12-lead ECG Classification Using Time Series Motifs

2^nd June 2021, 11:30
Hanadi Aldosari

Abstract

In recent decades, there has been a significant growth of interest in automated classification and prediction of many cardiovascular diseases based on various strategies and techniques that employ machine learning. Electrocardiogram (ECGs) are the primary source of information for cardiovascular analysis and classification, normally 12-lead ECGs . Most ECGs classification problems are related to the extraction of relevant features for the posterior induction of classifiers. The ECG feature extraction can take a number of forms such as statistical, wavelet, or any other signal-based approach. This work presents an approach based on motifs to classify ECGs time series data using machine learning. Motifs are frequently recurrent subsequences that are assumed represent relevant patterns. A combination of motifs is used to describe the data and as input to machine learning. A case study is presented using on the China Physiological Signal Challenge data from 2018.