🤖 AI Summary
This study addresses the limitations of conventional contact-based pulse oximetry—namely, reduced user comfort, limited accessibility, and unsuitability for remote monitoring—while overcoming robustness challenges posed by ambient illumination variations and skin tone diversity. We propose a non-contact SpO₂ estimation method using standard RGB cameras and introduce an end-to-end deep learning framework that jointly models physiological signals in both time and frequency domains, extracts dynamic photoplethysmographic (PPG) features, and incorporates a skin-tone-adaptive normalization mechanism. To our knowledge, this is the first work to systematically evaluate cross-population generalizability on a large-scale, multi-ethnic dataset. Experimental results demonstrate a mean absolute error (MAE) of under 2.1% on both internal and external validation datasets—meeting clinical acceptability thresholds—and significantly outperforming traditional PPG-based approaches. The method establishes a scalable, unobtrusive, and remote-capable solution for continuous physiological monitoring.
📝 Abstract
Blood oxygen saturation (SpO2) is a crucial vital sign routinely monitored in medical settings. Traditional methods require dedicated contact sensors, limiting accessibility and comfort. This study presents a deep learning framework for contactless SpO2 measurement using an off-the-shelf camera, addressing challenges related to lighting variations and skin tone diversity. We conducted two large-scale studies with diverse participants and evaluated our method against traditional signal processing approaches in intra- and inter-dataset scenarios. Our approach demonstrated consistent accuracy across demographic groups, highlighting the feasibility of camera-based SpO2 monitoring as a scalable and non-invasive tool for remote health assessment.