Problem with scales on source dynamics in certain Scouts

Discussions
, ,
1

Hello,

I am contacting you regarding a technical issue. In my work, I use the Desikan-Killiani atlas, which we modify to divide the left precentral and left postcentral regions into two subregions of equal size. This gives a total of 70 regions.

We have a set of 70 dynamics from a simulation. A dynamic corresponds to the average activity of a region; in other words, we allocate the same dynamic to each vertex in a region.

The idea is that in the precentral2 and postcentral2 regions, the dynamics are linked by the presence of peaks of the same amplitude but with a time lag.

Thus, a region containing 50 vertices will have the same dynamic replicated 50 times. If we average the activity of the region, we find the original activity.

In Brainstorm, I observe discrepancies with the exported sources.

When I observe the average activity of the region sources, there is a weighting on some of them. The right precentral has a weighting factor of 10^-5 and the left postcentral has a weighting factor of 10^-1.

I only observe this problem in Brainstorm. If I look at the exported sources or even the source file in the Brainstorm database, I don't see this weighting.

In addition, I compared the correlation between my regions in Brainstorm and Matlab by calculating the difference between the instantaneous correlation in Matlab and the correlation in Brainstorm with an averaging of the activity per region upstream to obtain a [70x70] matrix. The difference is zero everywhere except for my subregions postcentral1 and 2, precentral 1 and 2, and the right precentral region.

I would like to point out that if I take the .mat file from the Brainstorm database, I do not have this problem.

Thank you for your attention to my message.

Best regards,

Maël

2

Since there is only one dynamic (aka signal) per vertex, we're talking of constrained sources right?

One thing to consider here is the orientation of the vertices in each Scout. For constrained sources, their orientation is normal to the cortical surface. So, for large Scouts with sulci and gyri, there are vertices that have an opposite orientation, but the same signal, to obtain the Scout average, adding the signals without considering the orientation would not be correct, to avoid so, when computing Scout averages in Brainstorm, the sign of the signal is adjusted to reflect the vertex orientation and then the signals are averaged. Check this link:
https://neuroimage.usc.edu/brainstorm/Tutorials/Scouts#Sign_flip

This is why you are seen an attenuation on the Scout average.

  • Correlation is computed with the same Scout averages?
  • How are the Scout averages obtained?