Questions about defined number of dipoles for BEM and compromises of using less vertices in meshes

Discussions
, , ,
1

Hi everyone! I'm trying to generate some BEM solutions for my masters thesis, but I'm having trouble doing so for lack of memory, I'm working on a workstation with 32 Gb RAM and keep getting into crashes due to a high number of vertices in my meshes.

My goal is to compare the activity sources of a SSEP with different values of the conductivity ratio, and my supervisor advised me to use high resolution meshes so the results could be comparable with the ones of another student who is working on something similar.

So my questions are, I see that the mesh that generates the dipoles has 15002v and I'm trying to make the model with at least 8000 triangles in each surface. What is better to compromise on? The number of dipoles used for the dsm command on openmeeg or the number of vertices on one of the meshes?

Also I noticed that outer and inner skull must be the same to have better accuracy, so which is better to set a lesser number of vertices on, the scalp or the pair of skull meshes?

And for last, I'm also curious why Brainstorm uses the 15002v mesh for the cortex and thus the dipoles, I read here Open MEEG BEM on Colin27 - Discussions - Brainstorm (usc.edu) that the value was chosen to have a good compromise between accuracy and usability. Do the dev team has information on this? I would like this info to back up why I'm choosing these dipoles.

Thanks in advance for the responses and I hope I'm not asking dumb questions, also I would like to thank the dev team and researchers behind Brainstorm, it's a terrific piece of software that has helped me to streamline my thesis protocol.

Best,

-Óscar

2

Thank you for the appreciation, Oscar!

The cost factor is the resolution of the surface meshes used for the head envelopes, not so much the number of elementary sources distributed over the cortex. Hence you may consider reducing the number of triangles for the scalp first. If this is not enough, then reduce resolution of the skull envelopes (they are the most sensitive in terms of accuracy because of the thin skull bone layer).

The magic 15002 number is purely pragmatic: with this amount of vertices, the anatomy of the cortex remains relatively well resolved spatially, with all folds etc.

1 Like
3

Thank you for the reply Sylvain! Your response was really helpful.

And I'm sorry for replying so late, I got caught up in other stuff and just recently started working on my project again.

Cheers!

-Óscar