Functional brain imaging by EEG graph-matching.

Our goal is to image the brain activation and function by the mean of electroencephalogram signals. The work's originality is that we build a data structure, a graph, that sums up the brain activity in the spatial, temporal and frequency domain. This graph is formed from the information included in the EEG time-frequency map. Contrary to methods trying to reproduce or analyze the whole complexity of the signal, our method is based on a multi-scale approach. In this order the level of information extraction could be adapted. So as to obtain a pattern of activation or to compare different EEG signals, we use some techniques of graph-matching on our data structure. The developed algorithm is based on the A* algorithm that allows us to compare variations of the recorded EEG answer in term of latency, frequency, energy and activated areas. The first results of this new project show on the one hand that the graph is a good choice to sums up the cortical activity and on the another hand that the graph-matching offers some interesting perspectives in order to describe the functional brain activity.