Validation of a new, minimally-invasive, software smartphone device to predict sleep apnea and its severity

Résumé fourni par la source

To assess the performance of Apneal®, an application that only requires the patient to tape a smartphone on the thorax for one night, compared with polysomnography. Methods: We concomitantly recorded polysomnography (PSG, following AASM standards for recording and scoring) and smartphone signals (microphone, accelerometer, and gyroscope) with the Apneal® v0.1 software for one night in patients referred to a sleep center. PSG were scored manually by trained sleep specialists. Apneal recordings were scored by non-specialist scorers who did a first manual scoring based on various signals extracted from the smartphone sensors, like the probability of snoring, or the filtered acceleration. We also used a deep learning model to produce an automatic scoring, using as input features extracted from the smartphone signals. We compared the Apnea Hypopnea Indexes (AHI) obtained by these different scorings with the PSG as a reference. Results: We included 44 patients (mean age 57.5, mean BMI 28.7, men 66%) with suspected OSA. Pearson correlation was 0.90 (p-value=1.25 x 10-17) for manual scoring, and 0.87 (p-value = 1.7 x 10-12) for automatic scoring. Using a cut-off of AHI≥ 15, we found a sensitivity of 0.91 (95% CI [0.8, 1]) and a positive predictive value (PPV) of 0.89 (CI 95% [0.76, 0.97]) for manual scoring, and sensitivity of 1 (95% CI [0.95, 1]) and PPV 0.9 (95% CI [0.8, 0.9]) for automatic scoring. Conclusions: The first results obtained by this new minimally-invasive method to diagnose sleep apnea are promising and offer a proof of concept for further larger studies.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.