A Supervised Approach for Multiple Sclerosis Lesion Segmentation Using Context Features and an Outlier Map

Automatic multiple sclerosis (MS) lesion segmentation in magnetic resonance imaging (MRI) is a challenging task due to the small size of the lesions, its heterogeneous shape and distribution, overlapping tissue intensity distributions, and the inherent artifacts of MRI. In this paper we propose a pipeline for MS lesion segmentation that combines prior knowledge and contextual information into a boosting classifier. The prior knowledge is introduced in terms of atlas distribution of the main brain tissues while the contextual information is based on a large set of features describing the spatial context in the lesion neighbourhood. Besides, we investigate the inclusion of a probability map describing the likelihood of a voxel to be an outlier, i.e. not being part of any healthy tissue. The experimental results, performed using a set of 30 MRI volumes of MS patients with very different lesion load, shows the feasibility of our approach. Besides, the results demonstrate the benefits of taking the outlier map into account.