create ROI from TT_desai_dkpmaps atlas and convert to MNI space


I want to create a binary mask of the left rostralmiddlefrontal region from the Desikan-Killiany probabilistic atlas, TT_desai_dkpmaps+tlrc, so I ran the following command (to get a mask of all voxels > 25% in the probabilistic atlas):

3dcalc -a ‘TT_desai_dkpmaps+tlrc[19]’ -expr ‘ispositive(a-0.25)’ -prefix leftMFG

My first question is: Does this step yield a binarized mask (which is what I want)? Or do I have to run an additional command?

Next, I attempted to convert the above mask (left MFG) to MNI space. My resting functional data is normalized to MNI space and I could only find this specific atlas in Talaraich space. I found this old AFNI message board post (,70173,80922#msg-80922) and tried the first couple of commands posted there.

3dWarp -tta2mni -prefix leftMFG_mni leftMFG+tlrc

3dWarp -matvec_out2in mni_fsl2tal.1D -prefix lMFG_mni2 lMFG25+tlrc

These steps both ran successfully. Can I use either option for converting the Desikan-Killiany ROI to MNI space? I am new to AFNI, and looked at a couple Message Board posts but I want to make sure I’m doing this correctly before running my seed-to-voxel resting fMRI analysis. Thanks!


The most accurate approach is to use the 3dfractionize method suggested in that previous post. That does an exact transformation of the N27 anatomical dataset used there to go from MNI to Talairach space.

Hi Daniel,

Thanks for the suggestion. The two 3dWarp commands simply took a mask of leftMFG+tlrc as input and applied a transformation to produce an output file in MNI space with the specified prefix. I am looking over the 3dfractionize example and want to make sure I properly specify the commands (Example: 3dfractionize -input dset+tlrc -warp ~/abin/TT_N27+tlrc -prefix dset_mni -template template_mni+tlrc).

For -input, I assume I would specify leftMFG+tlrc as I did previously. But the post references that the user must create a template dataset (presumably in MNI space). Does this consist of my resting functional scan(s) already in MNI space? I’m confused. Also, can I use this method if my ROI is from the TT_desai_dkpmaps+tlrc dataset (and not TT_N27 dataset as the post mentions)? Sorry for these two follow-up questions, but I want to make sure I properly specify the appropriate input, images for the -warp step and the appropriate template images. I appreciate your help!


For the template parameter, you should be able to use the MNI template. Here, the template’s only purpose is to define the grid and the name of the space for the output (MNI), so you can use the MNI_N27 dataset or one of the MNI152_T1_2009c dataset or almost anything. If you would like it to be on another grid, say a transformed EPI dataset, with lower resolution, you can create a new dataset with lower resolution and use that as the template. The TT_N27 dataset is definitely the warp dataset you want here. That contains the manual transformation in its header to go from MNI to Talairach (TT_N27) space. The Desai atlases are all based on alignment to that template, so it’s the most accurate transformation.

Thanks so much, Daniel. That’s clear. I should be good to go!

Hope you had a nice weekend,