EEG source localization sensitivity due to brain lesions modeling errors

For accurate electroencephalogram-based (EEG) localization of neural sources correct modeling of brain lesions geometry and tissue conductivity is required. Lesion properties are derived from anatomical images like CT or MRI. According to imaging modality, lesion can appear of different size and shape. Conductivity parameters are taken from standard references, despite the large variability in the available data. The uncertainties in lesion conductivity assignment (LCA) and in determining exact lesion geometry affect source localization accuracy. The aim of this paper is to quantify the combined effect of these uncertainties on EEG dipole source localization accuracy. The study was conducted using an eccentric-spheres model of the head in which a modifiable eccentric bubble approximated various brain lesions. In 32 simulated pathological conditions the inverse dipole fitting procedure was carried out assuming an incorrect (under/overestimate) lesion dimension and conductivity. Errors in lesion modeling led to markedly wrong source reconstruction even for small differences between the actual lesion and its model. Localization errors up to 15.4 mm demonstrate the requirement of an accurate parametric setting of the model to achieve localization accuracy within few millimeters.