Superpixel Combining Region Merging for Pancreas Segmentation

Pancreas segmentation is an essential predetermination for effective treatment of pancreac organs. To boost the capability of pancreas segmentation, this paper proposes a pancreas segmentation method combining superpixel and region merging. Firstly, a medical superpixel segmentation method is proposed to preprocess the CT image to ameliorate the verge features of the pancreas in the images. Then, this paper uses the New Maximal Similarity-Based Region Merging (NMSRM) way to obtain the pancreas segmentation results on the preprocessed images. We evaluated it on the NIH dataset and compared it with the latest methods, where the Dice-Sørensen coefficient (DSC) of 94%.