Dreem 3S

The Dreem 3S is a wearable headband device designed for at-home or clinical sleep assessment in adults, utilizing dry EEG electrodes and an accelerometer to record brain activity and movement. Its core technology relies on an artificial neural network algorithm to automatically analyze these physiological signals, generating hypnograms and sleep metrics with performance showing high agreement with lab-based polysomnography for most sleep stages.

The Dreem 3S is a prescription-use wearable headband designed for the assessment of sleep in adult individuals (22-65 years old) in both home and healthcare environments. It functions by measuring, recording, displaying, transmitting, and analyzing the brain's electrical activity (EEG) to differentiate between sleep and wake states. The device comprises six dry EEG electrodes (frontal, occipital, reference, and ground) and a 3D accelerometer to detect head movements, all integrated into a flexible casing. Data captured, including raw EEG and accelerometer readings, are transferred to Dreem's servers for subsequent analysis, and the device can also generate user-defined reports, including a hypnogram detailing sleep scoring by 30-second epochs and a summary of derived sleep metrics. The Dreem 3S employs an artificial neural network algorithm to perform automatic sleep staging according to AASM classifications, analyzing frequency spectrums computed from the raw EEG and accelerometer data through recurrent and attention layers. This algorithm outputs robust sleep scoring by combining predictions for multiple 30-second epochs. Clinical performance investigations demonstrated its capabilities by comparing its sleep staging output to a laboratory polysomnography (PSG) system. In a study with 38 subjects, the Dreem 3S showed high positive agreement with expert-scored PSG for Wake (88.5%), N2 (83.4%), N3 (98.2%), and REM (91.57%) sleep stages, with N1 at 58.0%. Furthermore, an EEG data quality study confirmed that 96.6% of epochs per night were acceptable for manual scoring, and a usability study indicated successful use and tolerability in a home setting.

AI Regulatory Concerns

Powered by

Jev reads the full FDA letter and returns calibrated concern probabilities per regulatory category. These are AI estimates, not regulatory advice.

• Q1. How does the performance of the Dreem 3S's automatic sleep staging algorithm, particularly for N1 sleep, compare to manual scoring by multiple independent, certified polysomnographic technologists in a larger and more diverse patient population, including those outside the 22-65 age range or with specific sleep disorders?; What is the long-term reliability and signal quality of the dry EEG electrodes used in the Dreem 3S after repeated use and under various environmental conditions (e.g., humidity, temperature)?; Could the accelerometer data be further leveraged to differentiate between sleep stages or identify specific sleep-related movement disorders beyond general head movement?; What are the specific neural network architectures (e.g., types of recurrent and attention layers) and training datasets used for the sleep staging algorithm, and how robust is the algorithm to inter-individual variability in EEG patterns?; How does the Dreem 3S's performance in a home environment, with potential for increased artifacts and patient non-compliance, compare to its performance in the controlled sleep lab setting presented in the clinical investigation?