Design with long-duration events for MVPA

,
1

Hi all,

I’m designing an fMRI study with the main goal of performing MVPA on the data. The phenomenon I want to measure is a perceptual event that must unfold over a long interval (approximately 40–45 seconds). Because of this constraint, the most appropriate approach appears to be a block design with one long continuous event per block, repeated through my 4 conditions in several runs.

Since I can’t include multiple trials within each block, I will obtain only a single beta estimate per block. With four conditions, the design may require a large number of runs to obtain enough estimates for MVPA.

I have experience with MVPA for event-related designs but much less with long-duration block designs of this type. Would have any recommendations on how to design and optimize such a paradigm for effective MVPA? I thought about artificially splitting the blocks into several events, but this seems like a poor solution because it would reduce the independence of the events.

Thank you in advance for any advice you can share !

2

If you want to models some effects occurring during the blocks you may want to take polynomials (constant, linear, quadratic…) to model “what happens” during the blocks, or indeed chunk the blocks into time segments.