Browsing by Subject "Computational Modeling"
Results Per Page
Sort Options
Item Embargo Computational Tools to Improve Stereo-EEG Implantation and Resection Surgery for Patients with Epilepsy(2024) Thio, BrandonApproximately 1 million Americans live with drug-resistant epilepsy. Surgical resection of the brain areas where seizures originate can be curative. However, successful surgical outcomes require delineation of the epileptogenic zone (EZ), the minimum amount of tissue that needs to be resected to eliminate a patient’s seizures. EZ localization is often accomplished using stereo-EEG where 5-30 wires are implanted into the brain through small holes drilled through the skull to map widespread regions of the epileptic network. However, despite the technical advances in surgical planning and epilepsy monitoring, seizure freedom rates following epilepsy surgery have remained at ~60% for decades. In part, seizure freedom rates have not increased because epilepsy neurologists do not have appropriate software tools to optimize stereo-EEG. In this dissertation, we report on the development and analysis of foundational models and software tools to improve the use of stereo-EEG technology and ultimately increase seizure-freedom rates following epilepsy surgery.We developed an automated image-based head-modeling pipeline to generate patient-specific models for stereo-EEG analysis. We assessed the key dipole source model assumption, which assumes that voltages generated by a population of active neurons can be simplified to a single dipole. We found that the dipole source model is appropriate to reproduce the spatial voltage distribution generated by neurons and for source localization applications. Our findings validate a key model parameter for stereo-EEG head-models, which are foundational to all computational tools developed to optimize stereo-EEG. Using the dipole source model, we systematically assessed the origin of recorded brain electrophysiological signals using computational models. We found that, counter to dogma, action potentials contribute appreciably to brain electrophysiological signals. Our findings reshape the cellular interpretation of brain electrophysiological signals and should impact modeling efforts to reproduce neural recordings. We also developed a recording sensitivity metric, which quantifies the cortical areas that are recordable by a set of stereo-EEG electrodes. We used the recording sensitivity metric to develop two software tools to visualize the recording sensitivity on patient-specific brain geometry and to optimize the trajectories of stereo-EEG electrodes. Using the same number of electrodes, our optimization approach identified trajectories that had greater recording sensitivity than clinician-defined trajectories. Using the same target recording sensitivity, our optimization approach found trajectories that mapped the same amount of cortex with fewer electrodes compared to the clinician-defined trajectories. Thus, our optimization approach can improve the outcomes following epilepsy surgery by increasing the chances that an electrode records from the EZ or reduce the risk of surgery by minimizing the number of necessary implanted electrodes. We finally developed a propagating source reconstruction algorithm using a novel TEmporally Dependent Iterative Expansion approach (TEDIE). TEDIE takes as inputs stereo-EEG recordings and patient-specific anatomical images, produces movies of dynamic (moving) neural activity displayed on patient-specific anatomy, and distills the immense intracranial stereo-EEG dataset into an objective reconstruction of the EZ. We validated TEDIE using seizure recordings from 40 patients from two centers. TEDIE consistently localized the EZ closer to the resected regions for patients who are currently seizure-free. Further, TEDIE identified new EZs in 13 of the 23 patients who are currently not seizure-free. Therefore, TEDIE is expected to improve the accuracy of the evaluation of surgical epilepsy candidates, result in increased numbers of patients advancing to surgery, and increase the proportion of patients who achieve seizure freedom through surgery. Together, our suite of software tools constitute important advances to optimize stereo-EEG implantation and analysis, which should lead to more patients achieving seizure freedom following epilepsy surgery.