How to select ROIs scientifically

Hi all,

Another dummy question. People usually select ROIs to calculate the function connectivities between different brain regions. However, in REST, we do not have enough information to pre-set the ROIs like what we do in Task states. If we use AAL atlas, the brain regions are too large, and it should not be a good way.

So, if we want to calcualte the FC, how to select the ROIs scienfitically in REST studies.

ps. How to view the FC results in REST?  or we need to use other softwares?



I would call it not  "scientifically", but rather "unbiased".  

I can think about followis alternatives:

1. To take a central sphere from your anatomical region (which is according to your hypothesis, let's say some region in the default network). Clearly, if the region is huge then a central sphere cannot reflect the activity in the whole region. But then you probably better from the begining focus on a smaller a-priori region. 

2. To have independent data-set for finding most selective voxels in the regions (let's say most correlated) and then using them as seeds for your analysis (something like here:  

3. To take a coordinate from published and cited paper (e.g. see methods in : So, the reader would be able to relate your findings to what was published previously (since many others also use this seed).

Regarding the REST I cannot answer since I do not use it.





Thank you, That's great answers and inspired me alot.


Region of interest analysis for fMRI by Russell  Poldrack

Selecting ROIs is based on your hypothesis -- you can have some ideas from the literature.

You can view FC maps, with REST Slice Viewer, MRIcroN, xjview, FSL view and many other viewing programs.