Graph cut based automatic prostate segmentation using learned semantic information

We propose a graph cut based automatic method for prostate segmentation using image feature, context information and semantic knowledge. A volume of interest (VOI) is first identified using supervoxel oversegmentation and their subsequent classification of the supervoxels. All voxels within the VOI are labeled prostate or background using graph cuts. Semantic information obtained from Random forest (RF) classifiers is used to formulate the smoothness cost. Use of context and semantic information contributes to higher segmentation accuracy than competing methods.