Cortical Parcellation and Classification using PageRank Clustering and the Small-Worldness of ADHD

More Info
expand_more

Abstract

Networks and graphs are all around us and can represent a variety of models: from airline connections, the World Wide Web to the way people are connected on Facebook. There are are a few classes of networks with interesting properties such as structured and random networks. One particular interesting class is the small-world network, with a short path length between nodes and a high degree of clustering. This network emerges in many (real-world) phenomena, like the models mentioned above, as well as biological models such as protein-protein interaction and neu- ronal connections in the human brain. We discuss the differences in small-world properties in the anatomical brain for typically developing children and children diagnosed with Attention-Deficit/Hyperactivity Disorder. Using the differences we investigate the possibilities for classification between the populations. Secondly, we propose a clustering algorithm that is based on spectral clustering and uses the im- portance of nodes in a network to find clusters. This algorithm is similarly applied to fMRI data and used to reduce the dimensionality before classification.

Files