A Multimodal Cardiac Quiescence Prediction Method based on Artificial Neural Networks

Abstract

A multimodal cardiac quiescence prediction method is developed to improve the prospective cardiac gating accuracy. An artificial neural network is applied to fuse the predicted quiescence derived from the cardiac-motion-based signal, seismocardiography, with that from electrocardiography (ECG). Results from a cardiac patient demonstrate less prediction error when the proposed weighted fusion is applied. Thus, our method bears promise for triggering real-time data acquisition more accurately than solely use ECG. In addition, this method may eventually be applied to other cardiac imaging modalities such as magnetic resonance imaging and computed tomography.