A shoulder-worn sensor detects oxygen deprivation using less power than a nightlight.

Doctors need low-power devices that can detect low oxygen levels, known as hypoxia, directly on wearable sensors. The researchers built a brain-inspired computer chip design using an FPGA, which is a reconfigurable microchip. Their system mimics brain signals using spiking artificial neurons. Early stages process sensory patterns through shared local filters called convolutions, while deeper stages combine everything to make a final decision. A custom software tool trains the network and prepares it to run efficiently on the hardware. First, the chip identified standard image datasets with up to 98% and 86% accuracy. Then, it monitored blood signals and skin tone from a shoulder sensor. It detected hypoxia with 88.26% average accuracy while consuming only 1.455 W of dynamic power.

Why it matters

This enables wearable medical devices to monitor for hypoxia in real-time without draining batteries, making continuous health tracking possible in remote or low-resource areas.

Questions answered