Only few features are supported

[khanshigri]

I am using this plugin to calculate various FC features. However, this plugin only calculates MIM, MIC, and GC. For other features such error appears e.g. for CS "Warning: Only the first principal component will be used to determine CS "

[nguyen-td]

Hey there,

CS and other FC metrics are still being computed, but only the first PC will be used.

After source reconstruction, the sources are assigned to their respective regions of interest determined by an atlas, e.g., to 68 regions if the Desikan-Killiany atlas is chosen. There are usually many sources that correspond to a single region. We therefore use PCA to reduce the dimensionality, the default is three princple components. Since most FC metrics are univariate quantities (and based on the cross-spectrum), they only compute connectivity between a single signal in region 1 with a single signal in region 2, so connectivity is only computed between the first princple components of the connected regions.

MIM and MIC can compute connectivity between two multivariate signals so all principle components can be used.

[khanshigri]

Hi, thanks for the response. I am curious if there's any explicit manual to use this plug-in effectively? if so then please share as it's very hard for me to interpret these results from MIM, MIC and GC. Other parameters are not working for me

[nguyen-td]

The only written instruction we have is the README. We are working on a tutorial but it's still work-in-progress and it's uncertain how long it will still take.

As for the interpretation, there is a paper referenced in the README that compares different connectivity metrics in simulated noise settings. For some basic intro into connectivity, Bastos & Schoffelen, 2016 is a good place to start. Other toolboxes like MNE-Python have complete tutorials, e.g., this is an excellent manual on multivariate metrics (MIM and MIC) in Python. That implemention in particular is pretty much identical with the one in ROIconnect.