Published Paper Details:

This paper presents an automatic method for obstructive sleep apnea screening and sleep (SL) - wake (WA) stages classification based on the whole night PSG (polysomnography) recording of subjects with sleep-disordered breathing during sleep. The proposed algorithm can identify apnea (AP), hypopnea (HY), and Onset of apnea (OA) events within a period of a respiration cycle. In this work, underlying dynamics embedded in each physiological state (i.e., SL, WA) and different pathological events (AP, HY, and OA) are identified with the help of trajectory formed by assigning three signals, i.e., respiration, EEG, and ECG into three orthogonal axes. The spatial distribution of the trajectory is found to be varied with the different stages and different events of OSA subjects. The discriminating information regarding different stages and events is extracted from the spatial distributions by dividing the whole hyperspace into equal-sized subspaces. The persistence of the trajectory in a particular subspace found to be sensitive to the physiological state and different pathological events. Application of standard Bayesian classifier revealed its ability to discriminate different stages. An encouraging detection accuracy of 95.28% is observed across all the subjects.