Effect of MRI coregistration on serial short-term brain volume changes in multiple sclerosis.

Objective: To test the effect of serial magnetic resonance (MR) coregistration on short-term brain volume changes using different semiautomated and automated brain volume techniques in patients with relapsing-remitting (RR) multiple sclerosis (MS). Coregistration is frequently used to increase precision in serial MR imaging (MRI) analyses. However, the effect of coregistration on measurement of whole brain volume changes from serial scans in the short term has not been tested in MS patients. Methods: Twenty-eight patients with RR MS [mean disease duration: 4.9 years, mean age: 34.4 years and mean expanded disability status scale (EDSS): 1.4] were scanned at baseline and monthly for a period of 3 months with 2D spin-echo T1-weighted sequences obtained with nongapped 3 mm axial slices. Percent brain parenchymal fraction change (PBPFC) was calculated by a semiautomated (Buffalo) and, separately, by two automated (Buffalo automated and SIENAX) techniques, whereas percent brain volume change (PBVC) was calculated by the SIENA technique. For coregistration of serial images we used a robust, fully automated linear image coregistration tool. PBPFC and PBVC were calculated before and after coregistration, comparing scans from the following time periods: (1) baseline to month 3; (2) baseline to month 1; (3) month 1 to 2 and (4) month 2 to 3. Results: The highest median PBPFCs measured on non-coregistered images were detected for the baseline-to-month-3 time period and ranged from -0.11% for Buffalo semiautomated to -0.45% for Buffalo automated (p=ns). On coregistered images, the highest PBPFCs were detected for the baseline-to-month-3 time period and ranged from 0.3% for Buffalo semiautomated, -0.3% for Buffalo automated, 0.02% for SIENAX and -0.02% for SIENA (PBVC). At all time points of the study, no significant differences of median volume changes were measured on coregistered and non-coregistered images when comparing the results among the segmentation algorithms. Conclusions: Over a 3 month period we did not detect short-term changes in normalized brain volumes using different measurement techniques. A longer observation period is needed to assess whether coregistration can affect the measurement of long-term brain volume changes. [ABSTRACT FROM AUTHOR]