Ofer Pasternak1, Nir Sochen2, Peter Joel Basser3
1Blavatnik School of Computer Science, Tel Aviv University, Tel Aviv, Israel; 2Department of Applied Mathematics, Tel Aviv University, Tel Aviv, Israel; 3Section on Tissue Biophysics & Biomimetics, Eunice Kennedy Shriver National Institute of Child Health and Human Development, NIH, Bethesda, MD, USA
We study the question of metric selection for diffusion tensors by applying a tensor-variate statistical framework. The Log-Euclidean metric, which represent the affine-invariant metric family, is compared with the conventional Euclidean distance. By calculating variability maps for synthetic and real DTI data we show that the Log-Euclidean distance does not adequately model the effect of Rician noise in diffusion weighted imaging data. We suggest that the Euclidean metric provides variability maps in coherence with the expected type of noise.