Adaptive weighted fuzzy region based optimization for brain MR image segmentation

Abstract

The recent trends in medical image segmentation and analysis are often used in many real world applications for analyzing different objects of interest. Analyzing the brain Magnetic Resonance (MR) images is found to be difficult task because of the existence of intensity non-uniformity. Although numerous models have been proposed to handle brain MR image segmentation, it is still a challenge to effectively approximate the Intensity Non-Uniformity (INU) and improve greater segmentation accuracy. Hence, an integrated energy minimization approach, namely adaptive weighted fuzzy region based optimization algorithm is developed for brain MR image segmentation. These adaptive fuzzy regions are iteratively weighted to estimate their membership values assigned to each pixel with respect to energy. Also, the optimal weighting parameter, membership values to each region, and bias fields are iteratively estimated and updated. Further, this algorithm is compared with the recent energy minimization approaches in simulated brain MR image dataset. The results of the quantitative evaluations demonstrate that the proposed algorithm gives more reliable segmentation and better accuracy in spite of initialization, noise, and intensity non-uniformity.