Multi-Task Learning Mixed-Signal Classifier for In-situ Detection of Atrial Fibrillation and Sepsis

This paper presents an on-chip analog machine learning (ML) classifier IC for detecting atrial fibrillation (AFib) and sepsis from electrocardiogram (ECG) signal. The proposed technique allows continuous in-situ health surveillance using wearables with embedded AI for early detection of underlying health issues. The analog classifier uses custom activation function and performs in-memory computation (IMC) with switched-capacitor circuits for reduced data movement. Designed in 65nm, the test chip achieves average accuracy of 98.2% for AFib detection, and 90.7% for predicting sepsis 4 hours before onset. The energy efficiency of the test-chip is 12.9nJ/classification which is 4× better than state-of-the-art.