In the last decade, there has been a revolution in biosensor devices producing a wealth of new data from patients. Increased density of sensors, for example embedded in implantable organ-conformal bioelectronic platforms, can provide high-definition spatial and temporal data for increased accuracy of detection and diagnostics. However, gathering a large quantity of data is insufficient if it is not matched by powerful circuitry to process the data and apply a therapeutic response in sometimes as fast as milliseconds. The next generation of biomedical technologies requires novel computing that can be embedded non-intrusively and can translate large quantities of time-sensitive biomedical data into a life-saving response rapidly and energy-efficiently. Distributed neuromorphic computing can provide faster, lower-power and more compact implementations than conventional computing, especially if implemented with emerging device technologies like resistive switches or memristors. Such neuromorphic chips designed for distributed computing can be co-located with the sensors and actuators for closed-loop diagnostics and therapy. This project will provide the fundamental investigations into the requirements and architecture for such distributed closed-loop computing. This technology will impact research in biomedical engineering, opening the path to computing of large amounts multi-physics data gathered in-vivo from organs and artificial tissue. Long term, this work could benefit healthcare, when embedded neuromorphic chips are integrated in a stand-alone implantable system that can enable real-time diagnostics and painless therapy for patients. These findings are also directly transferable to other applications in need of embedded computing hardware such as microrobots and Internet of Things (IoT).
This project will develop a neuromorphic computing solution that can be reliably embedded with existing sensor and actuator organ-conformal platforms. The proposed technology is based on hybrid neuromorphic chips organized in a cellular neural network with recurrence. To demonstrate its potential, the computing platform is be prototyped and tested for the analysis of cardiac wavefronts, which have stringent time constants of milliseconds. Initially, theoretical investigations will focus on developing a hardware-mappable algorithm that can distinguish electrical storms from normal electrical wave patterns. Then, experimental work focuses on taping out a cellular neural network processing unit in a hybrid memristor / transistor (CMOS) technology and testing a small distributed network of such units. This work benefits from the use of cardiac animal and human data for testing, thanks to collaborators in the Biomedical Engineering Department at George Washington University. The prototype demonstration and the supporting simulations serve as hands-on materials in a class on neuromorphic hardware and in various outreach efforts planned as part of the new GWU Center for Women in Engineering. A PhD student is being recruited, and an undergraduate student and a high school student are to be involved in the design and testing of this computing platform.
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.
