Optimization of wavelet transform in the task of intracardiac ECG segmentation

Abstract

My work deals with the selection of an appropriate wavelet transform setting for feature extraction from intracardiac ECG recordings. The studied signals were obtained during electrophysiological examinations at the Department of Pediatric Medicine, University Hospital Brno. In this paper, several wavelets are tested for feature extraction which is followed by adaptive thresholding to detect atrial activity from the extracted features. The procedure is evaluated using the F-score. Although the presented procedure does not appear to be overall effective for intracardiac signal segmentation, it certainly does not reject the use of wavelet transforms in combination with advanced machine learning, neural network, or deep learning techniques.