We recently discussed with Alex about the importance of checking the noise covariance prior to estimate the sources. While exploring one of our dataset in BST, we visualized the covariance (right click on the Noise covariance, then “Display as image”) but couldn’t really read it because all of the sensors were on the same image (color scale cannot be the same for all modalities). We were therefore wandering if you would kindly arrange a way to visualize the covariance matrix per modality (channel types)?
Francois is currently attending the HBM meeting in Seattle. This is a great suggestion you have here and it should not be too much of a problem for Francois: we could show one figure per sensor type in addition to the entire covariance array. In the meantime, I don’t think it would be too difficult for you to load the covariance array into Matlab (Right-click over file and select: File > Import in Matlab) and display the contents of a subset of rows and columns using the matlab’s imagesc command.
Done. I've added some menus that allow you to show either all the sensors at once, or only the selected modality.
Not that that in the "All sensors" view, the values were already normalized separately for each modality. Attached an example of display for a noise covariance matrix containing MEEG+EEG+SEEG.
All the blue areas that you see in those displays are not due to a bad scaling, the value is really zero everywhere. When you use Brainstorm to calculate this noise covariance matrix, it calculates only the covariance within modality, for the following sensor types: MEG, EEG, SEEG, ECOG. All the other sensor types are ignored (as they can't be used in the source estimation process).
