Comparison of distributed source localization methods for EEG data

A Seeland1, S Straube2, F Kirchner1

1Robotics Innovation Center, DFKI GmbH, Germany
2Robotics Group, University of Bremen, Germany

Contact: anett.seeland@dfki.de

To improve the spatial resolution of EEG data distributed source localization can be used. Various methods have been proposed in the literature, but a key problem is that there is no standard procedure and metric how to evaluate and compare them. Moreover, along with simulations further evaluations on empirical data are required [Pizzagalli, 2007, in: Handbook of Psychophysiology, Cacioppo et al, Cambridge, Cambridge University Press]. Hence, we present a comparison of wMNE, sLORETA and dSPM reconstruction methods on movement intention data of 7 subjects. We compared the methods based on three criteria: (i) the distance of the nearest activation cluster to a reference region derived from literature, (ii) the number of found activation clusters and (iii) the difference in activation between the two conditions “movement” and “rest”. The comparisons were performed on the average ERP and single trial data, respectively. For both levels the same qualitative results were obtained: wMNE reconstructions had the smallest distance and highest contrast, followed by sLORETA and dSPM. However, by using wMNE the number of found sources was much higher than for the other methods. The proposed approach provides a framework for a fair comparison between existing distributed source localization methods.

Up Home