fMRIPREP - Different resolution between BOLD files in *space-T1w* and actual T1w


I am looking at the preprocessed output generated from fMRIPREP version 20.1.1 for a single subject. During processing I specified T1w as one of my output spaces, and fMRIPREP generates files under the func folder with the space-T1w label, but the files I have looked at so far all have both voxel resolution and number of voxels matching the original bold scan not T1w scan.

I was thinking that these files would have been resampled to have the same resolution as my T1w image. Is this how it is supposed to be?

Also I am running fMRIPREP with the following command:
singularity run --cleanenv fmriprep_20.1.1.sif /bids_dir /output_dir participant --participant-label sub-10011 --fs-license-file fs_license.txt --mem_mb 60000 --use-aroma --output-spaces fsaverage T1w MNI152NLin6Asym:res-2 fsLR func --dummy-scans 4 --n_cpus 6 --medial-surface-nan --skip_bids_validation

Example files with same resolution as BOLD:

Thanks in advance,

Long story short, it is expected, but if you want to up-sample your data to match your anatomical (or to some other resolution), that will require implementation.

Generally when we say things are in the same space, we mean they are aligned such that the same (X, Y, Z) coordinates point to the same locations Xmm right, Ymm anterior, and Zmm superior to the same origin. If alignment went well, that means that you’re pointing to a BOLD or T1w sample of the same anatomical structure. But it does not necessarily mean that the same voxel indices (I, J, K) in each image match. These differences are encoded in each image’s affine.

Why do we do it this way? The main reason is to keep the amount of change to the data to a minimum. When you up-sample, you are effectively increasing the number of copies of the same read-outs, which then need to be accounted for, statistically.

That said, there are valid reasons to want up-sampled data when you resample it. We support it for MNI template-normalized images, but we don’t have it for subject-anatomicals. The closest issue would be FR: Arbitrary output resolution (for ``--output-spaces``) · Issue #1968 · nipreps/fmriprep · GitHub, and if you have the interest in contributing this feature, we’d be glad to help you get started!

1 Like