This project aims to develop a deep learning (DL)-based automated segmentation of epicardial and paracardial adipose tissue (EAT/PAT) in cardiac MRI data from the population-based “Hamburg City Health Study”. Based on these segmentations, we will evaluate the predictive power of EAT / PAT volume and fat characteristics (based upon T1 relaxation times) for cardiac scars and adverse cardiac events, such as myocardial infarction. EAT, the fatty tissue between the heart muscle and the pericardium, and PAT, the mediastinal fat outside the pericardium, are considered risk factors for cardiovascular disease. While quantification options exist for CT, they do not yet exist for the gold standard of cardiac function diagnostics: MRI. However, the automatic quantification of EAT and PAT is essential for an individualized risk classification with regard to myocardial infarction and necessary to initiate preventive measures at an early stage.