CBSU bibliography search
To request a reprint of a CBSU publication, please click here to send us an email (reprints may not be available for all publications)
Effects of normalization approach and global covariates on voxel-based morphometry: Comparing DARTEL and standard SPM approaches using age-related cortical change
PEELLE, J.E., CUSACK, R. & HENSON, R.N.A.
15th Annual Meeting of the Organization for Human Brain Mapping
Year of publication:
Introduction: Objectively characterizing patterns of regional cortical change is of great importance for many areas of neuroscience. This is often accomplished with voxel-based morphometry (VBM) using structural MRI scans that have been automatically segmented into different tissue classes and normalized into a common coordinate system. Here we investigate the effects of two methodological decisions on a VBM analysis of age-related gray matter volume change within a large sample of adults. First, we compared results using "unified" segmentation and normalization (Ashburner & Friston, 2005) with those using DARTEL, a more recent diffeomorphic registration algorithm (Ashburner, 2007). Second, for each type of normalization, we investigated different ways of controlling for global gray matter effects. Methods: Using SPM5, structural MRI scans were analyzed for 330 volunteers aged 18-78 years. Images were first segmented into tissue classes. Images of the gray matter (GM) segment were then normalized to either (1) an MNI template (the "standard" approach, similar to the "optimized VBM" of Good et al, 2001) or (2) to an average template using DARTEL, in both cases modulated and smoothed at 8 mm FWHM. The average of all subjects’ GM segmentations for each approach are shown in Figure 1. Note that the DARTEL average is less blurred than the standard average. For the VBM analyses, age-related change in GM volume at each voxel was modeled using a second-order polynomial expansion of age for each sex. Total GM volume (TGM) varies substantially with both age and sex. To distinguish local from global changes, for both types of normalization three analyses were performed: (1) with no correction for TGM; (2) by scaling the GM volume at each voxel by the TGM for that volunteer; and (3) by including TGM as a covariate in the model. The final analysis (3) removes any local changes that vary linearly with TGM. Importantly, the slope of this linear component is fit independently across voxels. In contrast, analysis method (2) effectively "removes" only a single slope for the linear correlation of TGM across the brain. Results: To assess normalization accuracy, an anatomical landmark was selected in one volunteer (Figure 2, yellow arrow) and compared to three additional randomly-selected individuals. Registration errors were noted in the standard images (a and b in Figure 2) but not in the DARTEL images. Analysis of linear age-related decrease in GM volume with no covariates (Figure 3) showed more focal results in the DARTEL relative to standard normalization (for which "rimming" effects suggested possible registration errors), as well as larger effect sizes. The inclusion of TGM as a covariate dramatically reduced T-values, while the alternative of proportional scaling by TGM resulted in less of a reduction (Figure 4). Conclusions: Normalization using DARTEL provides more focal and plausible statistical results and larger effect sizes than standard normalization in SPM. Due to sizable age-related decreases in total gray matter volume, differing approaches to controlling for this global change significantly impacted the age-related results. We suggest that proportional scaling of gray matter images at the whole-brain level is a better way to isolate regionally-specific effects than voxel-wise covariation. References: Ashburner, J (2007), 'A fast diffeomorphic image registration algorithm', NeuroImage, vol. 38, no. 1, pp. 95-113. Ashburner, J (2005), 'Unified segmentation', NeuroImage, vol. 26, no. 3, pp. 839-851. Good, CD (2001), 'A voxel-based morphometric study of ageing in 465 normal adult human brains', NeuroImage, vol. 14, no. 1, pp. 21-36.