Employing Neural Network Methods to Label Sleep EEG Micro-Arousals in Obstructive Sleep Apnea Syndrome

Well-designed studies are essential to screen suspected cases of Obstructive Sleep Apnea Syndrome (OSAS) using the widely-referenced questionnaires and then to confirm the diagnosis by means of full Polysomnography (PSG), and finally to assess various variables of treatment efficacy and safety. Defining the severity index of OSAS based on the Apnea-Hypopnea Index (AHI), sleep marco- and micro-structural features (i.e. hypnogram and cyclic alternating patterns or CAPs), and neurocognitive functions would help better explain the treatment outcome. Using the neural network models on sleep data in OSAS sufferers is potentially expected to help the above goals. Determination of neurocognitive impairments in OSAS subjects in relation with disease severity indices and subsequent changes in microstructural changes (i.e. CAPs) in sleep Electroencephalography (EEG), would therefore be useful in defining individualized care and cognitive rehabilitation plans. The present methodology paper has attempted to address the above hypothesis in a clinical population from a hospital-based sleep disorders laboratory.