To see this working, head to your live site.

Search

Mar 11

· Edited: Mar 11# Is tensor-valued encoding really necessary to extract specific microstructural information?

Is tensor-valued encoding really necessary to extract specific microstructural information?

1 comment

0

It depends. If one relies on a biophysical model, the associated constraints may limit the benefit of combining the complementary pieces of diffusion information yielded by multidimensional diffusion MRI acquisitions. However, it seems that these acquisitions can

alleviate degeneracy in certain modelsmake fiber dispersion estimates more robust to model inaccuraciesimprove accuracy on parameter estimationin certain cases.

When considering cumulant expansions of the signal, it has been demonstrated that at least two b-tensor shapes are required for unique estimation of

the covariance tensor in q-space trajectory imagingand that non-axisymmetric b-tensors are required for unique estimation ofthe skewness tensor in diffusion skewness tensor imaging. Also,diffusional variance decompositionrelies on the use of two b-shapes to disentangle the two sources of diffusion kurtosis appearing in the deviations from mono-exponential behavior of the powder-averaged tensor-valued diffusion encoded signal.Finally, for less constrained approaches such as

diffusion tensor distribution imaging, multidimensional diffusion MRI acquisitions provide more robust inversions, as indicated in previous work (seeFigure 1 herefor instance). This specific technique retrieves nonparametric distributions of axisymmetric diffusion tensors. Since such tensors are described by four numbers (isotropic diffusivity, normalized anisotropy, two angles for orientation), it makes sense to use an acquisition scheme comprising four tunable parameters to estimate these numbers. B-tensors do feature four tunable parameters,i.e.the b-value, the b-shape and two angles for orientation, matching the number of "diffusion unknowns".