Dear Sylvain, dear François,
I’ve finished building the atlas for the 1 year old infants. You can obtain a copy here ( https://www.dropbox.com/s/9w01amxzbtmh6x7/Atlas1yo.zip?dl=0 ) if you are interested in including it in Brainstorm. I wrapped the protocol with only a default anatomy since I don’t know how exactly you usually package your atlases. You have the “go” from the authors (i.e., Li et al.) for distribution within Brainstorm (with appropriate credits to their work, of course). Description of the procedure I used to build this Brainstorm-compatible atlas is as follow (including related credits for the original files):
The atlas is based on the work of Shi et al. (2011) who proposed an MRI template (grayscale average of 90 infants, recorded longitudinally after birth, at 1 year and at 2 years; only recordings taken at 1 year have been used here), tissue probability maps, and brain parcellation according to the division from Tzourio-Mazoyer et al. (2002). To make this atlas usable for EEG source reconstruction, its corresponding brain ribbon had to be reconstructed as a surface mesh. This process has been performed in a semi-automated fashion using the BrainVisa pipeline (version 2012). To provide at faithful reconstruction, BrainVisa had to be helped manually because of the poor white matter/grey matter discriminability at this age and the normal fuzziness due to inter-subject averaging of MRI volumes. Thus, we used a Python toolbox (NiBabel) to build the grey/white matter mask from the probabilistic maps provided by Shi et al. (2011) by merging the information from the two relevant maps (grey and white matter) in the following fashion: any voxel with an intensity lower than 25% of maximal intensity in the two maps was set to 0. All other voxels which intensity was higher in the probabilistic map of grey matter than in the white matter map was classified as grey matter. The remaining voxels were labeled as white matter. The MRI-space Tzourio-Mazoyed parcellation was propagated to the cortical mesh by coregistering every cortical vertex with the corresponding voxel. [François: Note that I used Brainstorm for that but I commented the two lines where you inflate (the cortex?) in the routine to import atlases based on MRI volumes. The result is that there are “unattributed” faces between cortical regions on the atlas mesh, but every vertex is associated with one and only one region. Thus, when averaging (or using other statistics) on vertexes of the scouts defined by these regions, the source activity computed at every vertex is used for the averaging of only one region. It is less nice visually (because it makes look the parcellation as if there were “gaps” between regions, but as far as vertexes are concerns, there are no such gaps. I guess you must have weighted the pros and cons of this “inflation” but it seems appropriate to leave it out for the usage I intended.]
BrainVisa provided a poor skull reconstruction. For this reason, Brainstorm was used for the reconstruction of the head. A 2.75 mm skull thickness was entered in its boundary-element method (BEM) algorithm for the reconstruction of scalp, outer skull interface, and inner skull interface. This value was based on the work of Li et al. (2015) who reported that skull thickness in one year old babies varies between 1.5 and 4 mm.
Li, Z., Park, B. K., Liu, W., Zhang, J., Reed, M. P., Rupp, J. D., . . . Hu, J. (2015). A statistical skull geometry model for children 0-3 years old. PLoS One, 10(5), e0127322. doi:10.1371/journal.pone.0127322
Shi, F., Yap, P. T., Wu, G., Jia, H., Gilmore, J. H., Lin, W., & Shen, D. (2011). Infant brain atlases from neonates to 1- and 2-year-olds. PLoS One, 6(4), e18746. doi:10.1371/journal.pone.0018746
Tzourio-Mazoyer, N., Landeau, B., Papathanassiou, D., Crivello, F., Etard, O., Delcroix, N., . . . Joliot, M. (2002). Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. NeuroImage, 15(1), 273-289. doi:10.1006/nimg.2001.0978
Of course, this methodology has some limitations and the results will probably be much more accurate with surface averaging than surface reconstruction of volume averaging. This has been proposed recently in http://www.ncbi.nlm.nih.gov/pmc/articles/PMC4332305/ of this age range, using FreeSurfer. However, their atlas does not seem to be available yet and the authors were unresponsive regarding when it should become available. Anyways, I think the atlas I’m proposing should do a satisfactory job for the relatively coarse process of computing EEG cortical sources based on atlases of infants and should already improve upon reporting only sensor level activity!
Best,
Christian