Time Series-based Clustering of ECG Heartbeat Arrhythmia using Medoids

Abstract

Abstract: The symptoms of heart diseases are sparse and infrequent. So, the analysis of wearable long-term ECG recordings over hours, days and months is obligatory for detection of these infrequently occurring symptoms of heart diseases that would not be detected with short-term ECG recordings. Manual identification of these heart-beat classes by cardiologists is time consuming and cumbersome. These professionals rely on computer-based classification for deter- mination of these heart-disease types. Partitioning around medoids (PAM also known as K- medoids) clustering using dynamic time warping (DTW) distance method (PAM time series- DTW) using the unequal length (dimensional) full heart-beat time-series is proposed with no explicit feature extraction except PQRST wave detection, which saves lot of time and computation cost.

Index Terms: Clustering, ECG, Heartbeat, Time-warping, AAMI.