fMRIPrep to FSL: Different Results

Summary of what happened:

Hello!

I’m running into inconsistencies when analyzing fMRIPrep-preprocessed data in FSL. When I preprocessed the same data directly in FSL, the results differ.

How I used the fMRIPrep-preprocessed data in FSL:

  • I selected the functional output from fMRIPrep that was already in MNI space
  • I then registered this functional output to the preprocessed MNI-space T1-weighted image using FSL
  • Using the FEAT directories from the registration output, I ran a first-level analysis, including additional confound EVs: translation (trans_x, trans_y, trans_z), rotation (rot_x, rot_y, rot_z), and nuisance signals (CSF, white matter, global signal)
  • Using the cope3 outputs from the first-level analysis I ran the third-level analysis (the GLMs were the same for both analyses)

Version:

FSL 6.0.7.14

Data formatted according to a validatable standard? Please provide the output of the validator:

BIDS

Screenshots / relevant information:

Results from FSL preprocessing:

Results from fMRIprep preprocessing:

Unless I am misunderstanding, if the functional data is already in MNI-space, it has already undergone registration, normalization, etc. …so why perform registration again? It certainly incurred another interpolation step, which reduces the quality of the data. And while I’m not overly familiar with FSL, did you ensure the process of creating the FEAT directories doesn’t perform redundant steps that fMRIPrep already accomplished (i.e. McFLIRT, etc.)?

In other words, it sounds like the data went through both pipelines sequentially which degraded the quality.