Northwestern researchers said their cerebellum-inspired memtransistor detected abnormal heart rhythms within one-fifth of a heartbeat, identifying arrhythmias in ECG tests with more than 98% accuracy.
Roughly 10,000 times fewer computer operations were needed because the device ignores routine signals and reacts only when incoming data deviates from expected patterns.
The hardware combines memory and processing in one memtransistor and switches between excitatory and inhibitory modes by reversing voltage, mimicking cerebellar circuits that flag novelty.
Nature Communications published the study on July 10, extending the lab’s earlier low-power AI work beyond classification toward always-on monitoring for wearables, robots, vehicles and cybersecurity systems.
Next, the team plans to add adaptive learning so repeated surprises stop being treated as novel, moving closer to fuller cerebellum-like AI hardware.
This new AI spots sudden changes, but could it miss slower, more sinister threats in cybersecurity or health?
Is mimicking the brain's efficiency the ultimate solution to AI's massive and growing energy crisis?
When will this brain-inspired chip let our smartwatches run for months, not just days, on a single charge?
Northwestern’s Brain-Inspired AI Chip Enables Instant, Ultra-Low Power Arrhythmia Detection and Edge Intelligence
Overview
Northwestern University researchers have announced a brain-inspired AI chip that can instantly detect cardiac arrhythmias. This breakthrough device uses a cerebellum-inspired memtransistor, which combines memory and computation in a single architecture. Built with molybdenum disulfide and featuring an asymmetric transistor design, the chip can switch between excitatory and inhibitory modes by reversing voltage, mimicking the brain’s neural processing. During testing, the chip processed ECG data with remarkable speed, efficiently filtering out normal heartbeats and focusing on anomalies. This innovation marks a major step forward in fast, energy-efficient medical diagnostics.