EEG dipole source localization using artificial neural networks

Localization of focal electrical activity in the brain using dipole source analysis of the electroencephalogram (EEG), is usually performed by iterati vely determining the location and orientation of the dipole source, until o ptimal correspondence is reached between the dipole source and the measured potential distribution on the head. In this paper, we investigate the use of feedforward layered artificial neural networks (ANNs) to replace the ite rative localization procedure, in order to decrease the calculation time. T he localization accuracy of the ANN approach is studied within spherical an d realistic head models. Additionally, we investigate the robustness of bot h the iterative and the AWN approach by observing the influence on the loca lization error of both noise in the scalp potentials and scalp electrode mi slocalizations. Finally, after choosing the ANN structure and size that pro vides a good trade-off between low localization errors and short computatio n times. we compare the calculation times involved with both the iterative and ANN methods. An average localization error of about 3.5 mm is obtained for both spherical and realistic head models. Moreover, the ANN localizatio n approach appears to be robust to noise and electrode mislocations. In com parison with the iterative localization, the ANN provides a major speed-up of dipole source localization. We conclude that an artificial neural networ k is a very suitable alternative for iterative dipole source localization i n applications where large numbers of dipole localizations have to be perfo rmed, provided that an increase of the localization errors by a few millime tres is acceptable.