Purpose: We investigated the accuracy and potential for serious error when representing cortical generators of epileptic spikes with the common single-dipole model. Spike generators were realistically simulated with cortical areas of different extents.
Methods: The source was simulated by using a patch that comprised small triangles on the cortical surface, each triangle having an elementary dipole generator with a moment corresponding to real intracerebral fields of spikes. The source-patch covered various clinically important parts of the temporal and frontal lobes, with an area ranging from 6 to 120 cm2. The scalp field was computed for each source-patch by using a realistic head model and was fitted by the single-dipole model to determine the best-fit dipole and the intracerebral distribution of residual variance (RV). Dipole modeling also was performed for the simulated scalp field with additional real EEG background.
Results: The RV after fitting a dipole to the scalp field without noise was at most 1.34%. Scalp spikes arising from sources of 6 cm2 were of small amplitude, and the dipoles estimated for these spikes were inconsistent. Extension of the source area was associated with increase of scalp potential amplitude, only very small increase of RV, and increased consistency of the estimated dipoles. When the source was very large, the dipoles clustered at very misleading locations.
Conclusions: Pitfalls in dipole source localization are caused by the procedure of fitting the simplistic dipole model to real cortical sources with spatial extent and complex configuration.