Swarm fuzzy inference system and R wave features for ventricular premature beat detection

Abstract

This article introduces a new strategy to detect a ventricular premature beat (VPB). The strategy utilized a swarm fuzzy inference system (SFIS) and features of the R wave of electrocardiogram. SFIS was a FIS optimized using particle swarm optimization (PSO). The PSO was used to find the optimal parameters of the FIS. The fuzzification part of the FIS used a Gaussian function. The inputs of the FIS were the width and the gradient of the R wave. Using clinical data, the proposed strategy performed well for VPB detection with sensitivity, specificity and accuracy of 99.05%, 99.64% and 99.59%, respectively. Keywords—ventricular premature beat, electrocardiogram, particle swarm optimization, fuzzy inference system I. INTRODUCTION A ventricular premature beat (VPB) is an abnormal electrical beat starting in the lower chambers of the heart and disrupts the normal rhythm of the heart. VPB in patients with heart diseases is dangerous as it is associated with life- threatening. Studies reported various techniques for detecting VPB. Nazmy at. al. proposed an adaptive neuro-fuzzy inference system (ANFIS) for VPB detection (1). A fuzzy neural network was investigated for VPB detection by Lim (2) and Shyu (3). In general two aspects were investigated for PVC detection. The aspects were algorithm and its inputs. The algorithm was used to process or to classify the inputs to find an output. Appropriate algorithm and inputs should be found to find a good performance of VPB detection. This article proposed a new strategy for detecting VPB. A swarm fuzzy inference system (SFIS) and features of the R wave of electrocardiogram were investigated. SFIS was a fuzzy inference system optimized using particle swarm optimization (PSO). Fuzzy system showed a good performance for different applications, including for heart problem (4). In this article, R wave features in terms of the width and gradient of the R wave were examined for the input of SFIS for detecting VPB. The rest of this paper is organized as follows. The proposed strategy is described in Section II. Section III and IV presents the experimental results and discussion, respectively. The conclusion for this article is presented in Section V. II. METHOD This article presents a strategy for detecting VPB of electrocardiogram. The strategy is presented in Fig. 1. There were three parts in the strategy: inputs, SFIS and output. The inputs were the width and the gradient of the R wave of electrocardiogram. SFIS was a FIS optimized using PSO. The output was one of two conditions: VPB or non-VPB.