Running Quantized Deep Neural Networks on Microcontrollers for Atrial Fibrillation Detection

Core clinical engineering and protocols for TinyML Arrhythmia Detection.

The Challenge of Continuous Wearable Cardiac Monitoring

Streaming continuous uncompressed electrocardiogram data to the cloud drains wearable batteries in hours and overwhelms cellular networks. Smart IoMT solves this by bringing intelligence directly to the sensor node.

Quantized 1D-CNN Architectures on Cortex-M Chips

By utilizing post-training quantization and depthwise separable convolutions, complex neural networks are reduced to fit into dozens of kilobytes of RAM. The embedded model analyzes every heartbeat in real time, waking the radio only when high-acuity arrhythmias (such as ventricular fibrillation or asystole) are identified.

Explorar Portafolio de Dominios Health