Hi,

I am interested doing an ROI analysis of a particular region. I have hand drawn the ROI and extracted the median activation value for each subject and noticed tons of variability across subjects and this may be hurting my statistical tests (I have only included non-zero values). One idea I had from reading other papers is to exclude those voxels within the ROI that in the original single subject Deconvolution results did not show a significant overall F-test. This way, these voxels do not contribute to median ROI value.

I am a little unsure how to do this. I can mask out all values that don't reach, say, F=3.79 (corresponding to p < .001) and this creates an image of 1's and 0's if I use the step function in 3dcalc. However, I am concerned that when I coregister this image to the T1 scan, the registration won't work well because there is limited structural information (is this true?). Is the solution to use 3dcalc to create a mask of 1's and, e.g., 100's? If so, how would I do this? Alternatively, is there a better way to run the analysis?

Thanks so much for your help.

Jatin