Transfer learning for the fully automatic segmentation of left ventricle myocardium in porcine cardiac cine MR images

Abstract

A fully automatic approach for the segmentation of the left ventricle (LV) myocardium in porcine cardiac cine MRI images is proposed based on deep convolutional neural networks (CNN). We trained a 56-layer residual learning CNN (ResNet-56) from scratch on a set of porcine cine MRI images acquired internally, and another CNN via transfer learning by fine tuning a network previously trained on a public human cine MRI dataset. A leave-one-out validation was performed on an 8-specimen porcine cardiac cine MRI dataset (3,600 slices). Comparisons with manual segmentations show that both CNN models are able to produce precise results (99.94% “good” segmentations), while the CNN trained through transfer learning performs better by achieving Dice similarity coefficient (DSC) of 0.86, Hausdorff distance (HD) of 4.01 mm, and overall average perpendicular distance (APD) of 1.04 mm on average.