Coupled Object Segmentation Using Entropy and Fuzzy Characteristics of Tissues

In this paper, we introduce a novel method based on level sets and fuzzy c-means clustering methods for the segmentation of medical images. Our new multi-structural method integrates the Herbulot’s entropy minimization and fuzzy forces. We introduce a prior shape based on the tissue segmentation results of each image. This incorporates interactive knowledge-based evolution that increases the accuracy and alleviates the need for the prior statistical information about the shapes of the structures. We have applied the proposed method to segment ventricles, caudate, and putamen in magnetic resonance images (MRI) of the brain. Comparative results show the benefits of the proposed constraints. Accurate results and independency from the initialization are obtained when using the proposed method.