| Accurate analytical EEG localization of neural signals could help reduce the need for surgery in some epilepsy patients. In order to achieve accurate EEG localization of neural signals using analytical techniques, it is imperative to know the head model's conductivity parameters. As new and more accurate head models are developed, their conductivity parameters must be specified. However, because the conductivities of human head tissue have never been determined to any appreciable extent, no standard approach has been available.; In this dissertation, model tuning is introduced as a means to specify the optimal conductivity parameters of an arbitrary EEG head model through a nonlinear least squares fit of the head model to in vivo EEG measurements. This method eliminates the difficulties of obtaining fresh human head tissue samples on which to measure conductivity and the subsequent problem of translating these measured conductivities into the conductivity parameters for the given EEG head model.; Model tuning makes use of intracranial depth electrodes that are stereotactically implanted into an epilepsy patient for diagnostic purposes. A known current is applied between two selected contacts on one of the depth electrodes to create a known artificial neural signal. Potential measurements are then acquired using EEG electrodes placed on the epilepsy patient's scalp.; Model tuning was tested on the four shell model with the participation of three epilepsy patients who had received depth electrode implants. The conductivity parameters that were empirically determined through model tuning were 0.2872, 1.2707, 0.0026, and 0.2048 for brain, CSF, skull, and scalp.; To illustrate the kinds of questions that can be addressed with model tuning, it was hypothesized that the four shell model conductivity parameters specified by model tuning would lead to more accurate localization. The localization accuracy of the four shell model using the new conductivity parameters was tested against that of the four shell model using the standard conductivity parameters. The difference in localization accuracy was insignificant, showing that the hypothesis was wrong. This result can be attributed to the geometric shortcomings of the four shell model.; With the possibility of noninvasive model tuning, EEG head models could be routinely customized to each individual for even better localization accuracy. Because of these advancements, the use of analytical EEG localization to reduce the need for surgery in some epilepsy patients seems a real possibility. |
