P-wave shape changes observed in the surface electrocardiogram of subjects with obstructive sleep apnoea

Automated detection of sleep apnoea from the surface ECG is a challenge with potentially significant clinical benefit. The objective of this study is to explore a possible relation between P-wave shape changes and the presence of Obstructive Sleep Apnoea (OSA) episodes in ECG records. Our working hypothesis is that apnoea-related phenomena (e.g., hypoxia, changes in intra-thoracic pressure) influence the depolarisation dynamics of the atria, and hence lead to detectable shape changes in the surface P-wave. For this purpose, a P-wave shape clustering algorithm is proposed. It uses the k-means method, an unsupervised classifier, coupled with a shape metric, the CISA (Corrected Integral Shape Averaging) distance, to distinguish apnoeic and non-apnoeic epochs. The method was applied to perform two-clustering on P-wave shapes extracted from ECG segments. The results obtained indicate that shape information from the ECG can be used in a classification approach for OSA detection. The study also raises the question of elucidating the mechanisms of how OSA may alter P-wave shape