One option is to use the HO maxprob th0 file that comes with fsl. 

Alternatively, you load the probabilistic HO atlas e.g. in matlab, then each volume is a Roi. If you take the max over the 4th dimension you see for each voxel which roi is more probable. You can see how I did it here https://code.google.com/p/funpsy/source/browse/#svn%2Ftrunk%2Fatlases%253Fstate%253Dclosed and if you download the file HO_2mm_rois.mat you get a struct with Rois and for each Rois the voxel ids. 

Best 

Enrico

--
Enrico Glerean (MSc.)
PhD Candidate, Brain & Mind Laboratory http://becs.aalto.fi/bml/
BECS, Aalto University School of Science
00076 AALTO, FINLAND
mobile: +358 40 1877599
email: [log in to unmask]
www: www.glerean.com

On 28 May 2014, at 20:12, "Yang, Daniel" <[log in to unmask]> wrote:

Dear all,

Or, alternatively, how to completely divide the brain (cortical area) into non-overlapping ROIs per the Harvard-Oxford atlas?

Best,

Daniel

On 5/28/14 12:56 PM, "Yang, Daniel" <[log in to unmask]> wrote:

Dear all,

I found that the 48 cortical ROIs in the Harvard-Oxford atlas tend to partially overlap each other, due to the nature of a probabilistic atlas.

However, it is possible to generate non-overlapping ROIs out of this atlas. Is it okay to do threshold/probability >= 50?

Best,

Daniel