Hello everyone,
I am currently trying to process an old dMRI dataset using TractoFlow, and I would appreciate some expert advice.
The dataset is a pseudo-multishell acquisition with a very unbalanced distribution of directions across shells: ~30 directions at b=1000, ~5 directions at b=2000, and ~5 directions at b=3000.
My goal is to determine whether this dataset is still usable for tractography with TractoFlow.
1. I first ran TractoFlow with the following parameters:
nextflow run tractoflow/main.nf --input ~/path/to/i_tractoflow --dti_shells “0 1000” --fodf_shells “0 2000 3000” -with-singularity containers_scilus_1.6.0.sif -profile fully_reproducible,bundling
The resulting tractography showed many anatomically implausible streamlines and a large number of false positives.
2. I then tried treating the data as single-shell and computing the FODF only from the b=1000 shell:
nextflow run tractoflow/main.nf --input ~/path/to/i_tractoflow --dti_shells “0 1000” --fodf_shells “0 1000” -with-singularity containers_scilus_1.6.0.sif -profile fully_reproducible,bundling
I tested this configuration with both deterministic and probabilistic tracking.
I am wondering whether using only the b=1000 shell for FODF is the right approach, whether deterministic or probabilistic tracking would be preferable, or whether this dataset may simply not be suitable for tractography.
Any recommendations on the best strategy for tractography in this situation would be very helpful.
Thanks in advance for your help!