Good morning everyone,
I’m analyzing a child population dataset and create a custom template was one of the suggestions that were given to me. During this week I’ve been studying some methods, and I learned that @toMNI_Qwarpar (and @toMNI_Awarp) was the pipeline used to create the Haskins pediatric brain atlas.
Reading @toMNI_Qwarpar script code, the ‘minpatch’ flag from 3dQwarp caught my attention since it seems to be a pre-defined value that decreases at each warping sequence (101, 49, 23, 13).
From what I understand, it’s likely related to an optimization problem of the non-linear warp cost function used by AFNI, and from the documentation; the -minpatch should be an odd integer, and the minimum is 5. I’d like to understand more about this specific parameter, what’s behind it and how it should be chosen with the purpose of creating a study-specific template. Actually, it would be great if someone posts some references here since I couldn’t find any paper for Haskins pediatric template.
I’m very thankful for whom can help me!
Hi José,
We’re still in the midst of writing the Haskins Pediatric Paper. You can checkout the poster here. But I’m happy to answer your questions about it.
-minpatch does go down to 5. But, as noted in the documentation, it’s hard to optimize a search problem on that scale for multiple processors. I did play around with other options when making the Haskins Template, and don’t see much benefit in going to smaller patch sizes. The nonlinear templates are already incredibly crisp.
What age group are you targeting with your template?
-Peter
Hi Peter,
Thank you for explaining it. So it’s really related to optimization of the warping process, my biggest question now is which criteria or what should I take into account to determine an ideal minimum patch.
We have a dataset of children between the ages of 5 and 8, they’re younger than ones used for the creation of Haskins’s atlas.
Many thanks!
José