Computational Tools for Improving Stereo-EEG Implantation and Resection Surgery

在340万美国癫痫患者中，超过30%的人没有从药物治疗中获益。手术 切除癫痫发作起源的脑组织（致痫区）是一种替代方法， 这些患者的1年无癫痫发作率为61%。然而，成功的切除需要 致痫区的定位，这是具有挑战性的，因为致痫区是不可区分的 临床图像（例如MRI）上的健康组织。立体脑电（sEEG）是一种微创记录技术 其中100-200个电极触点通过小的经颅钻孔插入到 大脑与活跃的神经元相距数百微米，从而产生比 常规EEG。尽管sEEG具有巨大的潜力，但癫痫切除手术的结果还没有 在过去的20年里有了很大的改善。这部分是由于对最佳选择的不完全理解。 sEEG植入策略和缺乏算法来利用sEEG的时空分辨率， 致痫灶定位该提案的目标是开发和部署临床上有用的计算 使用sEEG改善致痫区定位的工具。 第一个目标是开发一套计算工具，将可以记录的脑组织可视化 通过一组sEEG电极和优化电极轨迹的优化算法， 植入电极的数量，同时最大化皮层覆盖。我将开发出现实的计算 使用患者特定头部有限元建模的头部模型。我会把头部模型和一个 估计神经源强度，以估计和可视化可以由一组sEEG记录的组织 个电极然后，我将耦合头部模型和源强度估计到蒙特卡洛树搜索 该算法用于确定标测感兴趣区域所需的最小电极轨迹集。的 结果将是一对计算工具，其可视化可记录的脑组织并优化电极 植入轨迹 第二个目标是开发一种时空源重建算法来映射神经记录 进入大脑我将开发一个贝叶斯源重建算法，估计时间过程， 神经活动的空间范围。我将使用源重建算法记录癫痫样活动 来描绘大脑中与致痫区相关的区域。结果将是贝叶斯 源重建工具，癫痫病学家可以使用辅助外科切除决策。 这项工作的成功完成有望改善致痫区定位， 癫痫患者的无癫痫发作率更高。

More than 30% of the 3.4 million Americans with epilepsy do not benefit from drug therapies. Surgical removal of the brain tissue where seizures originate (the epileptogenic zone) is an alternative that can eliminate seizures in these patients with 1 year seizure freedom rates of 61%. However, successful resection requires localization of the epileptogenic zone, which is challenging because the epileptogenic zone is indistinguishable from healthy tissue on clinical images (e.g. MRI). Stereo-EEG (sEEG) is a minimally invasive recording technique where 100-200 electrode contacts are inserted through small transcranial burr holes into widespread regions of the brain hundreds of microns from active neurons, resulting in substantially higher signal fidelity than conventional EEG. Despite the enormous potential of sEEG, outcomes of seizure resection surgeries have not improved substantially over the past 20 years. This is due, in part, to an incomplete understanding of an optimal sEEG implantation strategy and a lack of algorithms to exploit the spatiotemporal resolution of sEEG for epileptogenic zone localization. The goal of this proposal is to develop and deploy clinically useful computational tools to improve epileptogenic zone localization using sEEG. The first aim is to develop a set of computational tools that visualize the brain tissue that can be recorded by a set of sEEG electrodes and an optimization algorithm that optimizes the electrode trajectories to minimize the number of implanted electrodes while maximizing cortical coverage. I will develop realistic computational head models using patient specific head finite element modeling. I will couple the head models to an established estimate of the neural source strength to estimate and visualize the tissue that can be recorded by a set of sEEG electrodes. I will then couple the head models and source strength estimate to a Monte Carlo Tree Search algorithm to determine the minimum set of electrode trajectories necessary to map a region of interest. The outcome will be a pair of computational tools that visualize the recordable brain tissue and optimize electrode implantation trajectories. The second aim is to develop a spatiotemporal source reconstruction algorithm to map neural recordings into the brain. I will develop a Bayesian source reconstruction algorithm that estimates the time courses and spatial extent of neural activity. I will use the source reconstruction algorithm on recordings of epileptiform activity to delineate zone of the brain that are associated with the epileptogenic zone. The outcome will be a Bayesian source reconstruction tool that that epileptologists can use aid surgical resection decision making. Successful completion of this work is expected to improve epileptogenic zone localization and result in higher seizure freedom rates in patients with epilepsy.