Dilating of brain mask and some other general queries

[Michiko]

@rsmith

In this sort of scenario, I would recommend dilating by only maybe 1-2 voxels, and then using the -seed_gmwmi option. The GM-WM interface seeding operates as described in the ACT paper. However, it relies on finding a local tissue partial volume gradient in order to localise the interface; this means that initial seeds drawn deep within the cortex will not lead to a successful streamline seed, which is problematic for FreeSurfer cortical ROIs where only voxels that predominantly contain cortex are included in the mask. So a slight dilation helps that algorithm to find the tissue interface more reliably, and with a more consistent coverage.

Hi @rsmith,

Thanks for your reply regarding this. Just to be really certain of what is going on with ACT, do you mean that in a typical scenario without specifying the seed_gwmi option, ACT does not do GM-WM interface seeing but only WM seeding?

Based on what you know (the literature, personal experience), is there a significant difference with dilating the cortical ribbon and providing this mask during ACT? I’ve also read that you mentioned in a previous post that tck2connectome ultimately does a 2mm radial search to find the nearest GM node - would this be sufficient then if one were to be doing a connectome type analysis?

Lastly, taking the example from the structural connectome tutorial, if one were to be using the tissue-segmented image provided by FSL for the 5ttgen step, how would you generate this, given that the GM nodes are specified in the aparc+aseg file by Freesurfer? It does get a little confusing with this since I have the impression you would want to dilate 1-2mm of the GM labels into the WM but then you would be providing a tissue-segmented image using FSL instead.

Thank you so much.