Integrated Circuits for Next-Generation Neural Implants

大约有360万加拿大人受到神经系统疾病的影响，如癫痫和帕金森病。神经植入物是新兴的医疗设备，其可以通过响应于实时检测神经信号中的不规则生物标志物（例如，癫痫发作）。神经植入物可以达到比传统药物更上级的临床疗效，特别是在治疗顽固性症状方面。集成电路（IC）已被开发用于将神经植入物集成到微型芯片中，这可以大大提高性能并减小设备尺寸（从而减少手术侵入性）。 虽然近年来用于神经植入物的IC设计已经显著进步，但是存在三个主要限制。首先，大多数设计中的神经接口仅限于电信号，缺少化学信号中不规则神经状态的关键生物标志物（例如，多巴胺水平）。其次，神经信号处理算法通常在IC集成期间被过度简化，从而在检测用于刺激的生物标志物方面表现不佳。虽然机器学习算法在离线研究中显示出优势，但由于缺乏节能的IC实现，它们没有被采用。第三，植入不同大脑区域的电极通常通过长导线连接到集中的IC，导致信号污染和导线移位的风险。 我们的长期目标是开发用于下一代神经植入物的IC，其神经接口模式（从单独的电信号到多模态），信号处理能力（从粗略检测到机器学习）和系统拓扑结构（从集中式到分布式）的范式转变。具体而言，我们有三个短期目标：（1）开发新型多模态神经接口IC，以捕获全面的神经活动，以改善诊断;（2）开发节能IC，以支持通过机器学习进行高性能生物标志物检测;（3）开发低功耗，短延迟无线网络，以实现分布式神经植入系统，消除长导线。为了实现这些目标，提出了具有根本创新的IC设计。 这项研究计划将通过新颖的IC设计来提高神经植入物的性能、安全性和可靠性。这些IC可以低成本制造，并使治疗能够使大量患者受益，特别是那些由于收入和医疗保健不平等而处于不利地位的患者。先进的集成电路设计技术也可以应用于广泛的应用领域，为创业和技术转移提供了巨大的机会。此外，该计划将为HQP提供多学科培训，包括神经接口，机器学习和无线通信的IC设计。

About 3.6 million Canadians are affected by neurological conditions, such as epilepsy and Parkinson's disease. Neural implants are emerging medical devices that can treat neurological conditions by providing therapeutic stimulations (similar to a heart pacemaker) to a patient in response to the real-time detection of irregular biomarkers in the neural signals (e.g., seizures). Neural implants can achieve superior clinical efficacy over conventional medications, especially in treating intractable symptoms. Integrated circuits (ICs) have been developed for integrating neural implants into miniature chips, which can greatly improve performance and reduce device size (thus reducing surgical invasiveness). Although IC designs for neural implants have advanced significantly in recent years, three major limitations exist. Firstly, neural interfacing in most designs is limited to electrical signals alone, missing key biomarkers of irregular neural states in chemical signals (e.g., dopamine levels). Secondly, neural signal processing algorithms are often oversimplified during IC integration, rendering poor performance in detecting biomarkers for stimulation. Although machine learning algorithms have shown advantages in offline studies, they are not adopted due to the lack of energy-efficient IC implementation. Thirdly, electrodes implanted in different brain regions are often connected to a centralized IC through long wires, causing signal contamination and risks of wire displacement. Our long-term goal is to develop ICs for next-generation neural implants with paradigm shifts in neural interfacing modalities (from electrical signal alone to multi-modal), signal processing capabilities (from coarse detection to machine learning), and system topology (from centralized to distributed). Specifically, we have three short-term objectives: (1) develop novel multi-modal neural interfacing ICs to capture comprehensive neural activities for improved diagnosis; (2) develop energy-efficient ICs to support high performance biomarker detection by machine learning; (3) develop a low-power, short-latency wireless network to enable a distributed neural implant system eliminating long wires. IC designs with fundamental innovations are proposed to achieve these goals. This research program will advance the performance, safety, and reliability of neural implants by novel IC designs. These ICs can be fabricated at low costs and enable treatments that can benefit a large patient population, especially those disadvantaged due to income and healthcare inequalities. The advanced IC design techniques can also be applied in a wide range of applications, providing great opportunities for entrepreneurship and technology transfer. Furthermore, this program will provide HQP with multi-disciplinary training, including IC designs for neural interfacing, machine learning, and wireless communication.