Epicardial adipose tissue (EAT) and its fatty acid composition (FAC) have been implicated in numerous cardiovascular diseases as saturated fatty acids are known to promote inflammation. FAC MRI techniques, while prominent, have not been applied to EAT due to extended scan times, thus, image acceleration is essential. Here, we demonstrate compressed sensing with a signal model-based dictionary (CS-DICT) to reconstruct psuedo-random undersampled images. Using CS-DICT, we achieve rate-3 acceleration while maintaining accurate EAT FAC maps and estimations in obese mice. These methods facilitate the application of FAC MRI to the EAT.