External Email - Use Caution
Hi,
Following up in reporting that it does appear to be an issue with Nibabel and not FreeSurfer. mri_diff between the metric files reveals that they are not empty. I have found an alternative for loading the morphological metrics in python, so consider this issue solved, though I am curious why the files are all the same sizes.
Best, Steven
On Feb 6, 2023, at 14:17, Steven Meisler smeisler@g.harvard.edu wrote:
Hello FreeSurfer developers,
(This issue is described in detail at the following link: https://secure-web.cisco.com/1z_dCK4QH-iYz_hvm5u5SeBYQFVo2AjBUeGX7L1n6t9FIat...)
I am preparing a surface-based morphometry analysis. I ran recon-all v7.3.2 with qcache enabled. It completes without error and I do see the output files in fsaverage space at the various smoothing kernels, as expected. However, the fsaverage space smooth files appear to be empty, based upon trying to load their data with the Python package Nibabel. Some other notes:
- When I load the QCACHE files, there are no errors in loading the file, but the data array returned is empty, so the files appear to be properly formatted morphological files.
- When I load the native space thickness file (e.g. surf/lh.thickness), I get the expected full output with all data there, so I doubt it is a Nibabel issue.
- This happens for all subjects, measurement types (e.g. area, volume, etc) and smoothing kernels.
- All qcache outputs are the same file size (668.6 kb), which doesn’t seem like it should be right.
I am planning on running models on the data outside of Freesurfer since I have to deal with complex tasks such as site harmonization and general additive modeling, which is why I cannot proceed using the typical mris_preproc method.
Please let me know if I can provide any other helpful information in addressing this.
Best, Steven