Sparsity based morphological characterisation of heartbeats

The sparsity of the representation of a heartbeat as a parameter for its morphological characterisation is considered. The approach relies on greedy pursuit strategies and dedicated dictionaries learned from examples of the classes to be identified. The dictionary rendering the smallest sparsity value characterises the morphology of that beat. The study focuses on a procedure for learning the dictionaries and compares several metrics of sparsity for morphological identification of heartbeats on the basis of those metrics. The suitability of the method is illustrated by binary differentiation of Normal and Ventricular heartbeats in the MIT-BIH Arrhythmia data set. In intra-patient classification the sensitivity score for Ventricular beats is 97.6%. In inter-patient assessment this score drops to 92.4 %. The results are competitive with the state of the art for both assessment schemes. This is encouraging, because the proposed binary identification is realised outside the usual machine learning framework. Thus, extensions of the approach to allow for combination with other features and other machine learning techniques are readily foreseen.