Differences between release 2 and 3?

While waiting for release 3, could you please tell us if we should expect differences between releases 2 and 3 due to changes in the anatomical pipeline steps? In particular, will the surfaces be strictly identical for the same subject, as well as the cortical ribbon and the Draw-EM segmentations?
Moreover, will additional registrations (generated with MSM in inter-individual but also intra-individual as in the study of Garcia et al, PNAS 20018) be provided in release 3?
Thanks in advance!
Best regards

Hi @JDubois,

I can speak to some of the points you’ve raised:

  • The surfaces won’t be identical between release 2 and 3. There were some alterations to the surface inflation process, so there will definitely be differences in midthickness, pial, and inflated surfaces.

  • I don’t think the registrations from Kara’s PNAS paper will be released, but we will be releasing the warps generated by registering native surfaces to the cortical surface template. Using these warps, it is possible to get all cortical surface measures from native to template space

  • The template that we are releasing is an updated version of the cortical surface template described in Jelena Bozek’s Neuroimage paper. Specifically, the template has been extended to cover 28-44 weeks’ PMA. Moreover, we have made the template left-right symmetric, meaning that there is vertex-wise correspondence between the left and right hemispheres. This new symmetric template is available here: Cortical Surface Template – Brain Development

  • Instead of directly registering native space surfaces to the 40 week template, we first register to a local template and use warp/deformation concatenation to get the surfaces into template space. This is similar to the approach described here. For example, a 28 week subject will first be registered to a 28 week template, and then we use the warps generated by registering 28 week template to 29 week template, and 29 week to 30 week template (and so on) to get the subject from 28 week surface space to 40 week space.

Does that answer everything? :slight_smile:

Best wishes,



That’s a comprehensive response Logan, but to further clarify

  • all release 3 surfaces will differ from release 2 because there is always slight differences in tesselation every time the surface extraction is run. Hence white surfaces will also differ, and there will be different numbers of vertices for each dataset for each release.

  • It might be that if you contact Kara directly she still has these surfaces and warps but these were not part of the dHCP and I never had these files. At some stage when we are happy with our fetal surfaces we will try to run aMSM for our fetal to neonatal data. However, we don’t have the intermediate time points which will make this challenging.


1 Like

Thank you for your quick answers!
Regarding the surfaces, we are thus looking forward to the release 3!
Regarding the realignment, maybe my question was not clear, sorry. With some colleagues we are interested in registering the longitudinal dHCP data as best as possible as in Kara’s article. We were wondering if you had tried intra-individual (no template) registration between the 2 (or more) time-points with MSM, for each dHCP subject with such follow-up. But the link to the LISA article seems to suggest that this is not the case? Do you think either approach would be more efficient?
Thanks again! Jessica

No actually we haven’t tried longitudinal alignment for the neonates but there is no reason it would not work. If you want to analyse the deformation fields as correlates of cortical growth then you should use aMSM like Kara did but beware it takes a long time to run. If not regular MSM will work perfectly fine. I’m not sure how well you can compare cross-sectionally as the pairs are not age matched. But perhaps you can find some close enough.

Thank you Emma for your reply! Best regards. Jessica