fMRIPrep Combining Echoes

Hi there,

I am trying to run fMRIprep with Tedana but fMRIprep appears to automatically combine the echoes when I run it.

I have tried using the --echo-idx flag to preprocess the echoes separately, but it is my understanding that all three echoes need to use the distortion correction parameters of the first. Is there a way to do this without automatically having the echoes combined so that I can then run them through Tedana for optimal combination? I am using multi-echo for motion correction and fMRIPrep seems to do a significantly inferior job of correcting for motion than does Tedana. Any help on this would be appreciated!


In fMRIPrep version 21.0.0 they introduced the --me-output-echos option (see here), which will output the individual, partially-preprocessed echoes. If you’re using an older version, see FAQ — tedana 0.0.11+0.g4d5c645.dirty documentation for a code snippet to pull the partially-preprocessed echoes from the fMRIPrep working directory.

Hi Taylor,

I used fmriprep version 20.2.1 and extracted data in native space by (From Tedana document). This script extracted files in native space from ‘/fmriprep-workdir/single_subject_{sub}_wf/func_preproc*task*_wf/bold_bold_trans_wf’. I wonder what steps has been applied on them.

They should have slice timing correction, motion correction, and distortion correction applied, and should be in the reference BOLD image’s space. You should just need to apply the scanner-to-T1w and T1w-to-standard transforms if you want them in standard space.

Thank you so much for your quick reply.

I noticed there is one sentence in the document of fmriprep version 20.2.1 “Please note that all routines for susceptibility-derived distortion correction have been excised off of fMRIPrep for utilization on other projects”. So I am a little confusing if it has gone through distortion correction or not.

Moreover, I am planning to extract these files by, then load these files into Tedana to get the combined-echo data. What preprocessing steps are still needed for the following analysis like calculating functional connectivity in your opinion? I guess it might be hard to send them into fmriprep again, so I may use other toolboxes like CONN. Except for coregistration and normalization, is there anything else I need to do?

Thank you so much for your help!!

That just means that fMRIPrep imports another tool (sdcflows for distortion correction. fMRIPrep’s pipeline still includes distortion correction.

It depends on what your analyses entail (e.g., further denoising with aCompCor), but in the most basic approach you can apply whatever transforms (coregistration/normalization) you want to the tedana-denoised data and feed those data into CONN. A lot of the fMRIPrep confounds, like the motion parameters, can still be used on the combined-echo data. You just don’t want to use any that use signal, like the aCompCor, global signal, or gray matter confounds.

1 Like