maartenmennes/ICA-AROMA

Using ICA-AROMA after preprocessing

Opened this issue · 3 comments

Hi,

We have recently completed preprocessing our functional data using the default fmriprep pipeline. However, we've noted that the current version of fmriprep does not include ICA-AROMA. We are interested in implementing ICA-AROMA on this already preprocessed functional data. Would it be possible to apply ICA-AROMA at this stage of our processing workflow?

Thank you in advance for your assistance!

I'm not too familiar with the default fmriprep pipeline, but I would say yes, this should be possible. Our experience was that AROMA was best implemented before highpass filtering and any further nuisance regression. If these steps have already been implemented in the standard pipeline I'd carefully check the output of the AROMA classifier to make sure the component selection still makes sense.

The default pipeline does include high pass filtering and compcor regressors

I would be hesitant to use AROMA after compcor. I'm wondering how separable signal related to head motion will still be after the compcor regression. This will no doubt have an impact on the AROMA classifier. We've never validated it for this kind of usage. That said, there is no penalty in trying, other than very carefully checking the output as indicated above.