A sensitive, noise-resistant method for identifying focal demyelination and remyelination in patients with multiple...

Magnetization transfer imaging (MTI) provides a reliable and histopathologically validated means for identifying important tissue changes in multiple sclerosis (MS), including demyelination and remyelination. However, most approaches to date have been based on a priori regions of interest (ROIs) and have been relatively insensitive to small focal changes or competing processes. More recent techniques have sought to address this through a voxel-wise approach, but have been limited in their detection capabilities by the amount of noise in standard MTR images. To address this issue while remaining sensitive to local changes, we propose the use of the recently introduced threshold-free cluster enhancement (TFCE) technique in combination with a Monte Carlo estimation approach. TFCE is first applied to enhance individual voxels based on their level of local cluster support, and then Monte Carlo estimation is performed to allow meaningful statistical interpretation of the resulting TFCE values. We validated this technique in three complementary ways: healthy control scan-rescan analysis, analysis of a "gold standard" simulated dataset, and analysis of a group of MS patients and healthy volunteers with 1-year longitudinal MRI scans. Scan-rescan analysis demonstrated a very low false-positive rate (1.44 mL increasing and 1.48 mL decreasing at the optimal detection threshold). Simulated dataset analysis yielded an area under receiver-operating characteristic curve of 0.942 (compared to 0.801 for a more conventional voxel-wise thresholding analysis). Finally, analysis of the real subject population showed highly significant differences (p<0.001) in volume of decreasing MTR between patients and controls. The proposed method provides a valuable means for quantifying MS-related tissue changes, particularly demyelination and remyelination, in vivo and without the use of highly complex or experimental MRI acquisition techniques. It improves on the sensitivity of other approaches, and may increase the statistical power of studies investigating the effects of therapy on MRI outcomes in MS.