20 February 2021 0 3K Report

Hello all

I am planning to run an fMRI study that is a single-factor three-level design. However, I do not have any prior experience in fMRI and I found people are doing it in some very different way.

We will have three sets of images (A, B, C, 18 images per set) + one set of blank (mix of pure black and white) and we will be comparing the BOLD signal between these sets (A-Blank, B-Blank, C-Blank, A-B, A-C, B-C)

The study will be more of a neuroaesthetic study, in which images were grouped by their unique characteristics.

One requirement for the study is I will like the participant to view finish viewing all images in a block before they see images in other blocks, which means I might need to use a blocked design. (Also because the effect I am looking at is rather weak) However, I am also interested in how the participant responds to each image (the participant will rate each image on their appearance level when the image is showing). After reviewing some literature I currently have three designs

  • Each block = 6* (1s blank + 1s cross + 6s image + 0/2/4s jitter ITI)
  • 6A -> 20s (rest) -> 6A -> 20s -> 6A -> 20s -> 6 Blank -> 20s -> 6B -> 20s -> 6B -> 20s -> 6B -> 20s -> 6Blank -> 20s -> 6C -> 20s -> 6C -> 20s -> 6C -> 20s -> 6Blank

    Total session time = 940s

    2. Each block = 1s blank + 1s cross + 6*(6s image + 1s ITI)

    6A -> 20s (rest) -> 6A -> 20s -> 6A -> 20s -> 6 Blank -> 20s -> 6B -> 20s -> 6B -> 20s -> 6B -> 20s -> 6Blank -> 20s -> 6C -> 20s -> 6C -> 20s -> 6C -> 20s -> 6Blank

    Total session time = 748s

    3. event-related design = (15A+15B+15C+15Blank)*(1s blank + 1s cross + 6s image + 20s ITI)

    Show all A then B then C, inserting Blank randomly

    Total session time = 1680s

    -> I feel this one might be too long so reducing each set to 15 images

    My questions are

  • Does this sound like a reasonable design?
  • Does this design allow me to look at the participant's individual responses toward a single image? I know that in general, the BOLD signal takes 20s to go back to normal, which a lot of recent studies suggest mixed design, but they are mostly mixing different trial-type which in my case I want to keep them separated.
  • Additional to looking at contrast, I am also interested in dynamic causal modeling and psychophysiological Interaction analysis, but I am unfamiliar with these analyses so I am wondering which design will work the best
  • Thanks for your attention

    More Jen-Tse Dong's questions See All
    Similar questions and discussions