NEAT 001

NEAT 001 is software that automatically scores sleep stages from EEG data according to American Academy of Sleep Medicine definitions. It is a client-server application using machine learning to analyze forehead EEG signals and classify sleep stages (Wake, REM, N1, N2, N3) for adult populations in physician offices or home settings.

NEAT is a software-only Class II device using machine learning on a single forehead EEG signal, similar to the predicate EnsoSleep. Both devices automatically score the same five sleep stages and produce sleep statistics. NEAT differs by not detecting leg movements or sleep disordered breathing events, which EnsoSleep does. NEAT's architecture uses a client-server model with Docker containerization, while the predicate uses a software-as-a-service cloud model.

AAMI/ANSI/IEC 62304 Standard was applied with Class B safety classification. FDA Guidance on 'Content of Premarket Submissions for Software Contained in Medical Device' was followed. Performance evaluation used bootstrapped resampling (2000 resamples) to calculate sensitivity, specificity, and overall agreement metrics with 95% confidence intervals.

NEAT and EnsoSleep perform equivalently for automated sleep staging from EEG, with differences in individual sleep stage accuracy (1-7%) that fall within expected human rater variability. Although NEAT excels at Wake, N1, and N3 classification while EnsoSleep is better at REM and N2, both devices achieve clinically comparable results. The functional limitation that NEAT lacks leg movement and apnea detection does not affect the core indication of sleep stage identification, making it substantially equivalent for its narrower, focused indication.