DUNEuro head model computation fails

[romboide2004]

Hi Takfarinas,

Unfortunately I am not allowed to share the data for ethical reasons. I will try and explore the problem myself - it might be related to electrode position indeed.

Thank you,
Ramtin

[tmedani]

If you run the tutorial data without a problem, then the issue is probably in your model/data.
Check the location of the sources and the sensors. Let me know if you need further help.

[romboide2004]

After a number of attempts, I obtained a successful outcome when selecting the low resolution pial as cortex, forcing the source space to gray layer, and building a regular (adaptive won't work) grid within the cortex.

I cannot tell if it is a casual outcome though, I hope this will consistently work for all the recordings.

To better understand how this process works: should the grid ideally be defined to have a size <= than the actual source space (which I presume depends solely on the sensors and should be within the cortex)? And this is to provide an anatomical constraint to the inverse model, right? Or please let me know if I am misunderstanding anything.

Thank you,
Ramtin

[tmedani]

@romboide2004

After a number of attempts, I obtained a successful outcome when selecting the low resolution pial as cortex, forcing the source space to gray layer, and building a regular (adaptive won't work) grid within the cortex.

This can explain why the first error, there are probably some sources outside of the brain mesh when using the other pial cortex.

@romboide2004

should the grid ideally be defined to have a size <= than the actual source space

Do you mean the Grid the FEM mesh?
if so, then the answer is yes; since the FEM source model uses the mesh nodes to model the dipole, so it's better to have a finer mesh than the source grid.

@romboide2004

which I presume depends solely on the sensors and should be within the cortex

For the SEEG you need to make sure that all the sensors are within the brain.

@romboide2004

And this is to provide an anatomical constraint to the inverse model, right?

So, Im not sure what you mean, but basically, yes, the seeg channels' locations and the brain sources are supposed to be within the brain areas.

[romboide2004]

@tmedani

since the FEM source model uses the mesh nodes to model the dipole, so it's better to have a finer mesh than the source grid.

In practical terms, this means that the surface I select as cortex should be contained within the source grid I define when setting up the head model computation, right? If this is the case, in principle DUNEuro should not fail if I select the gray surface as cortex, force the source space to gray and generate the grid within the whole head, am I right?

To clarify: I am working with EEG, not sEEG. The sensors are on the scalp.

[tmedani]

@romboide2004

this means that the surface I select as the cortex should be contained within the source grid I define when setting up the head model computation, right?

If you use the volume grid model, all the sources will be defined within the surface you define as the cortex (all the volume enclosed by the cortex surface).
If you use the cortical surface, the sources will be defined at the vertices of the surface.

@romboide2004

in principle DUNEuro should not fail if I select the gray surface as cortex, force the source space to gray and generate the grid within the whole head, am I right?

If the error is related to the source location, so then in principle, it should not fail,

[romboide2004]

I am working with the volume model. Thank you for the clarification!

[tmedani]

Great then, please take into account my previous message.

@tmedani

PS: Please also note that the FEM may have some instabilities on the results when the dipoles are located on tissue interfaces with high conductivity gradients. This issue is still under deep investigation in the FEM modeling field.

And you can also refer to the main paper and tutorial for more details or ask here