Accuracy of EEG source reconstruction in the presence of brain lesions: modelling errors and surface electrodes' placement.

Source localization techniques based on electroencephalography (EEG) use scalp potential data to infer the location of brain neural activity. A volume conductor model describing the electrical properties of the human head is needed. Lesions have conductivity considerably different from that of normal brain and should be included in the head model because the differences between the actual head and the model can cause source reconstruction errors. We performed a simulation study investigating EEG dipole source reconstruction errors, caused by brain lesions neglecting, using different measurement montages. The scalp was sampled by 64 electrodes (simulating clinical practice) and by 128 electrodes (extended configuration). The human head was represented by an eccentric-spheres model in which a modifiable eccentric bubble approximated various brain lesions. We analyzed 64 pathological situations. Results showed that neglecting brain lesions in source reconstruction procedures could cause large source localization errors which depended on source location and orientation, and varied with EEG montage. The maximum source localization errors (LE) were 2.5 cm and 1.4 cm for the 64 and 128 electrode configurations respectively. The largest errors occurred for sources nearby the lesion. LE was not systematically smaller with 128 rather than 64 electrodes. Maximum intensity errors were similar for 128 and 64 electrodes. The EEG inverse dipole solution was maximally sensitive to the electrode configuration on the scalp when the source was located deep in the brain (e.g. in the brain stem). We concluded that deep source localization needed an extended sampling of the scalp.