Topological Data Analysis Based Characteristics of Electroencephalogram Signals in Children With Sleep Apnea

This study aims to identify differences in the functional neural connectivity of the brain of paediatric patients with obstructive sleep apnea. Using EEG signals from 3673 paediatric patients, we grouped subjects into OSA or control groups based on sleep oxygen desaturation levels and apnea-hypopnea index (AHI), and applied topological data analysis (TDA) techniques. We evaluated our approach through statistical testing of TDA-based EEG features, which indicate fundamental differences in the functional neural connectivity of subjects with sleep apnea as compared to controls. There were statistically significant differences ( <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:semantics><mml:mrow><mml:mi>p</mml:mi> <mml:mo><</mml:mo> <mml:mn>0.05</mml:mn></mml:mrow> </mml:semantics> </mml:math> ) between EEG signals taken during apnea and hypopnea events as compared to those taken from healthy controls. No significance was found between the latent EEG signals within the same groups. We observed significant differences between EEG signals collected during oxygen desaturation as compared to the EEG signals of the controls. We additionally identified significant differences between the latent EEG signals (i.e., no oxygen desaturation event occurring) of subjects as compared with the EEGs from controls. Lastly, significant differences were additionally found in the awake before sleep portion of the polysomnograms when grouping subjects based on minimum oxygen saturation experienced during sleep. TDA techniques allow us to identify statistically significant differences between the EEG signals of subjects with OSA and healthy controls, including during awake periods. Our results provide novel insights on the effects of OSA on the central nervous system, and insights into potential novel methods for identification of sleep apnea.