An automatic screening approach for obstructive sleep apnea diagnosis based on single lead electrocardiogram
1. AN AUTOMATIC SCREENING APPROACH FOR OBSTRUCTIVE
SLEEP APNEA DIAGNOSIS BASED ON SINGLE-LEAD
ELECTROCARDIOGRAM
ABSTRACT
Traditional approaches for obstructive sleep apnea (OSA) diagnosis are apt to using
multiple channels of physiological signals to detect apnea events by dividing the signals into
equal-length segments, which may lead to incorrect apnea event detection and weaken the
performance of OSA diagnosis. This paper proposes an automatic-segmentation-based screening
approach with the single channel of Electrocardiogram (ECG) signal for OSA subject diagnosis,
and the main work of the proposed approach lies in three aspects: (i) an automatic signal
segmentation algorithm is adopted for signal segmentation instead of the equal-length
segmentation rule; (ii) a local median filter is improved for reduction of the unexpected RR
intervals before signal segmentation; (iii) the designed OSA severity index and additional
admission information of OSA suspects are plugged into support vector machine (SVM) for
OSA subject diagnosis. A real clinical example from PhysioNet database is provided to validate
the proposed approach and an average accuracy of 97.41% for subject diagnosis is obtained
which demonstrates the effectiveness for OSA diagnosis.