Physics-Constrained Deep Learning for Robust Inverse ECG Modeling

The rapid developments in advanced sensing and imaging bring about a data-rich environment, facilitating the effective modeling, monitoring, control of complex systems. For example, body-sensor network captures multi-channel information pertinent to electrical activity heart (i.e., electrocardiograms (ECG)), which enables medical scientists monitor detect abnormal cardiac conditions. However, high-dimensional data are generally complexly structured realizing full potential depends great extent on analytical predictive methods. This paper presents physics-constrained deep learning (P-DL) framework for inverse ECG modeling. method integrates physical laws system with infrastructure prediction dynamics. proposed P-DL approach is implemented solve model predict time-varying distribution electric potentials from measured by body-surface sensor network. Experimental results show that significantly outperforms existing methods commonly used current practice.