Reconstruction of 12-Lead Electrocardiogram Based on GVM

Abstract

With the development of intelligent hardware, people paying more attentions to health diseases, household electrocardiogram (ECG) monitor is becoming more and more popular. However, too many leads bring inconvenience to people. Reconstruction of 12-lead ECG system by few lead signals becomes a better solution. This paper applies general vector machine (GVM) to achieve 12-lead ECG system synthesis by using 3 subset leads (leads I, II and V2). Moreover, based on the study of 72 representative subjects selected from the PTB diagnostic ECG database, we compared GVM model with multiple linear regression (MR) method, backpropagation (BP) neural network and genetic algorithm optimized BP (GA-BP) neural network. In the effect of the reconstruction of the 12-lead system, GVM model has a maximum of 11% performance improvements than MR method, 5.5% performance improvements than BP neural network and 1.5% performance improvements than GA-BP neural network.