An automated method for segmenting white matter lesions through multi-level morphometric feature classification with application to lupus

Abstract

We demonstrate an automated, multi-level method to segment white matter brain lesions and apply it to lupus. The method makes use of local morphometric features based on multiple MR sequences, including T1-weighted, T2-weighted, and fl uid attenuated inversion recovery. After preprocessing, including co-registration, brain extraction, bias correction, and intensity standardization, 49 features are calculated for each brain voxel based on local morphometry. At each level of segmentation a supervised classifi er takes advantage of a different subset of the features to conservatively segment lesion voxels, passing on more diffi cult voxels to the next classifi er. This multi-level approach allows for a fast lesion classifi cation method with tunable trade-offs between sensitivity and specifi city producing accuracy comparable to a human rater. © 2010 Scully, Anderson, Lane, Gasparovic, Magnotta, Sibbitt, Roldan, Kikinis and Bockholt.