Detection of sleep apnea from multiparameter monitor signals using empirical mode decomposition

Abstract

For diagnosing obstructive sleep apnea (OSA), polysomnography (PSG) is used. Use of PSG is gold standard for detection of sleep apnea. This research is basically aimed at detection of sleep apnea from more commonly available physiological signals such as electrocardiogram (ECG) and photoplethysmographic (PPG) signals in any simple bedside multiparameter monitors. Respiratory activity extracted from ECG and PPG signals is used for the detection of apnea episodes. This process is useful in situations when recording of PSG is not possible or as a preliminary screening test of possible OSA in patients. In the present work ECG-derived respiration (EDR) and PPG derived respiration (PDR) signals, obtained using empirical mode decomposition (EMD) method, and are used to detect OSA episodes. Signals from MIMIC database were used for experimentation. The test results have revealed that the proposed method has efficiently extracted respiratory information from ECG and PPG signals for detection of obstructive sleep apnea syndrome (OSAS). The similarity parameters computed in both time and frequency domains have confirmed the same. High sensitivity and positive predictivity levels have revealed high degree of correctness.