Automatic cardiac RV segmentation using semantic information with graph cuts

We propose a fully automatic method for cardiac right ventricle (RV) segmentation using image features, context information and semantic knowledge using graph cuts. A region of interest (ROI) is first identified and pixels within it are assigned labels (RV or background) using Random forest (RF) classifiers and graph cuts. Semantic information obtained from the trained RF classifiers is used to formulate the smoothness cost. Use of context and semantic information contributes to higher segmentation accuracy than competing methods used on the MICCAI 2012 RV segmentation dataset.