An Automatic Tumor Segmentation System of Brain Tumor from MRI Based on A Novel Energy Functional

Abstract

In this paper, a novel energy functional for the level set model is proposed, which give consideration to Region and Edge characteristics of an image simultaneously (so it is called R-E model). This model is adept in dealing with heterogeneous features of MRI. Comparisons and experiments validate the performance of the R-E model. Aiming to the sensitivity to initial contours of all the active contours models including R-E model, an automatic segmentation system is developed to identify tumor-dominant slice and set proper initial contour automatically. And then R-E model is applied to segment tumor’s contours from all MRI slices autonomously. The automatic segmentation system is applicable to all active contour models. MRI brain tumor data are applied to validate the R-E model and the automatic segmentation system.