Connectivity - how to

Hello everyone!
For the first time I am trying to make some connectivity analyses on my resting state data.
I am facing an ‘out of memory’ error for each kind of connectivity analysis (correlation 1xN or NxN, coherence, etc…) I run, with standard options or custom ones. I tried selecting one scout or more. My workstation is provided with 16GB RAM and a good chipset.
I also tried to reduce the number of vertices in the anatomy folder, but I faced the same error. It only works with correlation AxB with a few scouts.
I thought that probably I make some mistakes in the way I proceed.
That is my pipeline:

  • refine using digitized head points and warp default anatomy
  • apply band pass and band stop filter, remove artifacts with SSP
  • calculate noise covariance matrix from empty room record
  • compute head model (overlapping spheres) and source model (2018, MN, constrained)
  • process the source map with connectivity analysis.

I hope you can help me!
Thanks a lot!

Angelantonio

Hello

Your analysis pipeline looks good.
However, connectivity in source space is very resource-consuming. You do not mention any epoching, so you are maybe processing very long time series, then even with 20 vertices per scout, it could go completely out of control if you have 50000 time samples.

You could reduce the data size by downsampling the signals, reducing the number of sources in your scouts, or use the option “Scout function: after”, somewhat documented in the time-frequency tutorial: https://neuroimage.usc.edu/brainstorm/Tutorials/TimeFrequency#Scouts

Cheers
Francois