Metrics Applied to Brain Neural Networks

This work is framed within the field of medical signal and image processing applied to patients with refractory epilepsy. In this context, a computational method is proposed to characterize the topology of the Epileptogenic Zone Network (EZN), responsible for generating seizures. Based on Granger connectivity, applied to stereo-electroencephalography (SEEG) records and their spatial location, different metrics are calculated that provide complementary information to make the stimulation process more efficient, reducing its error. Neurophysiologists must perform a visual analysis of the SEEG and build an imaginary network selecting those candidate nodes to be stimulated. This process is complex, subjective and time-consuming, so the team of specialists requested the development of a computational method that helps make such selection easier. This work aims to characterize a three-dimensional network that allows optimizing the evaluation process of the nodes, thus corroborating the results obtained by both methods (visual and computational). The computational methods applied to the studies of 2 patients showed a 60% similarity with respect to the analysis carried out by the medical team.