Identification of patients at risk of cardiac conduction diseases requiring a permanent pacemaker following TAVI procedure: A deep-learning approach on ECG signals

Abstract

This research aims to investigate the feasibility in predicting, at several stages of the transcatheter aortic valve implantation (TAVI) procedure, the risk to develop cardiac conduction defects (CCD) requiring a permanent pacemaker installation. Adopting deep-learning techniques specifically designed for time series signals, 2197 electrocardiograms for 631 TAVI patients have been analyzed. The most important leads for the perioperative stages have been identified. The experimental results have shown that more information is available during- and in the post-procedure stages. The highest AUC results have been obtained on the following leads and stages: 0.635 (lead V1) at pre-procedure stage, 0.711 (lead III) at during-procedure stage and 0.721 (lead aVF) at post-procedure stage. Overall, the highest AUC result has been obtained for the post-procedure stage reaching an AUC of 0.740 for the lead V1. This study confirms that it is possible to identify and stratify even further, the risk for patients that will develop CCD after TAVI, by using the information available in the ECG data immediately following the procedure.