Analysis of the Window Size Effect for T-Wave Alternans Detection Through Machine Learning Methods

T-wave alternans (TWA) is a phenomenon observed in the electrocardiogram (ECG), characterized by a consistent fluctuation of the ventricular repolarization segment. These episodes are regarded as a marker of high risk of ventricular vulnerability and sudden cardiac death. Detection of TWA on ambulatory recordings remains an open issue, so this work addresses the problem using a set of machine learning (ML) methods. Decision Trees (DT), Random Forest (RF) and K-Nearest-Neighbors (KNN) are fed with features extracted from three representative TWA analysis methods, namely the Spectral Method, the Modified Moving Average and the Time Method. Since ambulatory ECG exhibits high variability, this work investigates the impact that the analysis window size has in the performance, so, short-term frames of heartbeats (hb) with different sizes are considered. An ensemble dataset of 750 instances made of real ECG signals with added TWA episodes was utilized. Longer windows showed better performance in terms of F1-score. A non-parametric statistical test demonstrated significant performance increase for windows of 40 hb and 30 hb compared to 20 hb, but not for windows of 40 hb compared to 30 hb. All three ML models can learn from signal excerpts of varying lengths to identify alternant waves of
$35\ \mu V$
.