Hello!
we are planning in experiment with rs-fMRI acquisitions at 3T on a Siemens Vida scanner. We want to focus on subcortical structures and want to optimise our sequence. Following the advice of this paper, we want to use a single-band sequence with in-plane acceleration of 2. Another reason to avoid multi-band acceleration is B1+RMS limitations we have to follow.
There has been some controversy about which slice acquisition order to use. Interleaved as the benefit of less cross-talk, but our patients potentially move a bit more compared to healthy participants and thus, spin-history artefacts could be a problem. I am leaning towards ascending slice acquisition. Am I missing something important here or is this a valid reasoning?
Thanks!
The paper you link examines an important topic and the conclusions are plausible. However, I do wonder if there might be an alternate explanation for their findings. Specially, they acquired volumes of 72 slices with MB factors of 2, 3, 4, 6, 8, 9, and 12. Dividing the number of slices by the MB factor yields 36, 24, 18, 12, 9, 8, 6. With the exception of MB=8, these are all even numbers, meaning these are inherently unbalanced Multiband sequences (for details, see here, here and here). While the study used the 32-channel head-coil, MB=8 is quite aggressive, potentially leading to issues. For this research question, the choice of the partial Fourier=7/8 also may undermine the strength of their conclusion. One might also wonder if the 20-channel head-coil with restrained multi-band might also be worth testing specifically for subcortical signal (though I understand changing coils would add a lot of variance).
Again, I think that their findings are plausible. However, their conclusion have very strong implications for our field. I think there is a clear need to replicate their findings while attempting to address possible alternate explanations before embarking on large new studies.
You have correctly identified the trade off between sequential (cross-talk) and interleaved (spin-history) acquisitions. The solution to both problems is to add a bit of a slice gap, though this necessarily leads to some loss of signal (i.e. when the field of view and number of slices is kept constant, one acquires thinner slices with fewer protons if one introduces a gap between slices). The reason this debate remains enduring is one can make valid arguments for sequential or interleaved. My own preference for participants where movement is likely is to use a sequential acquisition.
3 Likes
Thank you @Chris_Rorden for this thoughtful and detailed answer and even the mini-review of the paper I linked. Our sequences currently currently has a distance factor of 25% with an isotropic resolution of 3 mm, so hopefully this will be enough to avoid meaningful cross-talk.