Sleep Apnea Detection by Slope Analysis

J. Han, H. Shin, D.-U. Jeong, and K. Park (Korea)

Keywords

Sleep, apnea detection, slope analysis, polysomnography, airflow

Abstract

Detection of sleep apnea is the one of the major tasks in sleep medicine. Various methods have been applied to detect sleep apnea by analyzing different bio-signals. This paper introduces an innovative algorithm that analyze slope of nasal airflow signal in polysomnography. In this study, slope is obtained by the moving average of an absolute differentiation of a signal with 3.8 seconds' time window followed by segmentation of 1 second. Apnea threshold is determined by 10% of the slope of normal breathing. If segments having slope below apnea threshold appear for 10 consecutive segments, those segments are marked as apnea. The presented algorithm analyzed polysomnography records from 13 subjects and detected 2500 apnea events. Performance of presented algorithm was evaluated by comparing its results against the sleep specialist's manual scoring on same records. The overall agreement rate between the two was 95.36% (=0.86). In considering its simplicity and lower computational load, the presented algorithm is robust and useful.

Important Links:

Go Back