Brain Networks

Deriving connectivity information from diffusion MRI has become a standard approach for understanding brain anatomy. We are currently developing tools to generate brain connectivity networks, but one technical challenge is how the networks are represented. This mostly affects the exchange of data between tools that generate and process the networks, tools for visualizations, and tools for statistical analysis. This page outlines some of the technical concerns for dealing with these networks.

Requirements

Multiple networks

• hundreds to thousands of subjects
• would like to compare in visualizations

Node attributes

• names of regions of interest
• several numerical anatomical measures

Edge attributes

• names of intersecting tracts
• a few dozen numerical anatomical measures

File Formats

Name	Pros	Cons	Support
CSV	Simple to parse, unix friendly	Can't effectively handle multiple networks, many flavors	Excel, Paraview, Cytoscape
GraphML	Flexible, rich features	Supports more features than necessary, larger file size	Gephi, prefuse
CML	Plays well with other connectome software	Doesn't play well with anything else	Connectome Viewer
NIfTi	Good for comparing multiple networks	A hack. Doesn't store node names	FSLView and others

References

UCLA Multimodal Connectivity Database

Human Connectome

Human Connectome Project

Open Connectome

Connectome Viewer

Connectome Mapping Toolkit

NiPy