Data-driven optimization for DBS programming in temporal lobe epilepsy

脑刺激疗法是一种改变神经和精神疾病患者生活的治疗方法， 包括帕金森氏症、抑郁症和癫痫。在这种治疗中，神经外科医生植入电极 在大脑内部，可以提供广泛的电刺激模式。然而，下一个问题是， 确定给定患者的最佳刺激设置。考虑到即使是基本的临床刺激设备 可以配置成数百万种不同的刺激设置，找到合适的刺激设置是一项挑战。这个问题 在癫痫中会加重，患者在癫痫发作之间不会表现出症状， 评估一个场景的效果就更加困难了。在其核心，这是一个优化问题，许多 存在工程解决方案。我们已经开发了一个框架，用于设计神经网络的优化系统。 这是一种可以应用于广泛的不同神经调制范例的神经调制。在这一提议中， 我们开发了一个优化系统，用于自动和有效地识别最佳刺激设置， 最大限度地抑制癫痫啮齿动物模型中的癫痫发作。我们之前的研究表明， 的刺激，异步分布刺激，可以减少大鼠破伤风毒素发作的频率 颞叶癫痫模型然而，仅对有限的刺激模式集进行了评价。在目标1中， 我们将在这项工作的基础上，更好地描述不同异步分布式 刺激参数对癫痫发作的影响这些实验有两个目的。首先，他们将澄清如何 不同的受试者受到刺激参数的影响，并确定是否如在其他神经系统疾病中那样 通过脑刺激治疗，最佳刺激设置将因受试者而异。第二个目的是 使用收集到的数据来创建用于原型优化系统的仿真平台。在目标2中，我们将 使用仿真平台原型并调整不同的优化系统。在确定哪些 优化系统在我们的仿真平台中表现最好，优化系统将被实现， 实时体内优化，以了解最佳降低癫痫发作频率的受试者特定刺激设置。

Brain stimulation therapy is a life changing treatment for patients with neurological and psychiatric disorders, including Parkinson’s disease, depression, and epilepsy. In this treatment, neurosurgeons implant electrodes inside the brain that can deliver a wide range of electrical stimulation patterns. However, the next problem is determining the optimal stimulation setting for a given patient. Given that even basic clinical stimulation devices can be configured to millions of different stimulation settings, finding the right one is challenging. This problem is exacerbated in epilepsy, where patients do not exhibit symptoms between seizures, making the process of evaluating a setting’s effectiveness even more difficult. At its core, this is an optimization problem for which many engineering solutions exist. We have developed a framework for designing optimization systems for neural modulation that can be applied to a broad spectrum of different neural modulation paradigms. In this proposal, we develop an optimization system for automatically and efficiently identifying the optimal stimulation setting to maximally suppress seizures in a rodent model of epilepsy. Our previous work has shown that a particular type of stimulation, asynchronous distributed stimulation, can reduce the frequency of seizures in the rat tetanus toxin model of temporal lobe epilepsy. However, only a limited set of stimulation patterns were evaluated. In Aim 1, we will build on this work to better characterize the differential effects of varying asynchronous distributed stimulation parameters on seizures. These experiments will serve two purposes. First, they will clarify how different subjects are affected by stimulation parameters and determine if, as in other neurological disorders treated by brain stimulation, the best stimulation setting will vary from subject to subject. The second purpose is to use the data collected to create a simulation platform for prototyping optimization systems. In Aim 2, we will use the simulation platform prototype and tune different optimization systems. After determining which optimization system performs best in our simulation platform, the optimization systems will be implemented for real-time in vivo optimization to learn the subject specific stimulation settings that best reduce seizure frequency.