Hi all,

I have some questions related to tractography. From DTI metrics we calculate FA, eigen values (by decomposition of DTI), MD. My query is which information is used for fiber reconstruction is it the DTI or the eigen values?

Hi all,

I have some questions related to tractography. From DTI metrics we calculate FA, eigen values (by decomposition of DTI), MD. My query is which information is used for fiber reconstruction is it the DTI or the eigen values?

Different tools use different methods, here is a good reference. For example, seminal tensor-based approach of Basser et al. 1994a generates streamlines in voxels where the fractional anisotropy (FA) exceeds a threshold and the angle between conjoining principal vectors is less than another threshold. This approach only requires a simple tensor acquisition. Longer and more sophisticated acquisitions will allow one to derive additional parameters that can be used to generate fibers.

Hi @Chris_Rorden . I have a query. Here, the principal vectors you mean is the highest eigen vector in a voxel among the three eigen vectors right?

Regards,

Ashwin

The principal eigenvector (V1) is the eigenvector with the largest associated magnitude, or eigenvalue (λ1).

Hi @Chris_Rorden @Steven. I have a question. It might sound nonsense, but I wanted to know about the possible outcomes. What will happen if we follow the other two eigenvectors instead of the principal eigen vector?

Also, can we have different thresholds of FA like say FA > 0.7 follow the eigen vector and for FA between 0.2 and 0.7 (0.2 < FA < 0.7) take the resultant of the two eigen vectors (V1 and V2).

The other two eigenvectors point perpendicular to the principal eigenvector. They will point in complete opposite directions than the supposed white matter. Do not advise

I suppose it is possible, albeit with a bit of coding that would be difficult given how the tracking software is built. But I wouldn’t recommend it. Here is why: if you are using *probablistic* tracking, the full shape of the tensor is taken into account when making possible streamlines. This is in opposition to *deterministic* tracking, which will only use the principal eigenvector in each voxel. However, V2 doesn’t have any tracking-relevant info to it, so using it could confound deterministic tracking to.

MANAGED BY INCF