Abstracts

Rationale: Identification of the seizure onset zone with stereo-electroencephalography (sEEG) can benefit from rigorous inverse source localization algorithms. Such algorithms require simplified models of neural sources. Dipoles are an appropriate source model for EEG because the recording sites are far from the neural sources. However, because the recording sites for sEEG are much closer to active neurons compared to EEG, an appropriate simplified source model for sEEG is unclear.

Methods: We simulated single neuron activity using a library of biophysically realistic cortical neuron models and population neural activity using a cortical column network model consisting of 336 spatially extended neuron models. We compared the spatiotemporal voltage distributions generated by the neural sources to an ideal dipole by computing correlation coefficients between their voltage distributions at increasing distances from the source. We then constructed an anatomically realistic volume conductor human head model with implanted SEEG electrodes. Using this model, we compared the voltages generated at the electrode contacts by a single dipole and by extended region of cortex (0.5 – 9.5 cm²) containing many dipole sources to determine the minimum distance that an extended region of cortex can be well represented by a single dipole. Finally, we compared source localization errors, using sLORETA, for a dipole source compared to realistic single neuron and cortical column sources, when the lead-field matrix was generated by dipole sources.

Results: At source to recording distances >0.5 mm, pyramidal cells were well represented by an ideal dipole at the peak of an extracellular recording. Over the time course of an action potential, all pyramidal neurons oscillated between being well represented by an ideal dipole down the somatodendritic axis to up the somaodendritic axis. These spatial and temporal findings were consistent for the cortical column model, where the column was well represented by a dipole at distances >1 mm and the correlation with an ideal dipole rapidly oscillated over the time course of the oscillatory activity in the column. At the spatial scale of cortical regions, a single dipole will yield >50 µV error within 1.5 cm of an electrode, compared to patches of active cortex ( >5 cm²). Additionally, source localization errors when using dipoles to represent realistic neural sources were < 5 mm.

Conclusions: Single dipoles are an appropriate source model to represent both single neurons and small populations of cortical neurons for sEEG. However, many dipoles are needed to appropriately represent large regions of active cortex.

Funding: Please list any funding that was received in support of this abstract.: Duke MEDex. Robert Plonesy Fellowship in Biomedical Engineering. Duke CTSA Grant UL1 TR002553.