EDABK's Edge AI Chip Battles the Silent Cardiac Kille
Today, an increasing number of scientific papers, media repoerts, and television health programs - highlighting the increasing incidence of heart attacks and strokes. What is often overlooked is that a large portion of these cases are linked to atrial fibrillation (AF). In fact, AF is associated with more than 80% of strokes and heart attacks, and it significantly increases both the risk of having a stroke and the risk of dying from it.Even more alarming is the fact that most people with AF are unaware they have it. Current statistics indicate that only around 30% of individuals with AF know about their condition, meaning millions of people worldwide are living with an undiagnosed, potentially life-threatening heart rhythm disorder.That is the reality that EDABK from Vietnam set out to change. Rather than relying on bulky ECG monitors or expensive medical devices, the team developed a low-power edge AI chip that could be embedded into the smartwatches and other wearables devices already used in daily life. Their innovation, which won a Silver Medal at Taiwan's 2025 Best AI Awards, is a signal of where healthcare is headed: real-time, always-on, and hyper-personalized. "We chose to focus on atrial fibrillation detection because we saw a real, unmet need - one that current solutions haven't adequately addressed," said Nguyen Duc Minh, representative of Team EDABK. "We wanted to make AF monitoring accessible, low-cost, and capable of running entirely on a chip, without relying on the cloud."AF is typically detected through electrocardiograms (ECGs), which requires hospital visits and professional equipment. EDABK's idea was to use photoplethysmography (PPG) - the same light-based technology found in off-the-shelf wearables. While PPG data is noisier than ECG, the team developed a processing pipeline that enables reliable AF detection.They began with preprocessing and quantization of the raw PPG signal, followed by a novel Pi-K Plot technique to extract irregularities in heart rhythm. These signal is transformed into spikes and fed into a Spiking Neural Network (SNN) - a brain-inspired AI architecture optimized for low power consumption and low latency.SNNs represent a fundamental shift in how AI operates. Instead of running continuous computations, they activate only "spike" when meaningful data appears, mimicking the behavipr of biological neurons. This event-driven architecture allows AI to run in real-time on small chips, using a fraction of the energy required by traditional neural networks. "Our model reduces RAM usage by more than 80%, while still detecting early warning signs of AF in real time," Minh explained. "It's built for wearables - devices with tight constraints on battery life, processing power, and form factor."The improvement is enormous, with no reliance on cloud services, users receive faster feedback, enhance privacy, and a more seamless user experience. And because the chip behaves like standard memory, it can also be easily integrate into commercial hardware.While still based at a university, EDABK is actively working with local partners to collect clinical data and validate the model beyond the laboratory. The team is already planning a smart ring prototype, complete with PPG sensors and wireless connectivity to deliver alerts directly to a smartphone.Beyond AF, the system's architecture is modular and generalizable. The SNN core and its toolchain can be retrained to detect other heart rhythm disorders or physiological abnormalities - simply by introducing new datasets and adjusting design goals."Our roadmap includes expanding detection to multiple arrhythmias, not just AF," Minh said. "The goal is to build a scalable platform that enables proactive health monitoring in daily life."For Team EDABK, the Best AI Awards did more than validate their work - it opened a new chapter. "We were amazed by how advanced Taiwan is - not just in chip design, but in edge AI innovation at the silicon level," Minh noted. "Some of our members are now considering graduate studies in Taiwan to take our work even further."This experience has also sparked early discussions around potential commercialization in Taiwan, where the supply chain for AI chips and medical wearables is already world-class.As wearable adoption continues to accelerate globally - forecasted to exceed 1.2 billion devices in use by 2028 - solutions like EDAB's EADK's point to a future where AI does not simply reside in the cloud, but operated quietly in the background of everyday life, watching for the warning signs users may not feel."We're not just building a device," Minh said. "We're building a safety net - one that helps users become aware of silent health threats before they become emergencies." In a world where heart disease remains the leading cause of death, that safety net cannot arrive soon enough.EDABK won the Silver Medal in the International Group IC Design Category at the 2025 Best AI Awards. Now it's your chance to shine - bring your innovation to the world and apply for the 2026 Best AI Awards! With global tracks open for both AI Applications and IC Design, students and companies worldwide can compete for the grand prize of up to USD 30,000 (NTD1,000,000). The deadline is March 16, 5pm (GMT+8), follow offcial Linkedin for the latest updates.
