Dear FreeSurfers,

I've been preprocessing, in FS5.3, 3T GE data in controls, obese and diabetes patients and adding the available flair to do the pial refinement as follows:

recon-all -i t1 -s bla -all -FLAIR flair -FLAIRpial

This works excellent for most of the scans. Unfortunately, for some scans adding the flair leads to some interesting refinements, either extending outside of the brain.mgz or going deep into the white matter.

I didn't notice any misregistrations of the flair to the T1 scan when checking the bb-registration, nor were there any errors in the recon-all.log. I have added some screenshots. The red lines represent the pial with the flair refinement, and the blue lines the pial without the refinement. As you can see, these interesting segmentations are not there when the flair is no used.

Does anyone know what is going on?

Thanks a lot,

Eelco