Workshop

Sleep is an essential part of our health, physical and mental, and overall well-being. There are a variety of sleep related disorders such as insomnia, sleep apnea, narcolepsy and hypersomnia that affect a significant percentage of the populace. Moreover, sleep disorders are often related to other ailments that include diabetes, stroke, obesity and cardiovascular diseases. About 50 to 70 million people is USA suffer from sleep disorders and undiagnosed sleep apnea alone costs 150 billion dollars annually. In addition to being a serious health concern, it reduces productivity and may cause accidents in workplace. 

Electroencephalography (EEG) signals are effective descriptors of sleep stages, a cyclic repetition of several patterns that can broadly categorized into AWAKE, non-rapid eye movement (NREM) and rapid eye movement (REM). NREM stage includes the slow wave sleep (SWS) stages. An important micro-structure in NREM is cyclic alternating patterns (CAP) which are basically periodic EEG activity lasting between 2 to 60 seconds, and closely associated with sleep stability and disruptions. These patterns are of two types: A –phase and B-phase while the A-phase has three subtypes. Detection of these patterns is important for diagnosing and monitoring of various sleep disorders and improving sleep health. It may also be helpful for various applications e.g. driver drowsiness detection, sleep tracking and study of food cravings etc.

 In this tutorial, first the sleep stages and subsequently CAP will be introduced. The need for the detection of CAP will be briefly explained. Subsequently, the state-of-the art in the identification of CAP will be discussed including the ones employing machine learning. Consequently, a deep-learning neural network model exploiting attention mechanism for detecting CAP will be described. Experimental results on publicly available CAP database from Physionet will be presented and compared with those of other recent methods. Limitations and future prospects will be outlined afterwards.