Diffusivity changes are predominantly proportional along all axes with early neurodegeneration in Alzheimer's disease

). 5 (FDT v2.0) was used to correct for eddy currents, fit the tensor and compute the diagonal elements, MD, RD and FA at each brain voxel. Warping errors due to anatomical discrepancies and brain atrophy were minimised by taking advantage of the tract-based spatial statistics (TBSS v1.1) approach 6 ; this method allows voxel-wise statistics to be applied only at the centre of each major fibre bundle defined by a study-specific WM tract skeleton. Nonparametric t-tests of reduced/increased DTI indices in patients were performed using randomise v2.1 7 ; we generated 5000 permutations of the data to test against. Correcting for multiple testing and choosing a meaningful threshold level is also a perennial dilemma in voxel-wise statistics; although there is a mathematical rationale for using falsediscovery rate (FDR) or family-wise error (FWE), determining a critical value for “clinical” significance is much more problematic. We therefore compared three increasingly conservative significance levels: (i) p<0.01 uncorrected for multiple comparisons, (ii) FDRcorrected p<0.05 and (iii) FWE-corrected p<0.05. Cluster-like structures were enhanced using the threshold-free cluster enhancement 8 (TFCE) algorithm. Results and Discussion: TBSS results for increased FA and reductions in diffusivities did not show any statistically significant difference at p<0.01 (uncorrected threshold). Results for reduced FA and increased λ1, RD and MD (Fig. 1) however, highlighted concordant clusters of significance for all three diffusivity indices, unlike FA for which there was an overall lack of sensitivity. Anisotropy changes not only were less extensive, but also were not significant when we controlled the rate of type I errors. In contrast, diffusivity differences were highly significant as they largely survived FDR and FWE correction. These observations are consistent with the hypothesis that early neurodegeneration in AD is associated with proportional changes in any plane of the diffusion ellipsoid. Abnormalities were bilateral and confluent involving the posterior cingulum bundle, parahippocampal gyrus and spreading into lateral posterior temporoparietal WM areas; significant differences were also found in the fornix and the splenium of the corpus callosum. Hypometabolism in adjacent grey matter areas is an established feature in AD 9, 10 ; it is believed that these areas degenerate as part of a