Sleep Apnea Event Detection System Based on Heart Rate Variability Analysis

Abstract

Sleep apnea is a disease related to sleep-disorder breathing. This work proposes a neural network sleep apnea detection system based on heart rate variability (HRV) analysis. HRV features are extracted from heart rate data derived from electrocardiogram (ECG) database. The neural network classifies each set of HRV features into sleep apnea event or non-sleep apnea event. This work analyzes the effect of window size on feature extraction, heart rate correction and feature selection on the detection performance. The feature selection algorithm applied in this work is correlation-based feature selection. Experimental result shows the maximum detection performance at 84.52% sensitivity, 83.72% specificity and 83.99% accuracy with HRV features extracted with 3 min corrected heart rate data.