Fig. 7.

The relationship between tract similarity difference (based on spatial correlation between two aligned tract density images) and warp deformation difference for the Rotterdam data (dashed) and the Oxford data (solid), computed for FNIRT operating at optimum parameters for each dataset. The largest difference is obtained when scaling the warp with a factor of 0.8. This translates into a similarity drop between fully and partially warped tracts. Deformation difference is computed by taking the deformation difference (vector) image for each subject, comparing the full and partial warps. The deformation difference in the graph is then the median Euclidean deformation difference distance (vector length), averaged over all subjects in each dataset.