fMRIPrep - Head movement estimation is performed in a different orientation than the output orientation

[valwmr]

Hello there,

the last couple of days I encountered some issues with my preprocessing pipeline.
After successful preprocessing, I wanted to use the estimated head motion regressors as covariates for DCM and RS analyses. While working with the output data, I noticed that my input data had a different orientation than the output images (LPS → RAS). Then I wondered at what point the head motion estimation was performed, and it seems that it is happening with the original orientation and not after a reorientation step (is reorientation of EPIs implemented or is it only performed during registration from EPI to T1w?).
Therefore, I preprocessed my data once in the original orientation (LPS - ‘fmriprep_unreorient’) and after reorientation to the initial space (RAS - ‘fmriprep_reorient’) and plotted the calculated motion regressors:

They are actually inverted for the x and y coordinates due to the different original orientation.

I think it might be a problem to use the motion regressors if the input image does not match the output orientation, or do I have a misunderstanding? It might be useful to align the EPIs to the template/output orientation before mcflirt calculates the head motion regressors.

Thank you in advance for your valuable comments on this topic!

[paulmccarthy]

Hi @valwmr, this shouldn’t matter - the GLM will just assign a negative beta value if there is any variance in the data which is negatively correlated with a motion parameter.

Having said that, it’s always a good idea to re-orient your data before preprocessing, using fslreorient2std or similar, as doing so can improve the robustness of subsequent preprocessing steps (e.g. registration), and having all of your data in the same orientation helps to avoid any orientation/L-R mismatch issues.

[valwmr]

Thank you very much for the answer.