Speakers
Description
Accurately measuring blood glucose levels without drawing blood remains difficult because glucose has weak optical absorption and biological tissue introduces strong background interference. In this work, we explore whether a compact, low-cost multispectral sensor (SparkFun Triad AS7265x, 410–940 nm) can detect glucose-related optical changes. Sugar solutions ranging from 5–35% w/v were measured at several sensor–sample distances, with an 8 mm spacing yielding the most consistent spectral data. After normalization, Partial Least Squares (PLS) regression produced excellent calibration performance (R² ≈ 1.00; RMSE ≈ 0.10%), demonstrating sensitivity to concentration-dependent changes stemming from scattering and refractive index variations. To better mimic physiological conditions, glucose was added to tissue-like optical phantoms composed of Intralipid and India ink. Preliminary measurements show subtle yet detectable spectral trends in the visible to near-infrared region. Overall, these results suggest that pairing low-cost multispectral sensing with multivariate analysis can provide coarse glucose estimates and represents a promising step toward portable, non-invasive monitoring technologies.
