Advanced Microsystems for Neural Information Processing

描述（由申请人提供）：开发先进的神经假肢系统和脑机接口，用于与神经系统进行高容量、实时、双向通信，是新兴神经工程学科面临的主要挑战。用于记录和刺激多个神经元细胞群的高密度微电极阵列的制造的最新进展引发了许多神经生理学的发现。然而，这些设备的成功在一定程度上受到其当前通信和信号处理能力的影响。高密度神经植入物的实时数据传输需要超高带宽遥测链路。因此，预处理神经信号应尽可能靠近信号获取的位置来实现，以便在数据流的早期推断出“有用”信息，并减少计算和通信成本。该项目旨在开发一种高度可扩展的模块化微系统，能够实时处理神经信号，并在保持最高信号保真度的同时实现大压缩比。该项目有三个目标。第一个目标是设计适当的阵列信号处理算法，以在皮外传输之前推断神经信号中的有用信息。一旦优化，第二个目标是将这些算法嵌入到定制设计的硬件平台上，以预处理神经信号。该模块将被设计用于颅内植入，因此将优化到最小的功耗和形状因素。该模块将作为分布式微系统的前端阶段，旨在连接感兴趣的皮质结构中的多个神经元群体。算法的适应性将在目标3中通过严格的测试进行评估，使用条件离线记录来模拟长期慢性实验中的不利条件。

DESCRIPTION (provided by applicant): The development of advanced neuroprosthetic systems and brain-machine interfaces for high-capacity, real-time, 2-way communication with the nervous system is a major challenge to the emerging neural engineering discipline. Recent advances in the fabrication of high-density microelectrode arrays for recording and stimulation of multiple neuronal cell populations have triggered numerous neurophysiological discoveries. Nevertheless, the success of these devices is mitigated in part by their current communication and signal processing capabilities. Data transmission in real-time from a high-density neural implant would require an ultra-high bandwidth telemetry link. Pre-processing neural signals is accordingly sought to be implemented as close as possible to where the signal is acquired to infer the "useful" information early in the data stream and to reduce the computational and communication costs. The proposed project is aimed at developing a highly scalable modular microsystem capable of processing neural signals in real-time and achieving large compression ratios while maintaining highest signal fidelity. The project has 3 aims. The first aim is to design adequate array signal processing algorithms to infer the useful information in the neural signals prior to extra-cutaneous transmission. Once optimized, the second aim is to embed these algorithms onto a custom designed hardware platform for the purpose of preconditioning the neural signals. The module will be designed for intra-cranial implantation, thus will be optimized for minimum power dissipation and form factor. This module will serve as the front-end stage of a distributed microsystem aimed at interfacing multiple neuronal populations in cortical structures of interest. Adaptation of the algorithms will be assessed in aim 3 with rigorous testing using conditioned offline recordings to mimic adverse conditions in long term chronic experiments.