Model Development for Prediction of Surgical Outcome in Temporal Lobe Epilepsy Patients: Incorporation of the Correlation between Post-Surgical Reorganization Phenotypes and Pre-Surgical Data

项目摘要 对于接受脑切除或消融干预的癫痫患者， 决定癫痫发作状态，无论是控制还是复发。然而，正是来自术前大脑的数据驱动了 术后预测过程-对于患者和医生来说都是一个关键的过程， 当癫痫发作结果被预测为最佳的药物决策时，这是有意义的。 因此，我们建议开发一个多步骤模型，建立更准确的预测后， 颞叶癫痫（TLE）的手术癫痫结局强调手术后状态，因为它是 在手术过程中保留下来的大脑形成了术后癫痫发作的神经基质。第二 激发我们项目的一个观点是，需要识别功能和结构脑网络中的这些变化 该组织支持脑外科手术后适应性与适应不良性癫痫发作的结果。这些都是 网络改变（例如，新的癫痫发作发生器），将潜在的外科手术候选人置于 具体的结果轨迹。因此，识别大脑重组和变化的表型， 将它们的状态纳入手术前预测结果的模型中，可能会证明对提高我们的 预测术后神经成形反应的能力。虽然现有的TLE结局预测模型 关注临床变量（例如，病变状态），我们选择专注于结构和功能 网络重组的措施（通信动态，区域互动，结构控制）。这 源于我们的信念，即捕捉整个手术后大脑的网络变化， 识别和预测潜在癫痫灶（癫痫发生）的实用方法， 手术通过机器学习技术，我们将提供一种算法，用于新的、潜在的外科手术。 患者，一种仅利用术前数据的算法，但结合了我们关于 术后脑组织因此，我们的方法提供了一个方法和概念 （重组表型）进展。导致我们假设的科学前提是， 文献来解释未切除/消融的脑区域的影响，并且大脑重组这些区域。 领域的强制性，严重阻碍了以前的结果模型的预测能力。

Project Summary For epileptic patients who undergo brain resection or ablation interventions, it is the postoperative brain that will dictate seizure status, whether controlled or relapsed. Yet, it is data from the preoperative brain that drives the postoperative prediction process – a critical process for both patient and doctor, and one that is only clinically meaningful when seizure outcomes are predicted presurgically to optimize surgical-decision making. Accordingly, we propose to develop a multi-step model that will establish more accurate predictors of post- surgical seizure outcome in temporal lobe epilepsy (TLE) emphasizing post-surgical status, for it is the areas of the brain spared during surgery that form the neural substrates generating postoperative seizures. A second perspective motivating our project is the need to identify those changes in functional and structural brain network organization that support adaptive versus maladaptive seizure outcomes following brain surgery. These are the network changes (e.g., the new seizure generators) that dispose and place a potential surgical candidate on a specific outcome trajectory. Therefore, identifying the phenotypes of brain reorganization and change, and incorporating their status into presurgical predictive models of outcome will likely prove crucial to enhancing our ability to predict postoperative neuroplastic responses. While existing outcome prediction models in TLE have focused on clinical variables (e.g., lesional status), we choose instead to focus on structural and functional measures of network reorganization (communication dynamics, regional interactions, structural control). This stems from our belief that capturing network changes throughout the whole postsurgical brain offers a better practical method for identifying and predicting the latent seizure foci (epileptogenesis) that will emerge after surgery. Through machine learning techniques we will deliver an algorithm to be used with new, potential surgical patients, an algorithm that utilizes solely presurgical data, but incorporates our innovative prediction about postsurgical brain organization. Accordingly, our approach provides both a methodologic and conceptual (reorganization phenotypes) advance. The scientific premise leading to our hypotheses is that the failure in the literature to account for the impact of unresected/ablated brain regions, and the brain reorganizations these areas compel, has seriously impeded the predictive power of previous outcome models.