Karki, Susan

Phonocardiogram (PCG) signals contain very important information regarding the heart condition. Recently, several automatic detection algorithms have been explored to profile the characteristics of heart sounds to aid in disease diagnosis. However, many of these methods has been demonstrated only on clean signals with limited test data and variety of PCG signals that can accurately provide information of diagnostic importance with higher sensitivity and specificity. In this work, we propose to characterize the multiscale frequency state of the normal PCG signals that can aid in accurate profiling of PCG to discriminate from pathological conditions.

Sleep apnea is characterized by abnormal interruptions in breathing during sleep due to partial or complete airway obstructions affecting middle-aged men and women on an estimated ∼4% of the population [1]. While the disorder is clinically manageable to relieve patients, the challenge occurs with diagnosis, with many patients going undiagnosed leading to further complications such as ischemic heart diseases, stroke etc. Sleep apnea also significantly affects the quality of day to day life causing sleepiness and fatigue. Polysomnography (PSG) technique is currently a used for detecting sleep apnea which is a comprehensive sleep test to diagnose sleep disorders by recording brain waves, the oxygen level in the blood, heart rate, breathing, eye and leg movements during the study. However, PSG test is very expensive, requires patients to stay overnight and is known to cause inconvenience to the patients.