Do preprocessing algorithms and statistical models influence voxel-based morphometry (VBM) results in amyotrophic lateral sclerosis patients? A systematic comparison of popular VBM analytical methods

Purpose: To study whether inconsistent findings in voxel-based morphometry (VBM) in amyotrophic lateral sclerosis (ALS) brain are due to use of different data preprocessing and statistical methods in two software packages.

Materials and methods: T1-weighted magnetic resonance imaging (MRI) was obtained during routine clinical imaging at 1.5T in ALS patients with frontotemporal dementia (ALS-FTD) (n=18) and in unaffected neurologic controls (n=15). Gray matter (GM) VBM analysis was carried out using FMRIB software library (FSL) 4.1.5 and statistical parametric mapping 8 (SPM8). Comparison of processing steps segmentation, registration, and statistical methods (nonparametric vs. parametric) between the two softwares was performed by subjecting the same dataset through standard VBM processing pipelines.

Results: GM volume was significantly (P<0.05) reduced in motor and extramotor regions of ALS-FTD when compared to controls. Percentage of atrophied GM voxels in the entire brain that reached statistical significance using FSL was 22.52% compared to 0.81% in SPM. Similarly, 0.81% (3308 voxels) reached statistical significance using nonparametric statistics when compared to parametric statistics (0.50%, 2056 voxels).

Conclusion: The differences in GM volume atrophy measures found by FSL and SPM analytic methods indicate that variable results in previous VBM studies may arise from differences in their image processing algorithms and statistical models.