Voxel based versus region of interest analysis in diffusion tensor imaging of neurodevelopment

There are two main methods of quantitative analysis in diffusion tensor imaging (DTI) studies: manual region of interest (ROI) and automated voxel based. The purpose of this study is to compare the results of each of these methods applied to the same data set. Linear correlative analysis was performed for mean diffusivity (Trace/3 ADC) and for fractional anisotropy (FA) versus age within 8-12 years (N = 32) and within 21-27 years (N = 28), as well as a group comparison. SPM analysis identified more structures changing with age, partly due to the limited regions measured with ROI analysis. In general, ROI and voxel-based analysis methods produced comparable results for widespread reductions of Trace/3 ADC and increases of FA with age, particularly for group comparison. The discrepancies (i.e., missed regions) were likely related to problems of spatial normalization for SPM analysis, and masking localized changes by averaging all the voxels within a region of interest for ROI analysis. These two analysis methods for DTI offer complementary results, but neither one yields the complete story of neurodevelopment.