Sampling functional data for the dentate gyrus

[viktor-pfaffenrot]

Hi hippunfold-team,

I am trying to sample my data in the dentate gyrus. However, the DG has only one layer output in native space, the mid-thickness. I was thinking of sampling between the inner surface of '_label-hipp' and the midthickness of 'label-dentate' as a crude firs step.

For the hippocampus layers, I use the label files 'label-hipp_atlas-bigbrain_subfields.label.gii' to mask the boundaries I want to sample in between. I.e. I index the inner/mid/outer surfaces by the numbers given in the label file to sample only within the e.g. subiculum. However, the label files exclude the DG and SRLM (number 6 and 7 in '_desc-subfields_atlas-bigbrain_dseg.nii.gz').

Are there some similar surface labels for DG and SRLM? I.e. is there an 'atlas-bigbrain_subfields.label.gii' for those regions such that I can mask the inner surface to have the same number of vertices as the dentate mid-thickness?

Or is there maybe a better way of sampling the DG and SRLM?

Many thanks!!

Cheers,

Viktor

[jordandekraker]

Hi Viktor, thanks for the good questions

DG and SRLM are both a bit different from the other subfields.

SRLM is technically part of the CA fields and Sub. It is the inner-most laminae of those subfields, and typically has a different intensity because it is high in myelin (however, it can also contain blood vessels and the hippocampal sulcus). One way to sample it could be to project each volumetric SRLM voxel onto its nearest inner surface vertex, and then average all the projected values per vertex. I haven't experimented much with this, but was interested in doing so since SRLM intensity was shown to be a good biomarker for AD in this paper. Please keep me posted if you find something interesting and perhaps we can add a method for this to the hippunfold_toolbox!

DG is also a bit different from the Sub and CA fields since its actually topologically separate, which is why we gave it its own unique surface file. However, its very thin (~300um, less than a typical voxel) and so separating it into distinct layers hasn't really made sense for our applications so far.

I hope that helps,

[viktor-pfaffenrot]

Hi Jordan,

thanks for the explenation, it really helped and sorry for a late reply from my end.

I played a bit around and I think that solution worked for me (disclaimer: I am sampling in matlab using spm_sample_vol so my approach likely needs adaption if other tools are used):

• I create a linear space from the inner to the outer surface coordinates to sample GM
• for CA1 and CA2, I extend that linear space by half the GM thickness into SRLM, i.e. beyond the inner surface
• I then sample in this linear space, explicitally taking the mid-thickness layer provided by hippunfold into account to account for the curvature

I hope this brief explenation is of use to you!

Many thanks,

Viktor