Slice timing correction (STC) in multiband fmri (resting) - CONN toolbox

Hey! I am studying functional connectivity in resting condition in a sample of older adults. In the sequence, we implemented a multi-band acquisition (Siemens) with a TR=1.

I am wondering if we this short TR is necessary to run slice timing correction or if there is not gonna be differences at all between running or not running it in CONN with multiband acquisition? I am trying to find something about this specific topic in the literature but so far I just found this paper ( in which they discuss the use or not use of STC depending in the TR, and with FSL and SPM and TR equal or lower than 1 they do not see differences.

Any thoughts / knowledge in this is more than welcome.

Thank u so much!


Hi @andreacopu, interesting question!

This question was also asked here: Slice time correction with multiband/short TR and here: Slice timing correction - TR threshold

In our fMRI Center, where we use TRs around 1s, we tend to not use STC. We were influenced in this regard with the HCP consortium, where they chose to not use STC in their analyses.

Details about the HCP consortium thoughts on STC are here: HCP fMRI slice-timing acquisition parameters - Connectome Data Public - HCP wiki

Here is what SPM is saying:

When to use slice timing correction?

  • Slice timing can be done on any fMRI data but it is particularly beneficial for studies with longer TRs (i.e. >2s). For studies with shorter TRs (≤2s), temporal derivative may be a better option.
  • Slice timing can be performed on multiband acquisition data, however, with multiband data and a short TR, slice timing correction can usually be skipped without much impact. If you decide to run slice timing correction on multiband data, it is necessary to use slice timings instead of slice order, since a slice order cannot represent multiple slices acquired at the same time in a vector. If you don’t know your slice timings, you can artificially create a slice timing values manually. Generate artificial values from the slice order with equal temporal spacing and then scale the numbers to the TR, so that the last temporal slice’s timing = TR - TR/(nslices/multiband_channels).
  • If you plan to use DCM, slice timing correction is necessary.

Additional interesting literature on this topic: