Does fMRIPost-AROMA replace xcp_d for ICA-aroma denoising?

[Tiny_Beagle]

Hello!

I am processing some resting-state functional images. Previously, I would run the images through fMRIprep:23.0.2, invoking the -use-aroma flag, and subsequently run the fMRIprep output with XCP-D to obtain denoised images. I am adjusting my pipeline to include a more updated version of fMRIprep (24.0.0), which brought me to fMRIPost-Aroma.

In the documentation of fMRIPost-AROMA, it seems to say that fMRIPost-AROMA would output denoised data. However, it also says that fMRIPost-AROMA outputs a set of confounds that can be used to denoise the data.

Is fMRIPost-AROMA outputting confound files for ICA-AROMA which require the use of XCP-D to actually denoise? Or does fMRIPost-AROMA replace XCP-D for the denoising step and the output from fMRIPost-AROMA could be directly used to extract connectivity matrices using tools like nilearn?

[tsalo]

Is fMRIPost-AROMA outputting confound files for ICA-AROMA which require the use of XCP-D to actually denoise?

My long-term goal is to make it so XCP-D accepts derivatives from multiple pipelines, including fMRIPost-AROMA and the other fMRIPost workflows I’ve been working on. At the moment, however, XCP-D can’t use regressors from fMRIPost-AROMA.

@Tiny_Beagle

Or does fMRIPost-AROMA replace XCP-D for the denoising step and the output from fMRIPost-AROMA could be directly used to extract connectivity matrices using tools like nilearn?

XCP-D does extra things that fMRIPost-AROMA and nilearn don’t, such as despiking, producing interpolated denoised data in scenarios where people want that, recalculating FD based on a custom head radius, etc., so you would only want to use fMRIPost-AROMA as the denoising tool if you are comfortable not including those steps in your denoising pipeline.