hey, does anyone know how fMRIPREP deals with soft failures in freeesurfer? like number of holes on the surface, etc
If recon-all fails, fMRIPrep should fail out. If you then fix things up, you should be able to resume using the same output and working directories.
Hey effigies, thank you for your reply! i checked on the Andy’s brain book website and they says:
‘Soft Failures, on the other hand, occur when recon-all has finished without any errors - but upon closer observation of the output there are obvious errors that need to be fixed. These include large holes in the white matter (“white matter segmentation errors”) and large discrepancies between the border of the white and grey matter of the anatomical volume and the reconstructed surface (“pial surface errors”).’
FreeSurfer Tutorial #12: Introduction to Failure Modes — Andy’s Brain Book 1.0 documentation (andysbrainbook.readthedocs.io)
just wondering if this would be a big problem?
Got it. For soft failures, it’s possible there will be downstream failures in fMRIPrep that result. If you fix up FreeSurfer results yourself and rerun fMRIPrep, it will pick up the fixes and rerun the appropriate nodes, assuming you still have the working directory. If you don’t have the working directory, it will also pick up the fixes, but will also spend a bunch of time redoing things that went okay.