Networks & Connectomes

Build the network a brain model runs on: connectomes, parcellations, coupling and heterogeneity.

2Specify

Part of the running example, where stage 5 adds a connectome and a coupling function.

A whole-brain model is dynamics on a network. Three things have to be decided: where the connectome comes from, how regions couple, and whether every region and connection behaves the same way.

The first decision is usually the quickest. TVB-O ships curated connectomes you can reference by IRI, so most experiments never build one. You build your own when the parcellation, the cohort or the tractography method has to match something specific.

Getting a network

Network specification

The Network spec itself: constructors, storage format, and the built-in connectome catalogue.

Build a connectome from tractography

A tractogram plus a parcellation atlas becomes a schema-compliant network.

How activity travels

Coupling functions

The rule that turns one node’s state into the signal its neighbours receive.

Coupling is where most first experiments go wrong, in two specific ways: the coupling key must be the model’s own input port rather than a name you invent, and a raw connectome carries streamline counts in the hundreds of thousands, which will drive any model to NaN unless it is normalized. Both are covered on that page.

When regions differ

Heterogeneous nodes

Giving regions their own parameters.

Heterogeneous edges

Giving connections their own behaviour.

Hierarchical & surface networks

A fine mesh layered onto a coarse parent network.

For a region with genuine internal structure rather than different parameters, see Beyond the mean field.

Reference

Network · Node · Edge.