A Transformer-Based Deep Learning Model for Sleep Apnea Detection and Application on RingConn Smart Ring

Obstructive sleep apnea hypopnea syndrome (OSAHS) is a prevalent and detrimental respiratory disease. Deep learning has emerged as a valuable tool for detecting physiological events, including OSAHS. However, current OSAHS detection relies on cumbersome polysomnography (PSG) devices. To address this, we propose a transformer-based deep learning model that utilizes multi-modal physiological signals from wearable devices. Our model achieved outstanding OSAHS detection results on public datasets, with an impressive F1-score of 76.6. We also evaluated the performance of signal combinations for oxygen desaturation event detection (DESAT) in wearable devices. Notably, no existing work validates OSAHS monitoring based on a smart ring. Therefore, we verified our model using data collected from the RingConn smart ring and performed polysomnography apnea hypopnea index (PSG-AHI) estimation. We observed a strong correlation between the AHI calculated by the smart ring and the PSG-AHI, as indicated by a high correlation coefficient of ρ=0.96. These findings open possibilities for the miniaturization and convenience of OSAHS detection.