creating atlas for older versus young adult comparisons

Hi all,

I am working on an fMRI study comparing older adults to young adults. We are running our analyses in AFNI and would like to make one atlas from our subject data that could be utilized for preprocessing all participants. This technique was suggested during my lab’s attendance of the AFNI Bootcamp at WashU in January.

I have gathered some resources from the workshop (specifically the Alignment and Atlases handout), the website for creating an atlas (, and have spent some time looking through the AFNI message boards, but I am still unclear on how exactly the atlas should be made and what the input should be.

I currently have two specific questions, and would also appreciate any additional insight into this process you all think we might need!

  1. To fairly represent the young and older adults in our sample, how many single subject scans are recommended to create this atlas? Should we use our entire sample, or would you suggest using a subset of our sample?

  2. The code for creating an atlas from your website is as follows:

@AfniEnv -set AFNI_SUPP_ATLAS_DIR ~/CustomAtlases/
@Atlasize -space MNI -dset hmat_spm_final.nii
-lab_file keys.txt 1 0 -lab_file_delim ‘;’
-atlas_type G -atlas_name HMAT -atlas_description ‘Motor Meta’

The ‘keys.txt’ file this code references to - what information should be included in this file, and how do we go about getting that information from our single subject scans?

Thank you in advance for any help or insight you all can offer - we really appreciate!


I wasn’t there, but I would guess the suggestion was to make both an older and a younger template, and maybe atlas segmentations for those groups too. You could get the segmentation from either manually drawn ROIs on each subject or from FreeSurfer segmentation for each. The template creation step can be done with the @toMNI_Awarp and @toMNI_Qwarpar scripts that do affine and nonlinear alignment to create an iterative template. There are many steps to this, but it’s a little involved for a messageboard post. Email me with more questions as you run into any issues.

Hi Dan,

Thanks for the response. Could you clarify further why we would need to create two separate templates? If we normalize each groups to a different template we would not be able to compare them, correct? And comparing young vs old is the primary goal of the study. Our goal end product is to have one atlas that merges young and old subjects. Do we create a young and old template, and then the programs will merge them together? Or do you recommend using two separate templates? If so, could you elaborate a bit as to why? Lastly, how many subjects from each group do you suggest we utilize?

Thank you again for your help. I hope the questions are clear.