Hello experts,

I have been following the FATCAT algorithm in analyzing a DWI dataset, but ran into problems when using Tortoise generated tensors and scalar metrics for tractography on AFNI. However, I heard that this pipeline has been tested extensively and worked well during the AFNI bootcamp. So I'm wondering if I am missing something.

For example, I ran a simple deterministic tractography (3dTrackID, OR, Tortoise generated FA> 0.2, Angle< 60, length >15) with a large corpus callosum ROI. The resulting tract only consists of the middle portion of the CC (connecting left and right hemisphere). The tract is cut off as the CC turns towards the superior direction. I wonder if the vectors have any problems.

Conversely, I took the Tortoise preprocessed DWI and rotated bmatrix, converted the bmatrix to gradient, ran tensor estimation on AFNI (3dDWItoDT), and then ran the same tractogrphy on AFNI. The resulting tract shows the entire CC as expected.

This same problem happened with other ROIs (e.g.. default mode network) and in probabilistic tractography as welll.

My questions are:
1) Am I supposed to use the second method described above for FATCAT (i.e. take the preprocessed DWI from tortoise and run the rest on AFNI)? Is that the way it was done in the bootcamp?

2) Is gradient preferred when running any DWI data on AFNI? I tried using the bmatrix from Tortoise (after making sure it is row-first or diagonal-first format as required by individual AFNI programs), but it did not work as well as using gradient. My understanding is that bmatrix and gradient are equivalent mathematically, except that bmatrix incorporates the bvalue while gradient is unitless. If gradient is preferred, what would be the rationale?

Thank you!

-akwl