Amplitude variability extraction from multi-lead electrocardiograms for improvement of sleep apnea recognition

This study is in the context of sleep apnea recognition from multi-lead ECGs. In 38 patients, 8-channel ECGs were recorded simultaneously to a polysomography (PSG). The ECG was classified in segments of one minute for occurrence of sleep apnea events by quantification of the regularity of characteristic oscillations in either heart rate or ECG amplitude. Diagnostic accuracy is compared by ROC-analysis against the expert annotations of the PSG, and its reproducibility was tested on the Physionet apnea ECG database. Whereas amplitude variations yield consistent results on both data sets (ROC-area 89.0% vs. 88.3%), a remarkable loss in performance is observed for heart rhythm (89.8% vs. 77.9%). Reasons for this difference are discussed and it is shown that factors like diabetes have a confounding influence on heart rate. With respect to aggregation of multi-lead information, simple averaging (89.3%) seems to be as appropriate as more complex PCA-based methods (87.2%)