A LOW-POWER NETWORK FOR ONLINE DIAGNOSIS OF HEART-PATIENTS

Implantable cardioverter defibrillators detect and treat dangerous car diac arrhythmias. Current ICDs, however, cannot distinguish between so me potentially fatal arrhythmias and benign conditions. Our system cla ssifies intracardiac electrograms to detect such arrhythmias and uses analog techniques to meet the strict power and area requirements of im plantable systems. A robust neural network architecture reduces the im pact of noise, drift, and offsets inherent in analog approaches.