EEG graph analysis for identification of ex-combatants: A machine learning approach

Emotional processing of ex-combatants is affected by chronic exposure to violent events. For a successful reintegration into society, it is necessary to discriminate their brain responses from civilian people, as a first stage to develop treatment strategies. This paper presents a comparative analysis between a Multilayer Perceptron Neural Network and a Fuzzy C-Means classifier to differentiate ex-combatant subjects from civilian controls using graph-based functional connectivity networks calculated from scalp EEG recordings. EEG data were acquired while participants performed an adaptation of a Dual Valence Task, a well-known experimental task commonly used to analyze emotional processing. We obtained up to 80%of accuracy with both proposed algorithms. These results are comparable with other approaches reported in the literature. Together with conductual tasks, the computarized techniques proposed in this paper provide a decision support system to psychologists for determining the level of exposure to the conflict of the ex-combatants, and consequently a tool for their posterior treatment.