External Email - Use Caution
Hi again,
At a loss for more things to try here, I tried the following: for the first subject in my sample, I replaced the 'post' scan (01_2.nii.gz) with a copy of the pre (01_1.nii.gz ), to rule out the unlikely cause that an actual geometries problem is at the root here, as reported in the warning/error I mentioned earlier.
While this removed the geometries message from the output of the base command, the error about norm.mgz not being there remained. Further searches of the FS mailing list revealed that this file's absence might be due to a bug in FS v6 (even though I am running 5.3.0, through VirtualBox, on Windows). Nevertheless, I followed the instructions https://www.mail-archive.com/freesurfer@nmr.mgh.harvard.edu/msg57033.htmlfor applying the bugfix patch, but when I redid everything (deleted subject folders, reissued the commands), the result was the same.
It thus seems the base step expects norm.mgz to exist, whereas the previous steps ('-i' and 'cross') do not create it.
I attach the recon-all logs for the pre, post and base of subject 01. Would remain thankful to anyone who can suggest what else I can try!
Best wishes, Tudor
On Thu, 21 Jun 2018 at 10:16, Tudor Popescu tudor3@gmail.com wrote:
Dear FS list,
( Apologies for asking my question again which I am still struggling to resolve and which I tried to clarify and condense below. )
I have pre and post-intervention data for a single group of subjects, and I'd like to extract the cortical thickness at both timepoints, for all standard regions in the atlas. I followed the standard approach https://surfer.nmr.mgh.harvard.edu/fswiki/LongitudinalProcessing#LongitudinalProcessing51.LongitudinalStream.28recon-all-long.29for longitudinal studies, however the third ('base') step immediately gives this error:
*"WARNING: Image geometries differ across time, maybe due to aquisition changes? This can potentially bias a longitudinal study! "*
I don't think the acquisition parameters of the pre-post images actually differ: I could verify that all images indeed have
- identical voxel size (1x1x1)
- identical image size (176x224x160)
- a (0,0,0) point that overlaps well-enough on the anterior commissure
This suggests trivial (if any) "geometry" differences, and at Bruce's suggestion, I used mri_convert to resample each subject's 'post' to their 'pre' baseline:
mri_convert -rl 01_1.nii.gz 01_2.nii.gz 01_2resampled.nii.gz
(and renamed the resampled file to 01_2). However, the base step,
recon-all -base 01 -tp 01_1 -tp 01_2 -all
still gives the same error about geometries, and also in addition, this one:
mghRead(/media/FS/01_1/mri/norm.mgz, -1): could not open file MultiRegistration::loadMovables: could not open input volume /media/FS/01_1/mri/norm.mgz.
Inside /media/FS/01_1/mri/ (where /media/FS is my $SUBJECTS_DIR), there is indeed no norm.mgz, which suggests something in my previous commands for subject1:
recon-all -i 01_1.nii.gz -subjid 01_1 recon-all -i 01_2.nii.gz -subjid 01_2 recon-all -subjid 01_1 -all recon-all -subjid 01_2 -all
..had gone wrong despite these having produced no errors. Although I thought the base step already does autorecon1 as part of the longitudinal pipeline, I tried doing it manually:
recon-all -autorecon1 -subjid 01
however this returns:
ERROR: It appears that this subject ID is an existing base/template from longitudinal processing (-base). Please make sure you pass all necessary flags. For example, if you are trying to re-run a -base: ' -base <templateid> -tp <tpNid> ... '
..whereas running autorecon1 and 2 for the first *timepoint *instead,
recon-all -autorecon1 -autorecon2 -subjid 01_1
returns
mri_em_register: could not open input volume nu.mgz
Should I in fact use, in mri_convert -rl , a *single *target for all the pre&post images in the study, like the study-specific template https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/FSLVBM/UserGuide#C_-_Creating_the_template:_fslvbm_2_template used by VBM? Or what else can I try to get the base step to continue? I'm very grateful for any suggestions!
Cheers, Tudor