Bias field estimation and segmentation of MRI images using a Spatial Fuzzy C-means algorithm

Abstract

Magnetic resonance imaging (MRI) images suffer from intensity inhomogeneity or bias field causes due to smooth intensity variations of the same tissue across the image region. This paper presents a new method called Bias Estimated Spatial Fuzzy C-means (BESFCM) algorithm for intensity inhomogeneity estimation and segmentation of MRI images at the same time. First, we formulate a new local fuzzy membership function that includes a probability function of a pixel considering its spatial neighbourhood information. Then, we introduce a new clustering center and weighted joint membership functions using the local and global membership values. Finally, MRI images are segmented and bias field is estimated by formulating an objective function using the new cluster centers and joint membership functions. The simulation results show that the resulting BESFCM algorithm estimates intensity inhomogeneity and improves the segmentation results as compared to other FCM-based clustering algorithms.