A tiny electronic device built by Northwestern University engineers can spot an abnormal heartbeat before the beat even finishes, using 10,000 times fewer computing operations than standard artificial intelligence.
The research team, led by Mark C. Hersam, the Walter P. Murphy Professor of Materials Science and Engineering, and Indira M. Raman, the Bill and Gayle Cook Professor of Neurobiology, published its findings July 10 in the journal Nature Communications. The study describes a new type of chip called a memtransistor that mimics the brain's cerebellum, the region responsible for split-second reflex reactions.
In proof-of-concept tests, the device identified irregular heart rhythms within one-fifth of a heartbeat and with more than 98% accuracy. It also detected anomalies more than twice as fast as standard software-based detection methods, according to the study.
"The cerebellum is excellent at ignoring the expected and reserving its resources for reacting to the unexpected," Hersam said in a university statement. "That approach ultimately translates into lower energy consumption, and that is where we achieve orders of magnitude improvement."
How it works
Most brain-inspired computing tries to replicate the cerebrum, the brain's center for complex thought. Hersam's team took a different approach, modeling the cerebellum's ability to tune out routine signals and fire only when something unusual happens.
The device is made from molybdenum disulfide, an atomically thin semiconductor. Its asymmetric design lets it switch between excitatory and inhibitory modes by reversing the direction of voltage, mimicking two competing neural signals that balance each other during normal activity but shift when an anomaly occurs.
Unlike conventional computers, which shuttle data between separate memory and processing units, the memtransistor handles both in a single device. That eliminates a major bottleneck and slashes energy use.
Building on earlier work
The new paper advances a line of research Hersam's lab has pursued for years. In October 2023, the team published a study in Nature Electronics showing that just two memtransistors could perform AI classification tasks that otherwise required more than 100 conventional transistors, cutting energy consumption roughly 100-fold. The 2026 study goes further by redesigning the device to detect novelties and make split-second decisions rather than simply classifying data more efficiently.
What comes next
Hersam said the team plans to explore ways to mimic the cerebellum's ability to learn and adapt over time, so that a once-unexpected event occurring repeatedly would no longer trigger an alert. Potential applications include wearable health monitors, autonomous vehicles, and cybersecurity systems.
The study was primarily funded by the National Science Foundation. Co-leads include Vinod K. Sangwan, a research associate professor at Northwestern's McCormick School of Engineering, and Amit Trivedi, an associate professor of electrical and computer engineering at the University of Illinois Chicago. Hersam also chairs the materials science and engineering department and directs the university's Materials Research Science and Engineering Center.
