Predicting Sleep Disordered Breathing Events

Abstract

Sleep is annotated by various physiological events however, the underlying generator processes responsible for their manifestation is unclear. Moreover, their relationship with each other is also unclear. We focus on Sleep Disordered Breathing Events which cause sleep fragmentation and evaluate a subset of these events namely, obstructive apnoea, hypopnoea and central apnoea events collectively called as apnoea events. Occurrence of such events may exhibit as disruption in the proper functioning of the electrocardiogram thereby, increasing risks to cardiovascular health. We further expand our analysis to consider cortical arousals and leg movements during sleep and particularly, the subset of periodic leg movements during sleep as bio-markers to occurrence of such events thereby, positing a hypothesis about their relationship. The larger set of events -- Sleep Disordered Breathing -- which has not been studied well is evaluated from the lens of significant breathing changes after controlling for apneic and arousal activities which enforces our view that these events are significant and may add to the increase in cardiovascular health risks. Numerous occurrence of apnoea events, defined as partial or full blockage of airflow during sleep, have been shown to be an indicator of cardiovascular health risks. The apnoea/hypopnoea index (AHI) is used clinically to define the severity for a night wherein an AHI >= 30 indicates severe obstructive sleep apnoea (OSA). However, their accurate measurement is dependent on use of nasal airflow and nasal pressure devices. This study seeks to add academic work and explore the relationship between these manifestations during sleep to increase understanding of the underlying processes. Furthermore, the work confirms that measurement of such events maybe easily monitored at home, instead of polysomnogram centers, and thereby, help in easier identification of indicative cardiovascular issues. The analysis and prediction of apnoea events is based on the public dataset from the Apnea, Bariatric surgery, and CPAP (ABC) study collected for 49 subjects (all of whom suffer from severe obstructive sleep apnoea) leading to a core dataset of 26 nights from the polysomnogram (PSG) before the subjects undergo bariatric surgery or CPAP treatment. The study is conducted in two parts: (i) prediction of apnoea events using ECG morphology and thereafter, improve the model by addition of more information from other channels to develop a relationship between these manifested events and their underlying processes (ii) conduct first principal analysis to identify a valid hypothesis for sleep disordered breathing events after controlling for apneic and arousal events to evaluate the changes in breathing around PLMS.