Organization of files output by fmriprep


I am preprocessing my data with Feat, but would like to re-organize the output in a similar way to fmriprep for use with another package. The document ‘Bids Derivatives RC1’ gives a lot of information on how to name the files, but not on how they are organized into different folders. I couldn’t glean this information from the fmriprep documentation either. Additionally, I couldn’t find information on how to name/where to place fieldmaps created from the phasediff and magnitude images. If anybody could point me in the right direction I would appreciate it!

Thank you!

I’ll let someone else chime in on the fieldmap placement question, but as far as how the fmriprep derivatives directory is organized, some of that information is here:

It follows the BIDS standard where in the parent directory /derivatives/fmriprep/, the top level is subject-specific folders (sub-xxx) containing session directories, each with ses-xx/anat/ and ses-xx/func/ subfolders. Additionally, there is a /logs/ folder containing the methods boilerplate in various forms (markdown, text, etc). Within each subject folder, in addition to the ses-xx/anat/ and ses-xx/func/ session subdirectories, you will have /figures/, which contains the images that are generated for the HTML report.

If you have multi-session data, as shown below, and you used the –longitudinal flag, you will also have a (non-session-specific) /sub-xxx/anat/ directory, since multiple T1w images are merged into a single template image.


│   dataset_description.json  
│   │   sub-101.html
│   └───anat
│   │   *<brain masks, preprocessed anatomical images go here>*
│   └───figures
│   │   *<figures for HTML reports>*
│   └───ses-A
│       │   └───anat
│       │   │   sub-101_ses-A_T1w_space-orig_target-T1w_affine.txt
│       │   └───func
│       │   │   *<confounds & preprocessed images go here>*
│   └───ses-B
│       │   └───anat
│       │   │   sub-101_ses-B_T1w_space-orig_target-T1w_affine.txt
│       │   └───func
│       │   │   *<confounds & preprocessed functional images go here>*

Hope that helps give you some idea of how things are organized. If you need to make sure your output matches exactly, I might suggest running fmriprep on one or two subjects in your dataset so you have that example of naming & structure as a more specific reference.


In addition to what @01010011 said above, you can check our latest preprint ( and the companion GitHub repo ( Both show how to use fMRIPrep outputs in an FSL-based analysis (via Nipype though).