Framewise displacement implementation with nilearn

rikaatgithub · August 5, 2021, 7:12am

Hi all,

I’m trying to make functional matrix with nilearn modules of Python.
I want to do frame censoring (high motion data removal) automatically.
I have tsv file of confound data since I performed preprocessing via fmriprep.

Do you have any idea to implement framewise displacement with nilearn module?

Thanks in advance.

NicolasGensollen · August 5, 2021, 7:59am

Hi @rikaatgithub

You might want to check load_confounds: GitHub - SIMEXP/load_confounds: Load fMRIprep confounds in python
We are planning to have some integration within Nilearn at some point, but this is still ongoing work (see this issue: Integration of `load_confounds` to `nilearn` · Issue #2777 · nilearn/nilearn · GitHub).
Note that, in the mean time, it shouldn’t be too hard to integrate it in your workflow.

Let me know if this is helping or not!
Nicolas

rikaatgithub · August 5, 2021, 9:39am

@NicolasGensollen
Thank you so much for your quick reply.
So we can not implement framewise displacement without nilearn’s module for now…
I’ll try it!

JohannesWiesner · January 30, 2026, 1:35pm

[ENH] Add function to compute framewise displacement

opened 05:07PM - 04 Dec 24 UTC

closed 04:52PM - 10 Dec 24 UTC

JohannesWiesner

### Is there an existing issue for this? - [x] I have searched the existing iss…ues ### Describe your proposed enhancement in detail. Are there plans to offer a function that calculates framewise-displacement (and possibly other measures?) I am aware that `nilearn.interfaces.fmriprep.load_confounds` exists but this relies on a previous fmriprep run. There is an R-package "fMRIscrub" that can do this, but I would love to see a python-implementation preferably in nilearn. Example: ```python import requests import numpy as np from io import StringIO from rpy2.robjects.packages import importr from rpy2.robjects import numpy2ri numpy2ri.activate() fmriscrub = importr('fMRIscrub') # download this file: # https://openneuro.org/datasets/ds000113/versions/1.3.0/file-display/derivatives:motion:sub-01:ses-forrestgump:func:sub-01_ses-forrestgump_acq-dico_run-01_bold_moco.txt url = ("https://s3.amazonaws.com/openneuro.org/ds000113/derivatives/motion/sub-01/" "ses-forrestgump/func/sub-01_ses-forrestgump_acq-dico_run-01_bold_moco.txt" "?versionId=oCUziRWPCbZh5XCF3R283ojCdw8BQRRJ" ) # Fetch the file content response = requests.get(url) content = response.text motion = np.loadtxt(StringIO(content)) # compute mean frame wise displacement # make sure R and rpy2 is installed and install.packages('fMRIscrub) fd = fmriscrub.FD(X=motion)[0] mean_fd = np.mean(fd) ``` ### Benefits to the change I haven't found any package besides fMRIScrub and this one that can calculate Mean FD: https://gist.github.com/JulianKlug/68ca5379935e0eedb9bdeed5ab03cf3a Especially for old in-house datasets it's complicated and/or not possible to bring the dataset into BIDS-format, run fmriprep and then load them using `nilearn.interfaces.fmriprep.load_confounds` ### Pseudocode for the new behavior, if applicable ```python # insert your code below ```