search

nipreps / smriprep

Structural MRI PREProcessing (sMRIPrep) workflows for NIPreps (NeuroImaging PREProcessing tools)
https://nipreps.github.io/smriprep
Apache License 2.0
131 stars 39 forks source link

Replace FSL FAST with ANTs-Atropos for brain tissue segmentation #109

Open effigies opened 5 years ago

effigies commented 5 years ago

Please see discussion in Neurostars #4859.

Summary:

1) ANTs N4 + fast -N results in poor segmentation 2) ANTs Atropos seems to work reasonably well 3) N4 + fast seems to work reasonably well, as FAST has built-in bias correction

Options:

1) Switch to Atropos - this has the advantage of removing an FSL, with its licensing issues 2) Use FAST, keeping bias correction in for safety, presumably at some cost 3) Give an option 4) Use FreeSurfer segmentation instead, when available

oesteban commented 5 years ago

I would go for 4 and fall back to 1 for --fs-no-reconall.

oesteban commented 4 years ago

Option 4 is now implemented via #180.

I leave this open, and edit the title to be explicit about option 1.

lucsanara commented 2 months ago

Hello!!

We are running aslprep, with freesurfer preexisting outputs. We have followed the node of the process and we thought that for tissue segmentation it was using fast thought, now having read you we are doubting. Indeed, we were wondering how to improve tissue segmentation. So, how can we make sure that the pipeline is using freesurfer preexisting outputs or ANTS atropos and not FSL fast?

effigies commented 2 months ago

The tissue segmentation is still FAST.