Electroencephalogram based Diagnosis Classification and Music Interventions on Obstructive Sleep Apnea

Abstract

Obstructive Sleep Apnea (OSA) is the breathing disorder where the upper airway newlinegot collapsed during sleeping. Due to the elevated risk of OSA-related illnesses such as newlinehypertension, stroke, and cardiovascular disorders, special attention is needed to OSA newlinepatients. Currently, the existing treatment options include surgery and Continuous Positive newlineAirway Pressure (CPAP) equipment, both of which are time consuming as well newlineas expensive. In this proposed study, a self-administered questionnaire and spectral newlinefeatures from Electroencephalography (EEG) signals based model for screening and newlinediagnosing OSA patients was developed as well the Carnatic music therapy interventions newlinewas analyzed. In the first study, a self-administered questionnaire was developed newlineto screen OSA using standard questionnaires including the Observed, Snoring, Apnea, newlineAge50 (OSA50), STOP-BANG Questionnaire (SBQ), and Berlin Questionnaire (BQ). newlineThe prevalence of OSA was assessed in this exploratory investigation based on age newlinegroup, smoking behaviors, gender, and working hours. Then, the second study utilized newlineEEG recordings from three independent sleep databases such as Sleep European Data newlineFormat (EDF), Cyclic Alternating Pattern (CAP) Sleep, and Institute of System and newlineRobotics-University of Coimbra (ISRUC) with machine learning algorithms to classify newlineOSA patients and normal subjects. Random Forest (RF) and Support Vector Machine newline(SVM) were the classifiers that have been used to classify OSA patients and normal newlineusing EEG signals based on their spectral and nonlinear features such as energy, newlineentropy, heart rate, brain perfusion, neural activity, and synchronization. Hence, the differences newlinebetween normal subjects and OSA patients were assessed in third study using newlinea total of 12 EEG spectral features from real-time EEG recordings such as heart rate, newlinebrain perfusion, neural activity, arousal index, central nervous system arousal, desynchronization