Perturbative Seizure Prediction and Detection of a Seizure Permissive State

描述（由申请人提供）：在200万美国癫痫患者中，尽管接受了治疗，但仍有近30%的人癫痫发作。现在越来越多的人接受刺激装置作为治疗神经调节的一种手段。为了提供复杂的反馈刺激，人们可以在癫痫发作的早期做出反应，以最小化其影响和传播，或者可以在癫痫发作之前对其他状态做出反应。在相当长的一段时间里，确定合适的癫痫发作前状态一直是一个中心主题，现在属于“癫痫发作预测”的范畴。这种癫痫发作前状态的识别可以指示何时刺激以避免即将到来的癫痫发作。但是，在癫痫发作预测界的经验是，迄今为止使用的所有措施都产生了显着的错误检测率。这项工作将在破伤风毒素颞叶癫痫模型中进行。我们已经开发了一种系统，用于应用低频电刺激来调节神经元活动，而不会干扰我们记录慢性植入动物神经活动的能力（Sunderam等人，2006年），因此可以应用反馈刺激。通过头部加速度测量，我们能够确定警戒状态（Sunderam, et al, 2007）。这项赠款的目的有三个方面。首先，研究警惕性状态作为一种鉴别特征的加入是否能提高对癫痫发作前状态的识别。第二，通过小幅度刺激实现并测试大脑状态的主动探针，以检测预示癫痫发作前状态的大脑状态变化。我们期望在实施被动和主动预测方法后，我们仍然会观察到显著的错误预测率。第三个目标是通过刺激来探索这些检测的本质(a)错误预测是简单的错误分类还是(b)识别出的癫痫发作前状态是允许癫痫发作的——一种偶尔过渡到癫痫发作的状态——以及“错误预测”是对这种状态的正确识别。从基础科学的角度来看，这应该能深入了解癫痫发作的产生过程和长期治疗。从更实际的短期应用的角度来看，检测癫痫发作允许状态和相关的过渡概率将在开发有用的反馈干预方面具有很大的效用。具体来说，一个目标干预-例如电刺激-响应于状态的检测来修改这种转换概率。在未来几年，这项工作的扩展将是测试一系列对癫痫发作前检测的反应性刺激，以了解它们预防癫痫发作的能力。公共卫生相关性：本资助的长期目标是改善癫痫发作控制的神经刺激。通过解决癫痫发作前状态的性质，更重要的是错误癫痫发作预测的性质，我们将提高优化反馈刺激和制作微创刺激的能力。

DESCRIPTION (provided by applicant): Nearly 30% of the two million Americans suffering from epilepsy continue to have seizures despite treatment. There is now a growing acceptance of stimulation devices as a mean for therapeutic neuromodulation. To provide sophisticated feedback stimulation, one can either respond early into the seizure in order to minimize its impact and spread, or one can respond before the seizure to some other state. Identification of a suitable preseizure state has been a central theme for quite a while that now falls under the rubric of "seizure prediction." Identification of such a preseizure state could both indicate when to stimulate in order to avert an oncoming seizure. But, the experience in the seizure prediction community has been that all measures used so far yield a significant false detection rate for significant levels of sensitivity. This work will be performed in the tetanus toxin model for temporal lobe epilepsy. We have developed a system for applying low frequency electrical stimulation for modulation of neuronal activity without interfering with our ability to record neural activity in chronically implanted animals (Sunderam, et al, 2006), and therefore can apply feedback stimulation. With head acceleration measurements, we are able to determine state of vigilance (Sunderam, et al, 2007). The aims of this grant are three fold. First, to investigate if the addition of state of vigilance as a discrimination feature improves identification of preseizure states. Second, to implement and test an active probe of brain state through small amplitude stimulations to detect changes in brain state indicative of a preseizure state. We expect after implementing both the passive and active prediction methods that we will still observe significant false prediction rates. The third aim is to probe through stimulation the nature of these detections (a) if the false predictions are simply misclassifications OR (b) if the identified preseizure state is seizure permissive - a state that only sometimes transitions to seizure - and the 'false predictions' are correct identifications of this state. From a basic science standpoint, this should give insight into the seizure generation process and long-term treatment. From a more practical short-term application standpoint, detection of a seizure permissive state and the relevant transition probabilities will have great utility in the development of a useful feedback intervention. Specifically, one then targets intervention - for example electrical stimulation - in response to detection of the state to modify this transition probability. The extension of this work in future years will be to test a range of responsive stimuli to preseizure detections for their ability to prevent seizure. PUBLIC HEALTH RELEVANCE: The long term objectives of this grant are to improve neurostimulation for seizure control. By addressing the nature of the preseizure state and more importantly the nature of false seizure predictions, we will improve the ability to optimize feedback stimulation and to craft minimally invasive stimuli.