NCS-FO: Super resolution Mapping of Multi-scale Neuronal circuits Using Flexible Transparent Arrays

了解作为感知、认知和行动基础的大脑的结构和功能组件，对于开发下一代神经假体、脑机接口和发现预防神经疾病的措施至关重要。光学技术使我们能够以单细胞分辨率记录和推断神经活动。然而，它们受到低时间分辨率的限制，并且往往无法准确地捕捉毫秒时间尺度上的神经动力学。另一方面，电生理学提供了更高的时间分辨率，但单细胞电生理学通常存在吞吐量低的问题，覆盖更大空间尺度的记录存在空间分辨率低的问题，使得从大区域破译细胞尺度的神经活动变得困难。认识到微观光学成像和宏观电生理记录在空间和时间分辨率方面具有互补性，这一多学科项目将利用神经工程、多模式成像和信号处理方面的创新将这两种记录方式结合起来，以了解以前未达到的时间和空间分辨率下的神经活动。这种能力将导致在大脑信息处理和神经系统疾病(癫痫、抑郁症、记忆障碍等)的电路功能障碍方面的新发现，影响全球10亿人。用增强的分辨率记录和解析神经活动可以推动下一代脑机接口的发展，以恢复视觉、听力和运动。该项目的成果还将被纳入编写跨学科教育材料，以培训下一代神经工程师、神经科学家和信号处理专家。该项目由理解神经和认知系统的综合策略(NSF-NCS)资助，NSF-NCS是由计算机和信息科学与工程(CEISE)、教育和人力资源(EHR)、工程(ENG)和社会、行为和经济科学(SBE)的主管部门联合支持的一个多学科项目。该项目有三个主要技术部分，包括开发新型电极阵列，仔细设计多模式成像实验，以及解决不适定反问题的高级信号处理技术。由新型电极阵列实现的同步多光子成像和电生理实验将生成全新的数据集，这些数据集将由新的数据驱动的超分辨率算法处理，该算法明智地利用两种成像模式的互补优势。其核心思想是将融合问题置于双线性问题的数学框架内，通过融合从光和电生理记录获得的数据集，利用潜在神经活动的稀疏性作为解决逆问题的关键因素。用于通过融合具有互补属性的信号来生成超分辨率图像的数学原理和算法具有比神经成像更广泛的适用性，并且可以用于为在不同成像应用中出现的不适定逆问题开发更有效的解决方案。

Understanding the structural and functional components of the brain that underlie perception, cognition and action, is crucial for developing next generation neural prostheses, brain machine interfaces, and discovering preventive measures against neurological disorders. Optical technologies have enabled us to record and infer neural activity with single-cell resolution. However, they are limited by low temporal resolution, and often fail to accurately capture the neural dynamics at the milli-second time scales. Electrophysiology, on the other hand, provides higher temporal resolution, but single-cell electrophysiology usually suffers from low throughput, and recordings that cover larger spatial scales suffer from poor spatial resolution, making it difficult to decipher neural activity at cellular scale from large areas. Realizing that micro-scale optical imaging and macro-scale electrophysiological recording possess complementary strengths in terms of spatial and temporal resolution, this multidisciplinary project will combine the two recording modalities using innovations in neural engineering, multi-modal imaging and signal processing, to understand neural activity at previously unattained temporal and spatial resolution. Such a capability will lead to new discoveries on information processing in the brain and circuit dysfunctions for neurological disorders (epilepsy, depression, memory disorders, etc.), affecting one billion people worldwide. Recording and resolving neural activity with enhanced resolution can drive the development of next-generation of brain computer interfaces for restoring vision, hearing, and movement. The outcomes of this project will also be integrated into developing interdisciplinary educational materials for training the next generation of neuroengineers, neuroscientists and signal processing experts. This project is funded by Integrative Strategies for Understanding Neural and Cognitive Systems (NSF-NCS), a multidisciplinary program jointly supported by the Directorates for Computer and Information Science and Engineering (CISE), Education and Human Resources (EHR), Engineering (ENG), and Social, Behavioral, and Economic Sciences (SBE).The project has three main technical components that consist of development of novel electrode arrays, careful design of multi-modal imaging experiments, and advanced signal processing techniques for solving ill-posed inverse problems. Simultaneous multiphoton imaging and electrophysiology experiments enabled by novel electrode arrays will generate brand new datasets which will be processed by new data-driven super-resolution algorithms that judiciously exploit the complementary strengths of the two imaging modalities. The key idea is to cast the fusion problem within the mathematical framework of bilinear problems, and exploit sparsity of the underlying neural activity as a key ingredient in solving the inverse problem by fusing the datasets obtained from optical and electrophysiological recordings. The mathematical principles and algorithms used for creating super-resolution images by fusing signals with complementary attributes have broader applicability beyond neural imaging, and can be used for developing more efficient solutions for ill-posed inverse problems that arise in diverse imaging applications.