Standard 12-Lead ECG Synthesis Using a GA Optimized BP Neural Network

Fangjian Chen,Yun Pan,Ke Li,Kwang-Ting Cheng,Ruohong Huan
DOI: https://doi.org/10.1109/icaci.2015.7184716
2015-01-01
Abstract:This paper presents a method to reconstruct the standard 12-lead ECG from a 3-lead subset (I, II and V2) by optimizing the back propagation neural network with genetic algorithm (GA-BP). The non-linear method BP network is more suitable for ECG signal processing and GA is utilized to optimize the initial settings of the weights and biases in the BP network. Based on the results experimented on the study population of 39 subjects randomly selected from the PTB diagnostic ECG database, the proposed GA-BP method is proved to achieve accurate synthesis of the standard 12-lead ECGs, showing significant improvements over the common BP network (p≤0.001) and linear transformation method (p≤0.001) in terms of correlation coefficient values and root-mean-square errors.
What problem does this paper attempt to address?