A Low-Cost Design for Wearable Glucose Monitoring, Tested in 428 Trials

Abstract

wearable glucose monitoring is a central problem in biomedical engineering, yet standard techniques scale poorly under distribution shift. Using equipment available in most teaching laboratories, we replicate and extend prior findings across 3965 experimental runs. Results show a 16% gain in accuracy alongside a marked reduction in variance between replicates. These findings suggest a practical path toward broader adoption under distribution shift.