We would like to use the Oreilly 1-year-old template for MEG analyses. The dropbox link for the copy of the template (see the post below) does not work any more. Do you mind sharing the template again?
We plan to run freesurfer on the template so it would be great if you can share the nifti format of the template to us. Thank you!
I recently found some issues with this atlas. I am currently doing different validation on results obtained using common average atlases for source reconstruction. I am not convinced for now that it is likely to produce very interesting results. I still need to run some tests.
The proposed atlas could be corrected manually, but although it is possible, I am not currently pursuing this avenue because I think the data used to make this atlas are not optimal. If you have MRI from infants and you just want an average to do group analyses infants like you would do with fsaverage in adults, I would recommend having a look at: https://surfer.nmr.mgh.harvard.edu/fswiki/infantFS
There is also a very nice dataset from the UNC: https://www.nitrc.org/projects/infantsurfatlas Unfortunately, they do not have BEMs right now for source reconstruction. I asked them a while ago if they could add these files. They are supposed to be looking into this.
@Francois I am not sure about the best course of action. If we had a better alternative, I would say yes but right now, as far as I know, there is no alternative atlas for this age. I think that using an adult template (or something line 3-shells) for analyses in infants would be way worse than using this atlas. Actually, the timing for this to come up is not bad because I am currently working on testing the impact of using such an atlas and I am looking at different possibilities to come up with an alternative atlas. If you are not in a rush to take a decision on this, I could come back to you by the end of the year with an update and we might then have better alternatives at that time.
@christianoreilly: Christian, thanks for the information! We would love to use UNC atlas because we are running large longitudinal and cross-sectional infant and toddler MEG studies so we need templates for different age (3, 6, 9, 12, 18, 24, 36, and 48 months). They only have surface folders available to be downloaded. Also the reason they don't provide inner_skull, outer_skull and outer_skin surface perhaps is because most templates have been skull stripped and there is no face information?
At the moment my brainstorm pipeline relies on the Oreilly_1y template. However for the longitudinal subjects, those infants are now toddlers, so I am running into issues whether or not to use the Oreilly_1y template for 2 year old kids who don't have MRIs, or to use a high-quality 24-months old MRI as a template for those older kids.
@Francois: We downloaded the template but it only provides analyzed format and not the nifti format, so we can't run freesurfer on it. Do you have the nifti format of Oreilly_1y template?
@yuhanmc Correct. All skull stripped. That is why in the current state it is not usable for source reconstruction and why I asked them to provided the BEM surfaces because without the original MRI we cannot reconstruct these surfaces by ourselves. But if they were to provide the BEM surfaces, I think we would have everything needed.
This is the problem:
I would need to go back to the reconstruction and check thoroughly for these kinds of reconstruction errors and correct them manually.
I have been working with the dataset of Richards et al and I am quite happy with the preliminary results. This is the result I got for 12 months:
They have MRIs for 3, 4.5, 6, 7.5, 9, and 12 month-old. I'll do the same process for the other time points, check the result in source reconstruction and, if the results are good, I'll ask Richard for sharing them within toolboxes. These days, I am more working in Python but I guess it should be not much additionnal work to provide a package for Brainstorm since I did all these surfaces in Brainstorm. With some chance, I'll manage to package all this soon enough to make it a Christmas present
Full disclosure: The brain surfaces have been reconstructed from averaged volume MRI using the standard recon-all. I would have prefered to use Zöllei infant's Freesurfer package, but so far I did not manage to get it running. The surfaces seems good and in agreement with the MRI when overlaid and eyeballed using Freeview.
@christianoreilly That was amazing! Such a great Xmas gift for all of us who study babies! I was about to suggest using John Richard’s templates (we are currently using his 3mos and 7.5 mos templates for other projects using a non brainstorm analyses pipelines). He has templates all the way up to school age (and also adults). We also have asked him to share his templates for school age but they are easy for us to run freesurfer in so we haven’t had issues.
We are also planning to use Zöllei’s infant freesurfer package on babies under 6 months old (some of our infant subjects do get sMRI). I also haven’t been able to get it running because among all the libraries that needs to be installed, one of them is Christos’ registration tool (I think) he or his student wrote a while ago and I can’t get to it. I also don’t know how her package work with T2 in babies < 6 months old. We would love to chat with you to see how to make it work if you are also interested in making it work. Please keep us updated and I’ll wait for your brainstorm Infant templates available (yay!!).
Yes! I am now a few doors away from your lab! Let's talk at some point if there are some occasions to collaborate on common goals! If I am not mistaken, you have already ongoing exchanges with Mayada for future projects.
Hi @yuhanmc. Sorry for not having provided any feedback to your last message. I was juggling with many balls at that time. I have a list of instructions I used to make Zollei's pipeline work for me. I can sent them to you if you'd like. Can you give me your direct email? If you still feel it would be useful to you to chat about this pipeline, I'm happy to do so at any time but I am not that knowledgeable about that pipeline so I am not sure to what extend I'll be able to provide a relevant feedback. I just tweak and fix so that it work for my use case. As for the templates, the 12mo works seems to work well. I got issues for the other time points. I resume working on it. I should be done with it soon.
@christianoreilly Thanks for your reply! Yes it would be great if you would like to share the instruction with us to install infantFS. My email is email@example.com or firstname.lastname@example.org. Let me see if I can follow your instruction to install the Zollei's pipeline and then I will be able to chat with you about specific questions when I use it.
Just an update. I got I thing good quality templates build for 8 time points from 3 month old to two years.
They passed all checks in MNE and looks visually good. I have not tried them yet in Brainstorm. I would use them to do some source reconstruction to see how they behave before putting them out-there, but if someone has an immediate need, they can be shared now (maybe just after final check with John Richards to validate that he his ok for these derived data to be shared).
That is great! I am still struggling with running infant_recon_all. Got stucked at fspython and our IS guy couldn’t fix it (I don’t have sudo right on building any libraries). I might have to ask you for more detail information on how you make infant_recon_all works on the template MRIs if you don’t mind.
I have a large database of infant MEG data from 0 – 48 months old (visual, auditory, resting-state, and face), source reconstruction using dSPM. Most of them are from our dedicated infant MEG machine. I can select one subject each age range and run the same analyses pipeline and see how the results are different from previously analyzed data (registered to your previous 1-yr O’reilly template in brainstorm). Let me know what you think and how you can share those templates?