As far as I know (4) is double-dipping, because you know from your
whole- brain analysis that effects in those coordinates are present or
not present, and you test ROIs on exact same data. Independence comes
from choosing ROIs based on different data set, so other publications /
meta analysis or testing hypothesis based on anatomical structures from
On 2017-03-15 17:11, Mike wrote:
> I know the definition for ROIs should be independent but there is a commonly encountered question and I would like to confirm my thinking.
> Suppose in an experiment, subjects had to see a picture, maintain a working memory for it (EV1: WM), and decide if the picture was included in a probe (EV2: Decision), which consisted of 3 pictures. Researchers performed a whole-brain analysis for WM period and found that amygdala and hippocampus were activated. They then wanted to examine if both regions were also activated during / contributed to the Decision period (whole-brain analysis during Decision did not yield activations in both regions). There are 4 possible ways to define amygdala and hippocampus ROIs:
> (1). Using coordinates from previous related fMRI studies, defining r=6 spherical ROIs centering at these coordinates.
> (2). Using coordinates from a meta-analysis focusing on similar studies, defining r=6 spherical ROIs centering at these coordinates.
> (3). Using anatomical atlas (i.e., MarsBar AAL).
> (4). Using exactly the coordinates identified during the WM period, defining r=6 spherical ROIs centering at these coordinates.
> I think (1)-(3) is ok but not sure if (4) is allowed or not. Is the definition of ROIs independent in (4)?
> Thanks in advance.
Laboratory of Brain Imaging (LOBI)
The Nencki Institute of Experimental Biology
Pasteur 3, 02-093 Warsaw, Poland
Tel.: +48 22 5892 551
mail: [log in to unmask]