Kolloquiumsvortrag: Dr. Carsten Wolters / Universität Münster
Institut für Informatik, Ludewig-Meyn-Str. 2, Übungsraum 2
Dr. Carsten Wolters
Modern numerical mathematics for source analysis from combined EEG and MEG data with applications in presurgical epilepsy diagnosis
Abstract:The accuracy of the inverse problem in Electro- (EEG) and Magnetoencephalography (MEG) source analysis strongly depends on the accuracy of the forward problem. For the forward problem, numerical approaches are needed to compute head surface field distributions from dipolar current sources in the human brain using realistic head volume conductor models derived from MRI sequences. For the forward problem, we will compare numerical and modeling accuracies of different procedures, under which are Lagrange and discontinuous Galerkin finite element methods, double-layer and symmetric boundary element methods. Both simplified multi-layer sphere models where quasi-analytical solutions exist as well as realistic head models will therefore be used. For the inverse problem, hierarchical Bayesian modeling and dipole scanning methods will be studied. We will then apply the developed methodology to a patient suffering from pharmaco-resistant focal epilepsy, where we will show that combined EEG/MEG source analysis using a calibrated realistic FE head model outperforms single modality EEG or MEG alone with regard to an improved non-invasive estimation of the seizure onset zone as well as an improved non-invasive reconstruction of the propagation pathway. Validation is carried out using additional depth electrode (sEEG) measurements.