Lesion type misidentification: EEG potential sampling and source reconstruction errors.

Abstract

Accurate EEG source reconstruction needs an appropriate volume conductor head model including, in the presence of a morphological brain lesion, a lesion compartment. Lesion electrical properties (conductivity) can not be measured in vivo and need to be retrieved from literature on the base of lesion type identification, performed by means of diagnostic imaging. However, different pathologies can appear similar at bioimages inspection, leading to uncertain diagnosis and in turn to wrong lesion conductivity assumption. Besides many factors, source reconstruction accuracy depends on lesion conductivity value and on electrodes placing (scalp potential sampling). We investigated the relation between electrodes sampling of the EEG and source reconstruction accuracy in case of uncertain lesion type identification. We adopted an eccentric-spheres head model (including the lesion) and a dipole source. We simulated several pathological conditions considering different electrode montages on the scalp: a "clinical like" sampling with 64 electrodes and an extended sampling with 128 electrodes. Source reconstruction was performed assuming wrong lesion type identification (i.e. introducing an error in the model parametrical setting). We found large errors for both source localization (LE) and source intensity estimations. Maximum LE could be reduced (from 30 to 27 mm) by the extended sampling. Conversely, the mean values of LE mostly increased using more electrodes. The benefit of enlarging the sampling was clear for the intensity estimation. In conclusion, only the most inaccurate source reconstruction could be improved changing electrodes placing; the inherent source reconstruction error due to model parameter setting could not be avoided.