Main Page

Wiki for Sharing Methods and Data - Brain Mapping Unit

This is a wiki, a user-editable Web site. You can treat it as a perfectly ordinary Web site (read it), but you are also able to make changes to any page that has an edit link at the top of it. The more or you contribute, the more useful this page will become for current and future members of our group, so please feel free to add any relevant information!

Basic Background Reading

The first thing to do if you are starting out at the BMU is probably to read a couple of easy introductory papers on network science in general, as well as applied to neuroscience. Some useful papers to start with can be found at Introductory Papers

Courses, seminars and groups

Statistics, Programming, Neuroimaging Courses

The Department of Experimental Psychology has a useful list of courses on statistics, programming and neuroimaging here

Sharing Data

In principle, we could put links to data located on the server so as to make it available to others in the group. Please make sure the data in question is free for you to share before doing this. Also, please try to document the data (a link to a paper describing the same dataset should do) and maybe add the name of the person that originally sourced the data.

Network Construction

In brain functional networks, each node corresponds to a different brain region, $i$, and edges or connections between nodes represent statistical associations, e.g., correlations, between the time series, $S_{i}(t)$, recorded at each of these regions. Once the $NxN$ association matrix of correlation coefficients has been evaluated (for $N$ brain regions), it is possible to draw a fully connected, weighted network where the weight of each link corresponds to the correlation strength between the pair $i,j$ of nodes it connects. This network, however, is not easy to analyze and contains many spurious connections resulting from noise rather than genuine correlations. It is therefore usually replaced by a sparser, unweighted network where, following the application of some filtering technique, only the most important connections have been retained as edges in a binary adjacency matrix. One difficulty arises from the fact that the method for filtering out less important connections is arbitrary and has a strong influence on the results obtained.

Beyond Correlations

This section discusses the various forms of statistical association that networks can be based on.

Filtering Methods

This section discusses the various filtering methods that can be used.

Network Analysis

All commonly used network measures have been implemented in MATLAB and bundled into a wonderful toolbox called the Brain Connectivity Toolbox by Olaf Sporns and collaborators. [[1]] They also welcome any additions to their toolbox, so feel free to contribute.

For the less computationally fluent: Yong Liu is developing a point-and-click toolbox of his own... it should be ready in late 2011.

Network Visualization

A variety of tools have been developed to generate efficient and pleasing graphical representations of networks:

• PAJEK is one of the most commonly used ones: [[2]]

• Gephi is looking increasingly poliched: [[3]]

• GUESS is my personal favourite [[4]]

Other Visualization tools

• Caret allows the creation of beautiful brain surface maps: [[5]]

How To Edit the Wiki

• You can find some help on editing this wiki on HowToEdit