EFRI BRAID: Optical Neural Co-Processors for Predictive and Adaptive Brain Restoration and Augmentation

神经系统疾病，如创伤性脑损伤、中风或脑瘫，是世界范围内残疾和死亡的重要原因。世界人口中近六分之一的人患有这些疾病。然而，可用于这些疾病的非常有限的治疗仅提供适度的治疗益处，并且通常与严重的副作用相关。受大脑启发的植入式计算设备可以为康复和治疗这些疾病提供解决方案。这种设备可以通过记录来自神经系统的电信号，处理它们并实时刺激大脑的另一部分来运行。这使得受伤或受损的大脑区域被绕过或康复。然而，现有的大脑启发式计算设备消耗太多的功率并且不够快以提供这种实时反馈和控制。该项目旨在通过两个方面的创新来创建“大脑协处理器”：首先，基于从大脑收集的神经信号创建新的算法，以提供更高的准确性;其次，通过使用光学硬件，不仅可以高速低功耗地处理信息，还可以通过利用大脑中的光控蛋白质直接与大脑接口。此外，该项目旨在改善本科生和高中生的培训和教育，特别强调将妇女和代表性不足的少数民族社区纳入光学，机器学习和神经科学的多学科研究。科学成果将通过研讨会、讲习班、同行评审的出版物，了解大脑如何工作，并利用这些知识来恢复或增强大脑功能，需要超快的并行算法，这些算法比当前最先进的算法先进几个数量级。该研究项目将建立“光学神经协处理器”。它使用光作为计算资源，并利用大脑启发的编码器-解码器循环神经网络，以大脑的多个自然时间尺度与大脑交互。结合理论神经科学，神经启发机器学习，光遗传学，神经康复，纳米光子学和集成半导体光学的专业知识，该研究项目将开发用于神经接口和协同处理的脑启发预测编码人工神经网络;设计和制造利用新兴半导体纳米光子学和集成光子学的光学神经架构;以及展示光学神经协处理器，它与非人类灵长类动物的大脑实时连接，用于康复。沿着技术的进步，神经伦理学的影响将在这个项目中进行调查。这个奖项反映了NSF的法定使命，并已被认为是值得通过使用基金会的智力价值和更广泛的影响审查标准进行评估的支持。

Neurological disorders, such as traumatic brain injury, stroke, or cerebral palsy, are an important cause of disability and death worldwide. Nearly one in six of the world’s population experience these disorders. However, the very limited treatments available for these disorders provide only modest therapeutic benefits and are often associated with serious side effects. Brain-inspired, implanted computing devices could provide a solution for rehabilitating and curing these disorders. Such devices can operate by recording electrical signals from the nervous system, processing them, and stimulating another part of the brain in real-time. This allows the injured or impaired area of the brain to be bypassed or rehabilitated. However, existing brain-inspired computing devices consume too much power and are not fast enough to provide such real-time feedback and control. This project aims to create a “brain co-processor” by innovating in two aspects: first, create new algorithms based on neural signals collected from the brain to provide higher accuracy; and second, by employing optical hardware that not only can process information with high speed and low power, but also directly interfaces with the brain by exploiting light-controlled proteins in the brain. Furthermore, this project aims to improve the training and education of undergraduate and high school students, with a strong emphasis on including women and underrepresented minority communities, in multi-disciplinary research on optics, machine learning, and neuroscience. The scientific results will be disseminated to a wide scientific audience via seminars, workshops, peer-reviewed publications, and conferences.Understanding how the brain works and using that knowledge to restore or augment brain function require ultrafast parallel algorithms that are orders-of-magnitude more advanced than current state-of-the-art. This research project will build “optical neural co-processors” that use light as a computational resource and leverage brain-inspired encoder-decoder recurrent neural networks to interact with the brain in multiple natural timescales of the brain. Combining expertise in theoretical neuroscience, neuro-inspired machine learning, optogenetics, neuro-rehabilitation, nanophotonics and integrated semiconductor optics, this research project will develop brain-inspired predictive coding artificial neural networks for neural interfacing and co-processing; design and fabricate optical neural architectures that exploit emerging semiconductor nanophotonics and integrated photonics; as well as demonstrate optical neural co-processors that interface with the brain in real-time for rehabilitation in non-human primates. Along with technical advancements, neuro-ethical implications of the developed technologies will be investigated in this project.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.