SCH: EXP: Collaborative Research: Exploring Sparsity and Spectral-Temporal Decomposition in Real-Time Network Modulation for Intractable Epilepsy

理解大脑活动和人类行为之间的关系不仅是我们这一代最重要的科学挑战之一，也是医学和公共卫生领域最重要的挑战之一。该项目开发的新技术可以解决神经元的微小尺寸，以及神经活动产生的大量数据。该项目利用莱斯大学和德克萨斯医学中心之间的合作环境，开发新的电刺激方法来自适应和选择性地调节癫痫发作网络。如果成功，最终结果将是一种利用固有大脑可塑性机制的修复疗法，有一天可能会独立于长期植入的电子设备。该项目开发了一种算法，通过使用皮质电图（ECoG）对大脑进行实时监测，捕捉大脑的动态、频率依赖性连接，然后确定低频电刺激（LFS）的“最佳”参数，以时间和空间精度调节癫痫网络的连接。首先将神经活动分割到不同的时代和光谱波段，然后在每个片段中导出稀疏连通性，从而控制实时建模的复杂性。使用格兰杰因果关系估计每个频谱时间段的有效连通性。LFS是在从模型中识别的空间位置检测间歇癫痫样放电（ied）后应用的。这些关键步骤导致了实时刺激原型系统的开发，在复杂性与准确性之间进行了自然的权衡，从而在电池寿命和效率之间做出了妥协。空间优化、活动触发的LFS的效果是通过测量癫痫发作网络的易怒性和比较治疗前后检测到的ied的比率来评估的。这些实验将指出治疗药理学上难治性癫痫的方法，而不需要手术切除脑组织，并导致利用大脑固有可塑性的修复疗法。提出的方法提出了其同类修复，实时和选择性网络调制治疗衰弱性疾病的第一个。

Understanding the relationship between brain activity and human behavior is not only one of the most important scientific challenges of our generation but also one of the most important challenges in medicine and public health. This project develops new technology that can address the minute size of the neurons, and the vast amount of data generated by neural activity. This project leverages the collaborative environment between Rice and Texas Medical Center to develop novel electrical stimulation approaches to modulate the seizure network, adaptively and selectively. If successful, the end result would be a reparative therapy that leverages inherent brain plasticity mechanisms and may one day be independent of chronically implanted electronics.This project develops algorithms that capture the dynamic, frequency dependent connectivity of the brain from real-time monitoring of the brain using ECoG (Electrocorticography) and then identifying the "optimal" parameters of the LFS (low-frequency electrical stimulation) to modulate the connectivity of the epilepsy network with temporal and spatial precision. The complexity of modeling such connectivity in real-time is managed by first segmenting neural activity into different epochs and spectral bands and then deriving the sparse connectivity in each of the segments. Effective connectivity in each spectral-temporal segment is estimated using Granger causality. LFS is applied after detecting interictal epileptiform discharges (IEDs) at spatial locations identified from the model. These critical steps lead to the development of a prototype system of real-time stimulation with a natural trade-off of complexity versus accuracy prompting a compromise between battery life and efficacy. The efficacy of spatially-optimized, activity-triggered LFS is evaluated by measuring the irritability of the seizure network and comparing the rate of IEDs detected during pre- and post-treatment periods. These experiments would point the way to treatment of pharmacologically refractory epilepsy without surgical resection of brain tissue and lead to reparative therapies leveraging inherent brain plasticity. The proposed methodology presents the first of its kind reparative, real-time, and selective network modulation to treat a debilitating disease.