On the Volume Conduction Model Validation with Stereo EEG Data

Volume conduction can be defined as the transmission of electric potential and magnetic field generated by a primary current source of brain activation in the surrounding medium, i.e., the human head. Volume conduction simulations are based on sophisticated models whose construction represents a current challenge within the neuroscientific community. Volume conduction models are used in various applications such as electroencephalography (EEG) or magnetoencephalography (MEG) source reconstruction, or in the optimization of the electrode placement in a transcranial electrical stimulation session. Clinical applications based on volume conduction models are, for example, the localization of the epileptogenic zone, i.e., the brain area responsible for the generation of seizures, in the presurgical assessment of focal drug-resistant epilepsy patients, and the antidepressant effects given by transcranial electrical stimulation. Increasing the accuracy of volume conduction simulations is therefore crucial. To the best of our knowledge, the accuracy of volume conduction models have never been validated directly with actual measurements in human patients. The main goal of this thesis is to describe a first attempt to validate volume conduction modeling using electric stimulation stereo-encephalografic (sEEG) data. This work therefore is focused on the research, investigation and test of tools and methods which can be used to describe the accuracy of volume conduction models used in both clinical and basic research. Given a dataset of one pharmaco-resistant epilepsy patient, composed by the anatomical T1 weighted magnetic resonance image (MRI), the electrophysiological signal recorded during electric brain stimulation sessions with sEEG technique and sEEG contact positions extracted by post-implantation CT image, the analysis conducted in this work can be split into three main steps. First, we built volume conduction head models and we simulated the electric potentials during the electric brain stimulations. In this step, we solved the so-called (s)EEG forward problem by means of the finite element method in its classical formulation, and we considered three different conductivity profile to assign to the computational domain, individually extracted by the T1-w MRI. Moreover we computed the solution in meshes with two different resolution, i.e., 1 mm and 2 mm, with three different ways to model the source term, i.e., the partial integration approach, the subtraction approach and Venant’s approach. Second, we extracted the responses to the electric brain stimulations from the actual sEEG measurements. Particular emphasis in this step was given to the optimal referencing systems of sEEG electrodes. Third, we compared the simulated and measured potentials for each of the three volume conduction head models, both in a single shaft and global comparison. The comparison results in overall high relative differences, with only slight modulations given by the distance from the stimulation site, the underlying volume conduction head model used and the compartment where the dipolar source is located. Simulation results show that the computation of sEEG forward problem solution is feasible with the same scheme adopted for scalp EEG in the duneuro software (http:// duneuro.org/), and it is stable for different mesh resolutions and source models also for intracranial electrodes, i.e., for electrodes close to the source positions. From this first validation attempt, we can conclude that the distance contact-source modulates the relative error between measured and simulated potential; for the contacts in the white matter compartment we observed the most accurate results, and the results relative to the three and four compartment results were more accurate than the ones relative to the five compartment results. While we achieved topographical errors within 10% for most of the shafts, the amplitude of simulated and measured potentials notably differs.