Zeros in imaging kernel matrix after dipole modeling

Bugs
1

Hello. I am attempting to compute sources with the option "Dipole modeling". The results did not make sense, and then I noticed after exporting the sources to the MATLAB workspace that the field ImagingKernel is mostly zeros. This reminded me of a problem with zeros in the Gain matrix in the head models from OpenMEEG that we saw before, so I thought there might be a problem here. But I can confirm that now all my Gain matrices from my head models do not contain zeros.

Please let me know if I can provide some data for someone to replicate this issue. Below is a figure and a command showing the zeros. Thanks,

Karl

Data exported as "dipoles"
>> figure;imagesc(dipoles.ImagingKernel);colorbar
>> numel(dipoles.ImagingKernel)

ans =

     4212230

>> numel(find(dipoles.ImagingKernel==0))

ans =

     4111229

kernel
kernel2333×1750 68.9 KB

2

Hello
Sorry for the response delay.
Do you get the same type of result if you compute a minimum norm solution?
What kind of source space / head model is this? (surface, volume, mixed?)

3

In fact I did notice that this does not occur with a MNI solution:

Data exported as "mni"
>> figure;imagesc(mni.ImagingKernel);colorbar
>> numel(find(mni.ImagingKernel==0))

ans =

     0

mni
mni3500×2625 180 KB

This is a mixed source space with a head model produced by OpenMEEG with default parameters. The mixed source space includes ~15,000 vertices of entire cortex + brainstem and thalamus vertices.

4

Ok, this is probably an issue with the mixed head models not being handled properly with the dipole modeling option.

@John_Mosher we need you here

5

Thanks @Francois, I am still waiting to see if there is a solution to this problem, assuming it is an effect of a bug. Are we still waiting for @John_Mosher?

6

No news from John… sorry…

Corresponding git issue:
https://github.com/brainstorm-tools/brainstorm3/issues/118