Accurate and flexible personalized head modeling and its validation using MR current density imaging

Personalized electric field calculations are the central element of strategies to characterize and optimize the physical dose of transcranial brain stimulation in the individual brain. The calculations require head models, which are usually automatically created from structural MR images. The accuracy of this procedure directly links to the accuracy of the estimated dose, its robustness affects how easily dose simulations can be embedded in clinical trials and its flexibility determines whether dose simulations can used in patient populations with altered brain anatomy.
We present a new pipeline that combines flexible and robust image segmentation and meshing methods to generate highly detailed tetrahedral head models for Finite-Element-based simulations. It robustly works for MR data from different scanners and segments >50 head and brain regions that are combined into nine tissues types for the simulations. The meshing achieves high-quality representations of the tissue regions and boundaries while at the same time offering flexibility to easily include information from specialized segmentation tools, e.g. for lesion delineation. The new pipeline will strongly facilitate personalized field simulations in particular in clinical research settings.
It is a major challenge to ensure that increasing levels of anatomical detail in the simulations indeed translate into a better accuracy of the estimated fields. MR current density imaging (MRCDI) is a new method for the non-invasive characterization of the current flow generated by transcranial electric stimulation (TES), based on measurements of their induced magnetic fields. We present new MRCDI acquisition and reconstruction approaches that enable stable assessments of individual current flow patterns. Interestingly, even for our state-of-the-art head models, we can demonstrate consistent differences in the accuracy of the simulated fields, depending on the electrode montage. We also explore ways to improve simulation accuracy based on MRCDI data, serving as further proof-of-principle of the usefulness of the new approach.