Reading Group: Pattern Recognition with Brain Connectivity Graphs

SpeakerJonas Richiardi
AffiliationEPFL, Switzerland
DateWednesday, 28 Sep 2011
Time16:00 - 17:00
LocationFoster Court 218
Event seriesMachine Learning for Neuroimaging Reading Group

Beyond classical "brain decoding" based on BOLD activations, another view of the data can be gained by looking at functional connectivity, which considers temporal correlations between brain regions of the brain at rest or during tasks. Such connectivity is typically represented as a graph, and much literature has focused on group-level, post-hoc analysis of such graphs in terms of properties such as local efficiency, clustering coefficient, and so on. However, it is of interest in many applications to be able to perform predictive modelling with connectivity graphs. This talk will focus on an emerging technique we recently proposed, connectivity-based decoding. This is an interesting tool for neuroimagers and provides complementary information to both activation-based decoding and qualitative analysis in terms of graph-theoretic properties. It is applicable to both brain state decoding and clinical applications such as diagnosis. After a whole-brain regional functional connectivity graph has been established, the problem can be cast as a weighted graph classification task. We will show that the graphs of interest form a restricted class of graphs whose properties prevent the application of classical graph matching techniques to elicit a useful distance or dissimilarity between graphs, and advocate for the use of modern graph embedding methods. We will present several vector space representations of graphs that are suitable for the class of graphs of interest, and discuss experimental results, in particular focusing on our recent NeuroImage paper.

iCalendar csml_id_56.ics