Conductivity estimation with EEG/MEG brain source localization in a finite element head model.

摘要

Brain source localization with EEG and MEG modalities provides a useful means of identifying and localizing bioelectric source in the brain and has been used as an important tool in neuroscience and in clinical applications. Due to modern imaging technology, one can construct a subject specific volume conductor model from a set of MRI or CT images that can improve the accuracy of source localization over generic models. Finite element method makes it possible to use the realistic geometry from the subject specific imaging data and to assign tissue conductivity in a flexible way. The dissertation works use the FEM volume conductor model and studied the following scientific issues in FEM source localization.;The first study is to investigate the impact of dipole models and numerical solvers on solution accuracy and computational efficiency. The accuracy of the forward solution has a direct impact on the accuracy of inverse localization that reconstructs current dipoles from head surface potentials by means of iterative forward problems. We studied the impact of dipole models (Venant, partial integration and subtraction) on the accuracy of forward solution by EEG simulation and evaluated the computational efficiency of the FE solvers (AMG-CG, IC-CG, Jacobi-CG). The second study is to estimate the tissue conductivity with EEG data during source localization. Bioelectric source analysis is sensitive to geometry and conductivity properties of the different head tissues. We developed a Low Resolution Conductivity Estimation (LRCE) method using simulated annealing optimization on high resolution finite element models that individually optimizes a realistically-shaped volume conductor with regard to the tissue conductivities. The third study is to stabilize the LRCE by adding MEG modality to the EEG. The combined analysis with an iteration scheme takes the source parameter from the MEG dipole fit that has much less sensitivity to conductivity and uses it as a prior constraint on the source for the EEG LRCE. We have shown the viability of an approach that computes its own conductivity values and thus produces a more robust estimate of current sources. Using the LRCE method, the individually optimized volume conductor model can be used for the analysis of clinical or cognitive data acquired from the same subject.