Magnetic resonance imaging (MRI) can reveal subtle microstructural changes in the heart muscle following infarction, a major cause of sudden cardiac death worldwide. We have developed preclinical models of chronic infarction and used diffusion tensor methods to efficiently probe biophysical MR signals. Here our specific goal was to identify subtle characteristics of the peri-infarct zone, where the mixture of viable cells and collagen fibrils form arrhythmia substrate. We induced myocardial infarction in n=9 Yorkshire swine using an occlusion-reperfusion technique. We occluded the left circumflex artery (LCX) in 5 pigs and the left anterior descendent artery (LAD) in 4 pigs. After 5-6 weeks healing time, high resolution diffusion tensor imaging (∼0.6x0.6x1.2mm spatial resolution) was performed ex vivo in explanted hearts. We then calculated pixel-based fractional anisotropy (FA) in 3 zones: infarct core (dense fibrosis), peri-infarct (arrhythmia substrate) and remote/normal tissue. For analysis, we selected four regions of interests (ROIs) from each tissue zone. A total of 36 ROIs selected from DT images were analyzed and further registered via anatomical markers with corresponding ROIs in histopathology slides. For histology, tissue samples were stained with collagen-sensitive stain to evaluate fibrosis under bright field and polarized light microscopy, as well as CD31 (microvascular integrity) and electrical gap junctions (Cx43). Our results demonstrated a significant decrease in FA (∼25%) in peri-infarct compared to normal myocardium (p<0.05), in agreement with quantitative histopathological and immunohistochemistry findings which showed a reduced CD31 density (∼33%) and Cx43 density (∼56%). These results provide clear evidence of altered microstructure (i.e., fiber disarray due to collagen deposition) and function (i.e., reduced electrical connections and perfusion) in peri-infarct, in large animal models relevant to scar-related human pathophysiology. Future work will focus on using these novel findings to parameterize MRI-based models to predict arrhythmia inducibility.