Vectorcardiographic Reconstruction from Standard 12-Lead Electrocardiogram Using Convolutional Neural Network

Vectorcardiography (VCG) is an alternative method of electrocardiography which provides useful spatial information for the cardiologists to diagnose and localise cardio-vascular diseases with higher accuracy. However, the method of measuring the VCG requires a different lead system which makes the monitoring devices more complicated. Thus, the method of reconstructing VCG from standard 12-lead ECG, which is more ubiquitous, was developed in past years. In this paper, we proposed a novel method of reconstructing VCG from standard 12-lead ECG by using convolutional neural network (CNN) which adds temporal information to the training process. The result indicates that adding convolutional layer before feeding into fully-connected MLP layers improves the performance of model significantly from the conventional methods.