Hi Theresa,
first, if you use the -hires flag in version 5.3 the processing pipeline will not start to downsample your data at some point (if I remember correctly during the normalization stage). As you stated
that you were expecting the automatic downsampling you should either omit the hires flag or you could follow the HiResRecon (https://surfer.nmr.mgh.harvard.edu/fswiki/HiResRecon) to process
your data in your native resolution. You may need to change or add a few flags to process your data with hippocampal subfield segmentation and changes to version 5.3 (compared to 5.1). In case you want your data to be downsampled you should at least use the
-3T flag additionally to correct for inhomogeneities in regards to your 7T data.
Second, the hippocampal subfield segmentation was, as far I read in the mailing list, vastly improved in version 6. As is it still in development, data processing should be done for testing only
and not publication. Version 6 additionally has full support for high resolution data (using the -hires flag should avoid downsampling to 1 mm). It can be downloaded as a developmental version.
Lastly, in regards to your issue, this usually happened to me if the resolution is not exactly 0.7 mm, but something like 0.69998. This results in a rounding conflict between conversion with mri_convert
and mri_nu_intensity.mni. To resolve the issue you have two options: You can either use mri_convert -cs <size in mm> on your dataset prior to feeding it to the recon-all script, this will result in resampling your dataset (using the size of actual voxel size
should avoid interpolation). Otherwise you can use mri_convert --cropsize <x,y,z> to remove one voxel in every dimension prior to the inhomogeneity correction stage.
Best,
Falk
Von: freesurfer-bounces@nmr.mgh.harvard.edu [mailto:freesurfer-bounces@nmr.mgh.harvard.edu]
Im Auftrag von Köbe, Theresa
Gesendet: Mittwoch, 21. September 2016 13:51
An: freesurfer@nmr.mgh.harvard.edu
Betreff: [Freesurfer] Recon hires 7T {Disarmed}
Dear FreeSurfer Developers,
I am attempting to run recon-all on 7T MP2RAGE images (resolution 0.7x0.7x0.7mm³) with hippocampus subfield segmentation. I use the freesurfer version
5.3 so that I expect an automatic down-sampling to 1x1x1mm³.
I used the data directly from the scanner (without reorientation to standard etc.) and used the following command:
recon-all -s $subject.freesurfer -hires -i $subject.mp2rage.nii -expert /media/NeuroMet_Project/scripts/expert.opts -all -hippo-subfields"
As expert option I used only:
mris_inflate -n 15
The processing starts fine but ends with the following error:
ERROR: dimension mismatch between input volume and seg
input 321 321 321
seg 322 322 322
Could you please help me dealing with this problem?
I attached also the log-file
Thank you so much.
Best Theresa
-------------------------------------------------------------------------
Charité – Universitätsmedizin Berlin
NeuroCure Clinical Research Center NCRC
Dipl.-Biol. Theresa Köbe
AG Kognitive Neurologie
Wiss. Mitarbeiter
Charitéplatz 1, D-10117 Berlin
Tel +49 30 450 560 185
Fax +49 30 450 7 560 280
Interne Besuchsadresse: CCM, Sauerbruchweg 5, E2