Hello,
Thank you for developing PRoNTo toolbox! I was analyzing Haxby's
dataset (participant 1) by using beta-images generated by SPM8 as
inputs for PRoNTo, and got some questions.
My goal was to perform a similar pattern classification analysis as
the example shown in PRoNTo manual (Chapter 11), but using beta-images
(instead of time-series data). I first performed normal 1st-level
statistics for the preprocessed data of participant 1 in SPM8. I got
96 beta images (8 conditions (faces, houses, ...) x 12 runs), and
these images were used as input images for PRoNTo.
My questions are,
In the "Data & Design",
1. I guessed that I should specify Group - G1, Subjects - S1, and
then, at "Modalities", I should specify multiple modalities (i.e.,
"Run1", "Run2",..."Run12"). For each modality (run), I put beta-images
(8 beta-images / run). Are these correct? or, I should specify single
modality (e.g., "fMRI') and put all beta-images into it?
2. I understand that I should choose "specify design" instead of
loading an SPM.mat file. I put number of conditions (8) and then,
image numbers as onsets (first beta image as 0, second beta image as
1, ...). Unit was "scan". I now wonder what I should put for
"duration" and "Interscan Interval (TR in s)". I guessed that duration
could be 1 (because there's only one beta image / run for each
condition), and any number can be put as Interscan Interval because I
don't need to correct HRF delay or overlap. Are these correct?
In "Prepare feature set",
3. I wonder which option should be chosen for "Detrend" and "Scaling".
"No" for both is the right choice as my inputs are beta-images?
In "Specify model",
4. which option should I choose for "Data operation" when single
beta-image/condition for each run was used as input? "Mean centre
features using training data" only?
Thank you very much for your help in advance!!
Michi
|