Reproducibility and consistency of evaluation techniques for HARDI data

The reproducibility of three evaluation techniques for high angular resolution diffusion imaging (HARDI) data, the diffusion tensor model, q-ball reconstruction and spherical deconvolution, are compared.

Two healthy volunteers were measured in a 3 T MR system six times with the same measurement parameters; one subject was measured with different b-values. The data was evaluated to compare the consistency and reproducibility of reconstructed diffusion directions and anisotropy values for the three investigated diffusion evaluation techniques. The angle difference between the reconstructed main directions of diffusion for the investigated techniques was evaluated. For q-ball and spherical deconvolution the consistency of the second dominant diffusion direction was additionally examined.

The differences between the tensor model and q-ball or spherical deconvolution in the estimated diffusion direction decrease with an increase in fractional anisotropy. Increasing the smoothing kernel in q-ball reconstruction renders the results similar to the ones from the diffusion tensor evaluation. Consistency in the reconstructed directions did increase for larger b-values.

The evaluation of HARDI data in clinical conditions with q-ball or spherical deconvolution shows consistency and reproducibility similar to the diffusion tensor model, but provides valuable additional information about a second dominant direction of diffusion.