Detection of Sleep Stages and Arousals Using Neural Network Classifiers

Conditions

• Sleep Apnea
• Sleep Disordered Breathing (SDB)

Eligibility Criteria

Sex

ALL

Age

18 Years - N/A

Healthy Volunteers

Accepted

Interventions

• SSAM — DEVICE
  Embedded software featuring neural-network classifiers designed to detect and classify sleep stages (N1/N2, N3, REM, Wake), autonomic arousals, and an Arousal Index based on data (specifically instantaneous pulse rate and full-resolution flow waveforms) collected during PAP therapy.
• Polysomnography — DIAGNOSTIC_TEST
  Polysomnography (PSG) is an overnight, in-laboratory diagnostic procedure that records multiple physiological parameters during sleep. It typically includes monitoring of brain activity (EEG), eye movements (EOG), muscle activity (EMG), heart rhythm (ECG), respiratory effort, airflow, oxygen saturation, and body position. PSG is used to evaluate sleep architecture, identify sleep stages, detect arousals, and diagnose sleep disorders such as sleep apnea. The collected data are analyzed manually by trained sleep technologists following standardized scoring criteria.
• EnsoSleep — DEVICE
  EnsoSleep is software-only medical device intended for use by physicians to assess sleep quality and aid in the diagnosis of sleep disorders. The software analyzes physiological signals and automatically scores sleep study results, including respiratory, sleep staging, arousal and movement events. Automatically scored events and physiological signals are analyzed, displayed, and summarized for review by clinicians.

Study Details

The objective of this clinical study is to evaluate the accuracy of the Smart Mask V1 System (herein 'Smart Mask') in measuring sleep stages-Stage N1/N2, Stage N3, Rapid Eye Movement (herein 'REM'), and WAKE-arousals, and the Arousal Index in adults diagnosed with sleep-disordered breathing, such as obstructive sleep apnea (herein 'OSA'). The Smart Mask operates in concert with a Wireless Access Module (herein 'WAM'), which is connected to a standard positive air pressure (herein 'PAP') device used in the treatment of OSA. Collectively the Smart Mask and WAM operate neural network classifier algorithms to determine sleep stages, arousals, and Arousal Index. These algorithms are coded into an embedded software system called the Sleep Staging and Arousal Module (herein 'SSAM') that operates directly on the WAM. The SSAM processes the following parameters, collected while the participant is asleep: 1) instantaneous values of pulse rate, determined from embedded optical sensors within the Smart Mask that measure photoplethysmogram waveforms (herein 'PPG'); and 2) full-resolution flow waveforms measured by sensors within the PAP device and retrieved by the WAM. During the study, volunteer participants (preferably those with OSA) will undergo an overnight sleep study in sleep testing facility located at three separate clinical sites. The test device (comprising the SSAM operating on the WAM) will retrospectively determine sleep stages and arousals, after the participant's sleep session has concluded. To evaluate the accuracy of the test device, its values of sleep stages, arousals, and Arousal Index will be compared to those parameters determined by polysomnography (herein 'PSG', a recognized gold-standard reference) and EnsoSleep (a FDA-cleared predicate device, and specifically a software package that uses artificial intelligence (AI) to determine sleep stages and arousals). Each volunteer participant will wear an FDA-cleared wrist-worn pulse oximeter called the 'Checkme O2' which generates data (specifically PPG waveforms and values of SpO2 and pulse rate) for the EnsoSleep cloud-based software platform. The main questions this study aims to answer are: * Can the Smart Mask accurately identify different sleep stages compared to the EnsoSleep device? * Can the Smart Mask accurately identify sleep arousals and calculate the Arousal Index compared to the EnsoSleep device? Answers to these questions will be derived through comparative statistical analysis involving the test device, the gold-standard PSG reference, and the FDA-cleared predicate device, employing methodologies similar to those used in the validation of the EnsoSleep. The study will include two cohorts. The first cohort will include approximately 75 participants from a single clinical site and will be used for device training purposes. The second cohort will consist of approximately 72 different participants, and will be used to validate the test device. Participants in the second cohort will be distributed roughly evenly across two separate clinical sites.

Key Dates

Start date

Sep 30, 2025

Status verified

Jun 2025

Primary completion

Dec 31, 2025

Completion

Dec 31, 2025

Study Design

Enrollment

150 participants (estimated)

Allocation

NON_RANDOMIZED

Intervention model

SINGLE_GROUP

Primary purpose

SUPPORTIVE_CARE

Arms

• Experimental: Phase I - Training
  Participants in this arm will undergo a one-night, in-lab sleep study while wearing the Smart Mask and WatchPAT while receiving positive airway pressure ('PAP') therapy. Standard polysomnography ('PSG') data will be collected and stored. Three independent registered PSG technologists ('RPSGTs') will manually score the data to estimate sleep stages and arousals, following the guidelines in 'The AASM Manual for the Scoring of Sleep and Associated Events: Rules, Terminology and Technical Specifications', which is the definitive reference for the scoring of PSG and HSATs. PSG data will also be uploaded to EnsoSleep's cloud-based software system. Data from this phase will be used solely to train and refine the Smart Mask's neural network algorithms for detecting sleep stages and arousals. No algorithm validation will occur in this phase.
• Experimental: Phase II - Validation
  Participants in this arm will undergo a single overnight, in-lab sleep study while wearing the Smart Mask and WatchPAT device, while receiving PAP therapy. Standard PSG data will be collected and stored. Data from the Smart Mask will be processed using trained neural networks to detect sleep stages (N1/N2, N3, REM, Wake), arousals, and calculate the Arousal Index. These outputs will be compared to: 1) manual scoring of PSG data by three independent, blinded RPSGTs; and 2) results from the predicate device, EnsoSleep. Data from this phase will be used validate the Smart Mask's neural network algorithms for detecting sleep stages and arousals.

Primary Outcome Measure

Agreement Between Smart Mask V1 System and EnsoSleep in Detecting Arousals and Arousal Index [ Time Frame: One overnight sleep study session (6-8 hours per participant) ]

Locations (3)

Find similar trials in Burlingame, CA

Related Studies

• Do Endotypes Predict Response and Sequelae in OSA Patients
  PHASE2 · Recruiting · University of California, San Diego · La Jolla, California
• Research Framework Exploring Sleep Health
  Recruiting · Scripps Translational Science Institute · La Jolla, California
• Behavioral Optimization to Overcome Obstructive Sleep Apnea Treatment Barriers
  Not Yet Recruiting · VA Office of Research and Development · West Los Angeles, California
• Effect of Oxymetazoline Hydrochloride in Combination With Fluticasone Propionate on the Apnea Hypopnea Index (AHI) in Subject With Persistent Nasal Congestion and Mild Obstructive Sleep Apnea
  PHASE4 · Recruiting · University of South Florida · Tampa, Florida