Automated segmentation of white matter lesions in 3D brain MR images, using multivariate pattern classification

This paper presents a fully automatic white matter lesion (WML) segmentation method, based on local features determined by combining multiple MR acquisition protocols, including T1-weighted, T2-weighted, proton density (PD)-weighted and fluid attenuation inversion recovery (FLAIR) scans. Support vector machines (SVMs) are used to integrate features from these 4 acquisition types, thereby identifying nonlinear imaging profiles that distinguish and classify WMLs from normal brain tissue. Validation on a population of 45 diabetes patients with diverse spatial and size distribution of WMLs shows the robustness and accuracy of the proposed segmentation method, compared to the manual segmentation results from two experienced neuroradiologists.