Unexpected FOD Orientation in scil_tracking_local with ISMRM-computed FOD

[tensor2023]

Hello,

I encountered an issue when using scil_tracking_local with an FOD computed from ISMRM data. Regardless of whether the FOD is generated using MRtrix3 or a deep learning method, when I input the FOD with the b-vector flipped along the x-axis or without flipping, the resulting tractography (Figure 1) does not follow anatomical correctness. However, when flipping the b-vector along the y-axis, the tractography appears reasonable (Figure 2), but in this case, the FOD in another view (Figure 3) still does not align with anatomical expectations. I would like to understand why this happens, and I hope that when the FOD appears correct, the corresponding tractography also follows anatomical correctness.

I would really appreciate any insights or suggestions on how to properly handle the b-vector flipping to ensure anatomically accurate tractography. Thank you in advance for your time and help!

[AntoineTheb]

Hey @tensor2023 , sorry for the late reply.

There is potentially a couple of things at play here:

• MRTrix tends to mishandle the tractometer's bvecs, resulting in a flip in x. See this page: https://tractometer.org/ismrm2015/dwi_data/ or this forum: https://community.mrtrix.org/t/mrconvert-flips-gradients/581
• MRtrix and scilpy/dipy use different SH bases which are not interchangeable or inter-compatible. MRTrix only supports their basis, while scilpy should support all of them. However, when calling scil_tracking_local, you should specify the correct basis using the --sh_basis option. Otherwise you will end up with flips.
• If the tracking is not broken, as in your first figure, then the ODF are probably properly handled. However, when visualizing, you may also run into SH basis issues. This looks like mrview ? If your FODs come from mrtrix then it should be compatible, but otherwise you will see flips.

Hope this helps and this answer is not too late !