Reassessing the Feasibility of PPG-Based Non-Invasive Blood Glucose Level Estimation

Summary

Non-invasive blood glucose level (BGL) estimation from PPG is one of the most commercially attractive claims in wearable sensing, and one of the least comparable across studies — datasets differ, leakage is common, and evaluation metrics are not standardized. This paper contributes the first reproducible, extensible evaluation pipeline for the task and uses it to reassess five representative PPG-based BGL methods on published datasets under three increasingly strict data-split protocols: random window-level splitting, participant-aware splitting, and leave-some-participants-out (LSPO).