A novel model for inhomogeneous brain MR image segmentation

Brain magnetic resonance (MR) images are significant for brain studies because of their excellent contrast of soft tissues, non invasive characteristic and a high spatial resolution. However, Intensity inhomogeneity is an undesired phenomenon that represents the main obstacle for brain MR image segmentation. In this paper, we propose a novel model which can overcome the intensity inhomogeneity problem of Brain MR images without the bias field correction. In this model, a simple and effective initialization method is taken to speed up the curve evolution toward final results; a new multiphase level set method is proposed to segment the brain tissues. This model not only extracts brain white matter (WM), gray matter (GM) and cerebrospinal fluid (CSF) simultaneously, but also provides smooth and accurate boundaries or surfaces of the object. We test our model on two-dimensional and three-dimensional brain MR images and obtain the satisfactory segmentation results. Furthermore, we also use our method to segment nature images and get the ideal results. Experimental results show that our method outperforms the state-of-art methods, yielding higher Tanimoto coefficient.