• Rose Gelineau-Morel, Valentina Tomassini, Mark Jenkinson, Heidi Johansen-Berg, Paul M Matthews, and Jacqueline Palace.
• Oxford Centre for Functional MRI of the Brain, Nuffield Department of Clinical Neurosciences, University of Oxford, United Kingdom.
• Hum Brain Mapp. 2012 Dec 1; 33 (12): 2802-14.

AbstractPrevious imaging studies assessing the relationship between white matter (WM) damage and matter (GM) atrophy have raised the concern that Multiple Sclerosis (MS) WM lesions may affect measures of GM volume by inducing voxel misclassification during intensity-based tissue segmentation. Here, we quantified this misclassification error in simulated and real MS brains using a lesion-filling method. Using this method, we also corrected GM measures in patients before comparing them with controls in order to assess the impact of this lesion-induced misclassification error in clinical studies. We found that higher WM lesion volumes artificially reduced total GM volumes. In patients, this effect was about 72% of that predicted by simulation. Misclassified voxels were located at the GM/WM border and could be distant from lesions. Volume of individual deep gray matter (DGM) structures generally decreased with higher lesion volumes, consistent with results from total GM. While preserving differences in GM volumes between patients and controls, lesion-filling correction revealed more lateralised DGM shape changes in patients, which were not evident with the original images. Our results confirm that WM lesions can influence MRI measures of GM volume and shape in MS patients through their effect on intensity-based GM segmentation. The greater effect of lesions at increasing levels of damage supports the use of lesion-filling to correct for this problem and improve the interpretability of the results. Volumetric or morphometric imaging studies, where lesion amount and characteristics may vary between groups of patients or change over time, may especially benefit from this correction.Copyright © 2011 Wiley Periodicals, Inc.

21 keywords...

hide…