and the appended claims should generally be construed to mean “one or more” unless specified otherwise or clear from context to be directed to a singular form. Also, the terms “first,” “second,” “third,” “fourth,” etc. as used herein are meant as labels to distinguish among different elements and can not necessarily have an ordinal meaning according to their numerical designation. When the term “about,” “substantially,” or “approximately” is used herein, this is intended to mean that the nominal value presented is precise within ±10%. [000289]Although the operations of the methods herein are shown and described in a particular order, the order of operations of each method may be altered so that certain operations may be performed in an inverse order so that certain operations may be performed, at least in part, concurrently with other operations. In another embodiment, instructions or sub-operations of distinct operations may be in an intermittent and / or alternating manner. [000290]It is understood that the above description is intended to be illustrative, and not restrictive. Many other embodiments will be apparent to those of skill in the art upon reading and understanding the above description. The scope of the disclosure should, therefore, be determined with reference to the appended claims, along with the full scope of equivalents to which such claims are entitled.

Claims

Attorney Docket No.: 46528.6 (L0002PCT)CLAIMSWhat is claimed is:1 . A method comprising: identifying, by a processing device, electroencephalogram (EEG) sensor data obtained by a wearable device associated with a head of a user; providing, by the processing device, the EEG sensor data as a first data input to a first trained machine learning model; obtaining, by the processing device, a first output from the first trained machine learning model, wherein the first output is associated with sleep quality data associated with the user; providing, by the processing device, a second input to a second trained machine learning model, wherein the second input comprises the sleep quality data and user input data; obtaining, by the processing device from the second trained machine learning model, a second output associated with sleep therapy data; subsequent to obtaining the second output associated with the sleep therapy data, identifying at least one of subsequent EEG sensor data or subsequent user input data; and responsive to determining that the at least one of the subsequent EEG sensor data or the sub sequent user input data meets a first threshold value, re-training at least one of the first trained machine learning model or the second trained machine learning model based on the at least one of the subsequent EEG sensor data or the subsequent user input data to generate updated sleep therapy data.

2. The method of claim 1, wherein the first threshold value comprises a predefined variance between a baseline sleep-quality metric comprised in the sleep quality data and a corresponding metric derived from the at least one of the subsequent EEG sensor data or the subsequent user input data.

3. The method of claim 2, wherein the baseline sleep-quality metric comprises at least one of: a sleep efficiency value; a sleep onset latency value; a wake-after-sleep-onset value;Attorney Docket No.: 46528.6 (L0002PCT) a total rapid eye movement (REM) sleep duration value; a total non-rapid eye movement (NREM) sleep duration value; or a total sleep time value.

4. The method of claim 1, wherein the wearable device is configured to store the EEG sensor data, wherein the EEG sensor data is obtained from the wearable device responsive to a wireless connection being established with the wearable device.

5. The method of claim 1, wherein the user input data comprises at least one of: user input associated with the user, wherein the user input is received via a user interface; or biometric data associated with the user, wherein the biometric data is different from the EEG sensor data, and wherein the biometric data is obtained by at least one additional sensor device.

6. The method of claim 1, further comprising: identifying historical EEG sensor data; identifying historical sleep quality data; and training a first machine learning model using data input comprising the historical EEG sensor data and target output comprising the historical sleep quality data to generate the first trained machine learning model.

7. The method of claim 1, further comprising: identifying historical sleep quality data, historical user input, and historical biometric data; identifying historical sleep therapy data; and training a second machine learning model using data input comprising the historical sleep quality data, the historical user input, and the historical biometric data and target output comprising the historical sleep therapy data to generate the second trained machine learning model.Attorney Docket No.: 46528.6 (L0002PCT)8. The method of claim 1, further comprising: receiving, by the processing device, the subsequent EEG sensor data associated with the user, wherein the subsequent EEG sensor data is received after providing the sleep therapy data; and determining, based on the subsequent EEG sensor data, whetherthe sleep therapy data is producing a targeted effect, wherein the targeted effect comprises at least one of improved EEG sensor data or improved sleep quality data.

9. The method of claim 1, wherein the wearable device comprises a memory configured to store between about 20 hours and about 80 hours of EEG sensor data.

