Abstract by Chris Fortuna
Deep Sleep Apnea Screening With Adapted Transformer Networks
Current methods for diagnosing Obstructive Sleep Apnea (OSA) require an expensive and intrusive overnight sleep study with expert review. Under the current system over 80% of OSA remains undiagnosed. We propose a novel screening system for diagnosing OSA accurately in the home using only pulse-oximetry data available on common smart watches and devices. We present a novel deep learning architecture for interpretable biosignals screening, inspired by SliceNet and Transformer networks.