Fully automated pipeline for quantification and localization of white matter hyperintensity in brain magnetic resonance image

Automated white matter hyperintensity (WMH) segmentation on magnetic resonance imaging is greatly advantageous for various clinical studies using large-sample data. Accurate localization of WMH can provide more beneficial information for clinical studies, as differences of regional WMH existence may be linked to clinical symptoms. We suggest a fully automated method for WMH quantification and localization without human interaction using T1-weighted and fluid-attenuated inversion-recovery (FLAIR) images. The known sources of false-positive results in the subarachnoid space and brain-cerebrospinal fluid (CSF) interface were removed by applying a WMH candidate-region mask. WMH segmentation was performed based on the Markov random field model. The intensity-substitution method was developed for the accurate localization of WMH, with proper tissue classification and nonlinear registration. The performance of the method was evaluated via comparison with manual delineation; the similarity index and the overlap ratio were 89.94 and 81.90, respectively. © 2011 Wiley Periodicals, Inc. Int J Imaging Syst Technol, 21, 193–200, 2011 (This work was supported by the Global Science experimental Data hub Center (GSDC) funded by the Korea Institute of Science and Technology Information (KISTI) (N-10-NM-IR05), in which main calculations were performed by using the supercomputing resource of the KISTI.)