SomnoMetry

SomnoMetry is an AI/ML-enabled software medical device designed to assist clinicians by automatically analyzing polysomnography data for sleep staging, AHI calculation, and the detection of obstructive apneas in adult patients. It operates by employing a range of signal processing, conventional machine learning, and deep learning algorithms on recorded physiological signals to provide scores and insights that aid in the diagnosis of sleep and respiratory-related sleep disorders, with all outputs subject to clinician verification.

SomnoMetry (V1.0) is an Artificial Intelligence/Machine Learning (AI/ML)-enabled Software as a Medical Device (SaMD) developed by Neumetry Medical Inc. to assist physicians in evaluating sleep quality and diagnosing sleep and respiratory-related sleep disorders in adult patients. The software automatically analyzes physiological signals recorded during polysomnography (PSG) studies to score sleep stages, calculate the Apnea-Hypopnea Index (AHI), and detect sleep-disordered breathing events, with a focus on obstructive apneas. It is designed for use in clinical settings under physician supervision, and all automated scoring results must be verified by a qualified clinician before final diagnostic decisions are made. The system employs a variety of signal processing techniques, data indexing, conventional machine learning, and deep learning algorithms to process PSG data and provide actionable clinical insights. Its architecture comprises a web Application Programming Interface (API) for secure PSG file uploads, a database for storing input and output data, an AI/ML engine for analysis, a dashboard for clinicians to review, manage, edit, and verify results, and a reporting API to generate sleep reports. Performance data from a retrospective clinical evaluation involving 201 PSG files indicated that SomnoMetry's sleep staging performance is non-inferior to the AASM gold standard of manual scoring, and its diagnostic performance for sleep apnea (based on AHI) is substantially equivalent to that of a predicate device, showing consistent agreement across different age groups and sleep apnea severity levels.

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 SomnoMetry AI/ML algorithm differentiate between obstructive apneas and other types of respiratory events (e.g., central apneas, mixed apneas, hypopneas) when calculating the Apnea-Hypopnea Index (AHI), especially since the predicate device required manual marking for some of these?; What specific physiological signals from the polysomnography (PSG) are prioritized or weighted most heavily by the AI/ML engine for accurate sleep staging and respiratory event detection?; Considering the performance data, particularly the confidence intervals and likelihood ratios across different AHI thresholds and age groups, what is the expected range of diagnostic accuracy variability when SomnoMetry is applied to broader, more diverse patient populations with various comorbidities not detailed in the study?; How does the SomnoMetry system manage and process PSG recordings with significant signal artifacts or noise, and what mechanisms are in place to flag studies where data quality might compromise the AI/ML algorithm's reliability?; What are the specific computational demands of the SomnoMetry AI/ML engine, such as average processing time per sleep study and cloud infrastructure requirements, and how do these factors scale with increasing PSG data volume or the complexity of sleep disorders?