A Deep Sparse Capsule Network for Non-Invasive Blood Glucose Level Estimation Using a PPG Sensor

Summary

This paper from LSU Health Sciences Center and the University of Glasgow introduces a Deep Sparse Capsule Network (DSCNet) for non-invasive blood glucose level (BGL) monitoring using PPG sensor data. Capsule networks, originally proposed by Hinton for image recognition, use dynamic routing between capsule groups to encode part-whole relationships — applied here to encode the relationship between different PPG morphological features relevant to glucose estimation. The sparse capsule formulation reduces parameter count and computational requirements compared to standard capsule networks.