Markovian segmentation of 3D brain MRI to detect Multiple Sclerosis lesions

This paper proposes a new method to detect multiple sclerosis (MS) lesions on 3D multimodal brain MR images. MS lesions are detected as voxels that are not well explained by a statistical model for normal brain images. These outliers are extracted using the trimmed likelihood estimator (TLE). Spatial regularization is performed using a hidden Markov chain (HMC) model. Tests on real brain MR images with MS lesions have been carried out and results have been compared to manual expert segmentation to validate the proposed method.