Leveraging Stochastic Resonance for High Spatial-Resolution Extracellular Monitoring

神经系统是一个由数十亿个神经元组成的网络，有几种类型的信号在其中传播。通过监测在网络中传播的电信号，可以探索神经系统的连通性以及信息在神经系统中的表现方式，这可以在广泛的应用领域取得重大进展，从理解神经系统网络在神经或精神疾病中的中断到解码意图。在过去的几十年里，电子设计和数据分析的努力导致了节能、实时和准确的神经监测系统。然而，与理论数量（200）相比，这些系统的每个电极可观察到的神经元数量明显较少（约5到10个），因为(i)错过了远离电极的神经元的小幅度电信号，（ii）活动水平非常低的神经元的信号被高度活跃的神经元的信号掩盖。提出的研究旨在设计一种节能和低复杂性的神经监测系统，解决这些问题，以最大限度地提高每个电极可以检测到的神经元数量，从而提高可扩展性。这项技术可以改变神经科学和神经技术研究中进行细胞外监测的方式。该结果可用于功耗预算有限的其他弱信号检测应用，例如物联网（IoT）和传感器网络应用的传感和通信模块。拟议的跨学科项目将与教育和推广活动相结合，通过参与和培训从K-12到研究生水平的教育劣势学生，在神经监测、弱信号检测和脑机接口方面促进多样性。这个项目解决了植入式脑机接口技术的圣杯之一：如何准确、高效地捕获大脑的神经活动，并在空间和时间上以高分辨率进行分类？为了解决这个问题，该项目将探索和测试以下假设：通过设计多个尖峰探测器，每个探测器通过利用随机共振（SR）来最大限度地提高对不同尖峰幅度和神经元活动水平的灵敏度，这是在包括生物神经元在内的各种系统中观察到的一种弱信号检测现象，应该可以最大限度地提高细胞外监测系统的尖峰检测灵敏度和尖峰分类特异性。为了验证这一假设，该项目将通过三个相互作用的目标进行：(1)展示sr增强的尖峰检测和分类，并设计细胞外监测框架。(2)为单个模块的物理实现设计节能电子设备。(3)进行系统级集成和验证。首先，对最佳噪声强度和探测器规格的理论分析将使用创建的细胞外神经记录的合成数据集进行验证。然后，将执行框架的物理实现。最后，物理系统的系统级集成将遵循与最先进的蜂窝外监测算法的基准比较。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

The nervous system is a network of billions of neurons, where several types of signals propagate through. By monitoring the electrical signals propagating in the network, the connectivity of the nervous system and how information is represented in the nervous system can be explored, which can lead to significant advances in a wide spectrum of applications ranging from understanding the disruptions in the network of the nervous system in the case of neurological or psychiatric conditions to decoding intentions. Efforts in electronics design and data analytics over the past decades have led to energy-efficient, real-time, and accurate neural monitoring systems. However, these systems suffer from significantly smaller number of observable neurons per electrode (~ five to ten) compared to the theoretical number (200) because of (i) missed small-amplitude electrical signals of neurons that are far from the electrode and (ii) signals of neurons with very low activity levels being masked by those of highly-active neurons. The proposed study aims to engineer an energy-efficient and low-complexity neural monitoring system addressing these issues to maximize the number of neurons that can be detected per electrode and therefore improve the scalability. This technology can transform the way extracellular monitoring is performed in neuroscience and neurotechnologies studies. The results can be useful for other weak signal detection applications with limited power budgets such as sensing and communication blocks of Internet of Things (IoT) and sensor-network applications. The proposed interdisciplinary project will be integrated with educational and outreach activities promoting diversity by engaging and training educationally disadvantaged students from K-12 to graduate levels in neural monitoring, weak signal detection, and brain-computer interfaces.This project tackles one of the holy grails of implantable brain-computer interface technologies: How can neural activity of the brain be accurately and energy-efficiently captured and classified at high resolution both spatially and temporally? To address the question, the project will explore and test the following hypothesis: By designing multiple spike detectors each tuned for maximizing the sensitivity to different spike amplitudes and activity-levels of neurons by leveraging stochastic resonance (SR), a phenomenon observed for weak signal detection at various systems including biological neurons, it should be possible to maximize spike detection sensitivity and spike sorting specificity of an extracellular monitoring system. To test the hypothesis, the project will be conducted through three interacting objectives: (1) Demonstrate SR-enhanced spike detection and sorting and design the extracellular monitoring framework. (2) Design energy-efficient electronics for physical implementation of individual blocks. (3) Perform system-level integration and validation. First, theoretical analysis on optimum noise intensity and detector specification will be verified using synthetic datasets of extracellular neural recordings created. Then, physical implementation of the framework will be performed. Lastly, a system-level integration of the physical system will follow benchmark comparisons against the state-of-the-art extracellular monitoring algorithms.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.