fMRIPREP not applying motion correction for multiecho data


FMRIPREP does not seem to be applying motion correction to all of the echos in our multiecho data set. A movie of the preproc_bold clearly shows motion in the image, and the logs show mcflirt run only on the first echo (see below). We have observed this in versions 20.1.1. and 1.4.0.

[Node] Setting-up fmriprep_wf.single_subject_999999_wf.func_preproc_task_smokecues_echo_1_wf.bold_hmc_wf.mcflirt" in "/u/project/elondon/data/mri/sexdiffs_smoking/tmp999999/fmriprep_wf/single_subject_999999_wf/func_preproc_task_smokecues_echo_1_wf/bold_hmc_wf/mcflirt".

[Node] Running "mcflirt" ("nipype.interfaces.fsl.preprocess.MCFLIRT"), a CommandLine Interface with command:

mcflirt -in /u/project/elondon/data/mri/sexdiffs_smoking/BIDS/sub-999999/func/sub-999999_task-smokecues_echo-1_bold.nii.gz -out /u/project/elondon/data/mri/sexdiffs_smoking/tmp999999/fmriprep_wf/single_subject_999999_wf/func_preproc_task_smokecues_echo_1_wf/bold_hmc_wf/mcflirt/sub-999999_task-smokecues_echo-1_bold_mcf.nii.gz -reffile /u/project/elondon/data/mri/sexdiffs_smoking/tmp999999/fmriprep_wf/single_subject_999999_wf/func_preproc_task_smokecues_echo_1_wf/bold_reference_wf/gen_ref/ref_bold.nii.gz -mats -plots -rmsabs -rmsrel

Is there extra specification required to enable motion correction for the multiecho preprocessing pipeline?

Thanks, Dara

I think we’ll need more information to figure out why there is motion in the movies of your data, but MCFLIRT is only run on the first echo for a reason. The transforms from the motion correction are retained and applied across all echoes later on in the workflow. You wouldn’t want to independently correct motion for each echo, since motion between echoes should essentially be nonexistent, while mistakes in motion correction could introduce differences between them that would cause problems later on. The contrast and signal coverage changes across echoes, so artificial differences in motion estimates across echoes are a real concern.

There were a couple of multi-echo-specific bugs fixed in 20.2.0rc0, although those were related to reference-image generation and susceptibility distortion correction. I don’t think either change would impact how motion correction was applied, but I could be wrong. @emdupre, does that sound right to you?

EDIT: Scratch that. Based on this comment of mine that I completely forgot about, HMC was being applied twice. Most likely, that would be the source of motion in your movies- not a lack of motion correction, but rather motion correction being applied multiple times. Anyway, using 20.2.0rc0 should fix the problem.

1 Like

Thanks very much, Taylor! We will use 20.2.0rc0.