I am analyzing a simulated DWI dataset that was provided as part of the ISMRM 2015 Tractography Challenge (http://www.tractometer.org/ismrm_2015_challenge/). I am using the “artifact-free” dataset for this (b-value - 1000 s/mm^2; 33 directions), but have been facing a few issues with the tractography :
- I cannot seem to track the callosal fibers very well. In fact, there seems to be a fair chunk of the corpus callosum that seems to go untracked.
- The corticospinal tract also seems hard to track.
I have attached images illustrating my problem, and have also explained some details below (post-script).
I have not faced these issues when tracking fibers either with our own human subject dMRI data, or with data from other public databases (e.g. HCP). I wonder if this is a specific problem with this artificially constructed dataset. If anyone has insights about this problem based on their experience with analyzing this dataset (or otherwise), I’d greatly appreciate that.
ps: Details of tracking:
I have a 5tt image (
5TT.mif) and brain mask (
dwi_mask.mif) that I generated using the
dwi2mask commands respectively. With these I’ve checked the FODs as well as 5 tissue segmentation mask - these look okay. The FOD lobes seem correctly oriented in the corpus callosum, yet I cannot seem to trace fibers in that region. The specific commands I am using are the following:
Estimate response function:
dwi2response -force dwi.mif response.txt
dwi2fod csd -force dwi.mif -mask dwi_mask.mif response.txt WM_FODs.mif
tckgen -force -act 5TT.mif WM_FODs.mif -seed_dynamic WM_FODs.mif -cutoff 0.02 -minlength 50 -maxlength 300 WB_csd.tck -backtrack -crop_at_gmwmi -select 1000000