Nitin B. Bangera1, Donald L. Schomer2, Steve Papavasiliou2, Don Hagler3, Nima Dehghani3, Istvan Ulbert4, Eric Halgren3, and Anders M. Dale3
1. Department of Biomedical Engineering, Boston University
2. Department of Neurology, Beth Israel Deaconess Medical Center, Boston, MA
3. Multimodal Imaging Laboratory, University of California, San Diego
4. Institute for Psychology of the Hungarian Academy of Sciences, Budapest, Hungary
Forward solutions with different levels of complexity are employed by clinicians and scientists to calculate the inverse solution, which gives the source of current generator responsible for electric fields, and magnetic field recorded using Electroencephalography and Magnetoencephalography respectively. The accuracy of the forward solution directly affects the localization accuracy of the model being used. The influence of tissue inhomogenity and anisotropy (of white matter) on the forward solution is poorly understood. Localization accuracy of forward models is difficult to test since the exact location of the source is unknown in most cases. In patients with intractable epilepsy, depth electrodes are clinically implanted to monitor epileptic activity for surgical pre-evaluation. Artificial dipoles are created at known locations in these patients by injecting biphasic square current pulsed though implanted electrodes (fig 1). The in vivo depth stimulation data provides a direct validation of the forward model.
The goal of this study is to report the influence of inhomogeneties in brain tissue and white matter anisotropy on the EEG forward solution by comparing experimental data with data predicted by a realistic head model solved using the finite element method (FEM).
– Recruit Patients undergoing depth electrode implantation
– Patients are scanned for CT, MRI and DTI images before implantation
– Post-implant CT images are obtained for electrode localization
– Simultaneous intracranial (iEEG) and scalp (EEG) electric fields are measured during current injection through contacts on depth electrode
– Current levels are below threshold for neural damage.
– Accurate segmentation of the brain tissue type is implemented using a semi-automatic method using multi-spectral MRI scans (different flip angles) in conjunction with the regular T1-weighted scans and CTs (fig 2).
– The electrical conductivity in the anisotropic white matter tissue is quantified from the water self-diffusion tensor, as measured by diffusion tensor MRI.
– The finite element model (FEM) is constructed using AMIRA(fig 2), a commercial segmentation and visualization tool, and solved using ABAQUS, a commercial finite element solver.
– Experimental data is compared with forward solutions calculated with the FEM model with increasing levels of complexity.
The final simulation results (like potential fields and electric current density vectors) can be interpreted and compared visually for different model types and simulation locations using appropriate visualization tools. The diffusion tensors obtained using DTI are visualized as ellipsoids and alignment with structural data is verified. Results from the FEM simulations (potential scalar fields and electric current density vector fields) are visualized using either ABAQUS or AMIRA on the head volume geometry(fig 3). In addition, stereoscopic renderings of the head geometry and electric current lines due to a dipole in the human cortex were done in collaboration with Ray Gasser at the Scientific Computing and Visualization group, Boston University.
Using experimental data for validation of forward model, factors affecting the accuracy of the intracranial and scalp measurements are quantified in a precise manner by studying the effects of different tissue types and anisotropy of white matter tissue.