Premature ventricular contraction detection using swarm-based support vector machine and QRS wave features

Abstract

A novel strategy for detecting Premature Ventricular Contraction (PVC) is proposed and investigated. The strategy employs a Swarm-based Support Vector Machine (SSVM). An SSVM is an SVM optimised by using Particle Swarm Optimisation (PSO). The strategy proposes new inputs. The inputs involve the width and the gradient of the electrocardiographic QRS wave. Experiments with different inputs and different SVM kernel functions are conducted to find the best one for PVC detection. On a test using clinical data, SSVM performs well in PVC detection with sensitivity, specificity and accuracy of 98.94%, 99.99% and 99.46%, respectively.