I have recently been looking into algorithms for optimal design of event-related fMRI experiments. I have found two noteworthy tools – optseq2 and Neurodesign – which seem good for optimal design for univariate analyses, where estimation efficiency of beta-weights or t-contrasts can be the target of optimization.
For multivariate pattern analyses, what would people recommend? e.g., for MVPA on 1) run-GLM beta-weights and 2) trial-GLM beta weights. My initial thinking is that for 1), optimization for beta estimation efficiency (e.g. with optseq2) would be preferable. For 2), I think that uniform medium-length ISI (e.g. 2-4 s) may be preferable.
Thanks for any help.
great and tough question.
Maybe you (& others having the same and/or comparable questions) find the following articles useful (just in case you haven’t read them already):
Effect of trial-to-trial variability on optimal event-related fMRI design: Implications for Beta-series correlation and multi-voxel pattern analysis
de Bruijn cycles for neural decoding
Trial timing and pattern-information analyses of fMRI data
The impact of study design on pattern estimation for single-trial multivariate pattern analysis
At least for me these were very good reads when starting with that topic.
Sorry for not providing detailed information and help, but I’m far away from being experienced enough to do so.
HTH, best, Peer
I think this paper has also some interesting things on design optimization for multivariate analysis.
Rebecca Saxe mentions this issue briefly in a quick aside during one of the fMRI bootcamp sessions at CBMM last year. She talks about the pros and cons and it could be useful to just hear a very quick and high-level note if the papers would be too tense to otherwise get a useful answer from.