FuSe-TG: Cross-layer Co-Design for Self-Evolving Implantable Devices

FuSe-TG：自我进化植入设备的跨层协同设计

• 批准号: 2235364
• 负责人: Yiyu Shi
• 金额: $ 30万
• 依托单位: University of Notre Dame
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2023
• 资助国家: 美国
• 起止时间: 2023-05-01 至 2025-04-30
• 项目状态: 未结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=2235364&HistoricalAwards=false
• 关键词: FuSe; TG; Cross; layer; Co

Medical implantable devices have been widely adopted in many different parts of the body for various applications in the past six decades. They are used to self-manage diseases before or after diseases occur based on the sensed physiological signals. However, the temporal patterns of the physiological signals vary greatly for different individuals (i.e., inter-patient variability). A hardened detection algorithm will not be adequate for all patients to accurately detect the targeting disease (or events) with such variability. The current implantable devices, once implanted, only allow the parameter values of the detection algorithms to be iteratively tuned by doctors. To address the challenges, this project aims to develop the next generation implantable devices that are self-evolvable, allowing dynamic, continuous, and automatic adjustment of both the detection algorithms and the hardware configurations based on individual patient’s circumstances (e.g., data features, ambient environments, device conditions) throughout the devices’ lifetime. A self-evolvable design paradigm with software-hardware co-exploration will be studied to enable low-power, low-overhead, and on-device model personalization and inference for implantable devices.The success of this project will lead to higher detection accuracy, longer implantable device lifetime, and higher quality of care for patients with implantable devices. More importantly, this project will develop a novel cross-layer co-exploration design paradigm for different types of implantable devices to conduct personalized disease management. The education impacts of the proposed research include the integration of various educational activities based on the resources available to the PIs such as DAC System Design Contest and TinyML Design Contest at ICCAD; outreach for local K-12 students through ND’s CS curriculum for K-12 students in Indiana and an in-person webinar named KidSemicon; an online social forum that allows students, researchers, and engineers to communicate, learn, and share knowledge in the domain of reconfigurable semiconductor technologies; undergraduate research with emphasis on minority participation, and course integration of the research outcomes.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.

在过去的六十年里，医疗植入式设备已被广泛用于身体的许多不同部位的各种应用。它们用于基于感测到的生理信号在疾病发生之前或之后自我管理疾病。然而，生理信号的时间模式对于不同的个体变化很大（即，患者间变异性）。硬化的检测算法不足以让所有患者准确检测具有这种变异性的靶向疾病（或事件）。当前的可植入设备一旦被植入，仅允许检测算法的参数值由医生迭代地调整。为了应对这些挑战，该项目旨在开发下一代可自我进化的植入式设备，允许基于个体患者的情况（例如，数据特征、周围环境、设备条件）。本项目将研究一种软硬件协同探索的自进化设计范式，以实现植入式设备的低功耗、低开销和设备上模型个性化和推理，该项目的成功将导致更高的检测精度、更长的植入式设备寿命和更高的植入式设备患者护理质量。更重要的是，该项目将为不同类型的植入式设备开发一种新的跨层共同探索设计范式，以进行个性化的疾病管理。拟议研究的教育影响包括基于PI可用资源的各种教育活动的整合，如DAC系统设计竞赛和国际CAD的TinyML设计竞赛;通过ND为印第安纳州K-12学生提供的CS课程和名为KidSemicon的在线研讨会，为当地K-12学生提供服务;一个在线社交论坛，允许学生，研究人员和工程师交流，学习和分享可重构半导体技术领域的知识;本科生研究，重点是少数民族的参与，该奖项反映了NSF的法定使命，并被认为值得通过使用基金会的学术价值和更广泛的影响审查标准。

项目成果

Enabling Ultra-Low Power Ultrasound Imaging with Compute-in-Memory Sparse Reconstruction Accelerator

利用内存计算稀疏重建加速器实现超低功耗超声成像

• DOI: 10.1109/biocas58349.2023.10389000
• 发表时间: 2023
• 期刊: IEEE Biomedical Circuits and Systems Conference (Bio-CAS
• 作者: Li, Wantong;Zhang, Xitie;Lee, Junmo;Degertekin, F. Levent;Li, Shaolan;Yu, Shimeng
• 通讯作者: Yu, Shimeng