Slice time correction and onset time: need adjustment?

,
1

If slice timing correction is applied to an fMRI run, is it recommended to adjust event file onset times?

2

Some softwares can account for that, such as nilearn (nilearn.glm.first_level.FirstLevelModel - Nilearn, slice_time_ref). Otherwise, yes you should adjust the onsets. If you slice time correct to the beginning of the TR then no shift should be necessary.

1 Like
3

SPM also adjust for this during model specification.

See an example from the SPM tutorial.

  • Highlight Microtime resolution and enter 24,
  • Highlight Microtime onset and enter 12. These last two options make the creating of regressors commensurate with the slice-time correction we have applied to the data, given that there are 24 slices and that the reference slice to which the data were slice-time corrected was the 12th (middle slice in time).
1 Like
4

Afni user. Looks like I should’ve done some more digging. If you use fmriprep, recommendation is to subtract onset times by ~TR/2 (actually: (Nslices-1)/Nslices * TR/2). Value can be fed directly to 3dDeconvolve via -stim_times_subtract.

Still wondering if this is valid for most processing done with fmriprep :thinking:.

Relevant threads:

https://reproducibility.stanford.edu/slice-timing-correction-in-fmriprep-and-linear-modeling/