Feature-based MRI data fusion for cardiac arrhythmia studies

Current practices in studying cardiac arrhythmias primarily use electrical or optical surface recordings of a heart, spatially limited transmural recordings, and mathematical models. However, given that such arrhythmias occur on a 3D myocardial tissue, information obtained from such practices lack in dimension, completeness, and are sometimes prone to oversimplification. The combination of complementary Magnetic-Resonance Imaging (MRI)-based techniques such as Current Density Imaging (CDI) and Diffusion Tensor Imaging (DTI) could provide more depth to current practices in assessing the cardiac arrhythmia dynamics in entire cross sections of myocardium. In this work, we present an approach utilizing feature-based data fusion methods to demonstrate that complimentary information obtained from electrical current distribution and structural properties within a heart could be quantified and enhanced. Twelve (12) pairs of CDI and DTI image data sets were gathered from porcine hearts perfused through a Langendorff setup. Images were fused together using feature-based data fusion techniques such as Joint Independent Component Analysis (jICA), Canonical Correlation Analysis (CCA), and their combination (CCA+jICA). The results suggest that the complimentary information of cardiac states from CDI and DTI are enhanced and are better classified with the use of data fusion methods. For each data set, an increase in mean correlations of fused images were observed with 38% increase from CCA+jICA compared to the original images while mean mutual information of the fused images from jICA and CCA+jICA increased by approximately three-fold. We conclude that MRI-based techniques present potential viable tools in furthering studies for cardiac arrhythmias especially Ventricular Fibrillation.