Dear FreeSurfer developers,
I'm trying to apply Fujimoto's pipeline to reconstruct surfaces of MP2RAGE structural images.
The
MP2RAGE sequence generates several images: a UNI or flat image (T1w but
with amplified noise outside the brain), a proton density image, and
other images.
To be able to run freesurfer on the flat image,
Fujimoto invokes freesurfer two times and then stops them immediately
after the brain masking step: the primary stream (input: the flat image)
and the secondary stream (the multiplicative product of the flat image
and the proton density, which removes the noise). Then, the brain mask
generated during the primary stream is replace by the mask generated
during the secondary stream, and the primary stream is resumed. I
implemented this as follows:
- run 'recon-all -s X -i Y -autorecon1' for the primary and secondary streams
- copy the brainmask.mgz file of the secondary stream to the primary stream (and replace the file)
- resume the primary stream with 'recon-all -s X -autorecon2 -autorecon3)
The
problem with this last step is that it gets stuck at correcting defects
(eg: CORRECTING DEFECT 0). Sometimes it adds "Xl defect detected". Each of these corrections takes several days
to run and this is not practical for our research.
Any ideas on how I could fix this? I have attached a recon-all.log file (I stopped the process because it takes too long).
FS version: freesurfer-Linux-centos6_x86_64-stable-pub-v6.0.0-2beb96c