Segmentation of MRI brain images by incorporating intensity inhomogeneity and spatial information using probabilistic fuzzy c-means clustering algorithm

Segmentation of magnetic resonance imaging (MRI) brain images is an important task to analyze tissue structures of a human brain. Due to improper image acquisition systems, MRI images are generally corrupted by intensity inhomogeneity (IIH) or intensity nonuniformity (INU). Conventional methods try to segment MRI images using only spatial information about the distribution of pixel intensities and are highly sensitive to noise and the IIH or INU. This paper presents a method to segment MRI brain images by considering the INU and spatial information using fuzzy C-means (FCM) clustering algorithm. Firstly, the INU of MRI brain image is corrected using fusion of Gaussian surfaces. The individual Gaussian surface is estimated independently over the different homogeneous regions by considering its center as the center of mass of the respective homogeneous region. Secondly, the IIH corrected image is segmented using probabilistic FCM algorithm, which considers spatial features of image pixels. The experiments using 3D synthetic phantoms and real-patient MRI brain images reveal that the proposed method performs satisfactorily.