10. The method of claim 1, wherein the wearable device is configured to store the EEG sensor data at predetermined time intervals and to transmit the stored EEG sensor data in a plurality of discrete packets to the processing device after a monitoring period has ended.

11. A system comprising: a memory; and a processing device coupled to the memory, wherein the processing device is configured to: identify electroencephalogram (EEG) sensor data obtained by a wearable device associated with a head of a user; provide the EEG sensor data as a first data input to a first trained machine learning model; obtain a first output from the first trained machine learning model, wherein the first output is associated with sleep quality data associated with the user; provide a second input to a second trained machine learning model, wherein the second input comprises the sleep quality data and user input data; obtain, from the second trained machine learning model, a second output associated with sleep therapy data; subsequent to obtaining the second output associated with the sleep therapy data, identify at least one of subsequent EEG sensor data or subsequent user input data; and responsive to determining that the at least one of the subsequent EEG sensor data or the subsequent user input data meets a first threshold value, re-train at leastAttorney Docket No.: 46528.6 (L0002PCT) one of the first trained machine learning model or the second trained machine learning model based on the at least one of the subsequent EEG sensor data or the subsequent user input data to generate updated sleep therapy data.

12. The system of claim 11, wherein the first threshold value comprises a predefined variance between a baseline sleep-quality metric comprised in the sleep quality data and a corresponding metric derived from the at least one of the subsequent EEG sensor data or the subsequent user input data.

13. The system of claim 12, wherein the baseline sleep-quality metric comprises at least one of: a sleep efficiency value; a sleep onset latency value; a wake-after-sleep-onset value; a total rapid eye movement (REM) sleep duration value; a total non-rapid eye movement (NREM) sleep duration value; or a total sleep time value.

14. The system of claim 11, wherein the wearable device is configured to store the EEG sensor data, wherein the EEG sensor data is received from the wearable device responsive to a wireless connection being established with the wearable device.

15. The system of claim 11, wherein the user input data comprises at least one of: user input associated with the user, wherein the user input is received via a user interface; or biometric data associated with the user, wherein the biometric data is different from the EEG sensor data, and wherein the biometric data is obtained by at least one additional sensor device.

16. A non-transitory machine-readable storage medium storing instructions which, when executed cause a processing device to perform operations comprising: identifying electroencephalogram (EEG) sensor data obtained by a wearable device associated with a head of a user;Attorney Docket No.: 46528.6 (L0002PCT) providing the EEG sensor data as a first data input to a first trained machine learning model; obtaining a first output from the first trained machine learning model, wherein the first output is associated with sleep quality data associated with the user; providing a second input to a second trained machine learning model, wherein the second input comprises the sleep quality data and user input data; obtaining, from the second trained machine learning model, a second output associated with sleep therapy data; subsequent to obtaining the second output associated with the sleep therapy data, identifying at least one of subsequent EEG sensor data or subsequent user input data; and responsive to determining that the at least one of the subsequent EEG sensor data or the sub sequent user input data meets a first threshold value, re-training at least one of the first trained machine learning model or the second trained machine learning model based on the at least one of the subsequent EEG sensor data or the subsequent user input data to generate updated sleep therapy data.

17. The non-transitory machine-readable storage medium of claim 16, wherein the first threshold value comprises a predefined variance between a baseline sleep-quality metric comprised in the sleep quality data and a corresponding metric derived from the at least one of the subsequent EEG sensor data or the subsequent user input data.

18. The non-transitory machine-readable storage medium of claim 17, wherein the baseline sleep-quality metric comprises at least one of: a sleep efficiency value; a sleep onset latency value; a wake-after-sleep-onset value; a total rapid eye movement (REM) sleep duration value; a total non-rapid eye movement (NREM) sleep duration value; or a total sleep time value.

19. The non-transitory machine-readable storage medium of claim 16, wherein the wearable device is configured to store the EEG sensor data, wherein the EEG sensor data is received from the wearable device responsive to a wireless connection being established with the wearable device.Attorney Docket No.: 46528.6 (L0002PCT)20. The non-transitory machine-readable storage medium of claim 16, wherein the user input data comprises at least one of: user input associated with the user, wherein the user input is received via a user interface; or biometric data associated with the user, wherein the biometric data is different from the EEG sensor data, and wherein the biometric data is obtained by at least one additional sensor device.

Images