Efficient 3D Deep Learning for Myocardial Diseases Segmentation.

Automated myocardial segmentation from late gadolinium enhancement magnetic resonance images (LGE-MRI) is a critical step in the diagnosis of cardiac pathologies such as ischemia and myocardial infarction. This paper proposes a deep learning framework for improved myocardial diseases segmentation. In the first step, we build an encoder-decoder segmentation network that generates myocardium and cavity segmentations from the whole volume, followed by a 3D U-Net based on Shape prior to identifying myocardial infarction and myocardium ventricular obstruction (MVO) segmentations from the encoder-decoder prediction. The proposed network achieves good segmentation performance, as computed by average Dice ratio overall predicted substructures, respectively: ’Myocardium’: 96.29%, ’Infarctus’: 76.56%, ’MVO’: 93.12% on our validation EMIDEC dataset consisting of LGE-MRI volumes of 16 patients extracted from the training data.