A Local Information-Based Fuzzy C-Means for Brain MRI Segmentation

Abstract

Segmentation of brain tissues from magnetic resonance imaging (MRI) is crucial for quantitative analysis of brain images. The fuzzy c-means (FCM) algorithm has proven to be an efficient approach for brain MRI segmentation. However, accurate segmentation results are hard to find due to the presence of noise. In this paper, we apply most commonly used local denoising filters to preprocess the image for obtaining better segmentation results. Then, we quantitatively compared various FCM-based state-of-the-art segmentation approaches with the proposed methods using Jaccard similarity (JS) and similarity index (ρ) on the simulated and clinical MRI. It is observed from the comparison results that the proposed methods provide more accurate segmentation results than the existing methods.