Where is an affine transformaiton matrix of EPI to T1w in fmriprep output folder

[ChrisGorgolewski]

@1112

I can find three txt file below which look like affine matrix /fmriprep_wf/single_subject_001_wf/func_preproc_task_rest_run_01_wf/bold_reg_wf/bbreg_wf/fsl2itk_inv/affine.txt

/fmriprep_wf/single_subject_001_wf/func_preproc_task_rest_run_01_wf/bold_reg_wf/bbreg_wf/fsl2itk_fwd/affine.txt

It’s both of the above. One is forward (EPI → T1w) and one is inverse (T1w → EPI).

@1112

Is there EPI image which is resampled with the same matrix of T1w image?

I think this EPI image is the result of bbregister.

There is no such option currently (since this would upsample the data and waste a lot of disk space). You are right that one of the FMRIPREP outputs is already aligned (while maintaining original resolution) with the corresponding T1w volume (which is why we did not see the need to output the affine matrix). In 90% cases there is no need to explicitly upsample the BOLD data, since you can use viewers that respect affine matrices (MAngo, FSLeyes etc.) and toolkits such as nilearn (for example for extracting time series from parcels with two aligned files that have different resolutions).

Relevant:

• BIDS -> fmriprep transforms to (3.5 x 3.75 x 3.75) MNI space?
• save transforms in subject derivatives output directory · Issue #1042 · nipreps/fmriprep · GitHub