Novel Approaches in 12-Lead Electrocardiogram Signal Reconstruction

Presentation Overview: A complete ECG consists of 12 leads of signal. Collecting all 12 signals manually can be time consuming and error pruning; If some signals are missed, physicians need to re-collect the 12-lead ECG again. What is a more efficient, accurate, and convenient way of  collecting 12-lead ECG? Our solution is to apply machine learning models to reconstruct the complete set of ECG using only a few leads of signals that can be collected relatively easily. We developed and tested 7 different models and evaluated their performances in terms of reconstruction accuracy, time, and computational cost.

Authors: Yan Zhu, Chunsheng Zuo, Guanghan Wang, Yunhao Qian, Christopher Cheung