Reliable Noninvasive Glucose Sensing via CNN-Based Spectroscopy

TL;DR

This paper introduces a dual-modal AI framework using SWIR spectroscopy and machine learning for reliable, non-invasive glucose sensing, achieving high accuracy and clinical relevance with cost-effective wearable solutions.

Contribution

It presents a novel dual-modal approach combining CNN-based imaging and photodiode sensors for non-invasive glucose monitoring, demonstrating state-of-the-art accuracy and clinical potential.

Findings

CNN achieved 4.82% MAPE at 650 nm

Photodiode system reached 86.4% Zone A accuracy

Framework balances clinical accuracy, cost, and wearability

Abstract

In this study, we present a dual-modal AI framework based on short-wave infrared (SWIR) spectroscopy. The first modality employs a multi-wavelength SWIR imaging system coupled with convolutional neural networks (CNNs) to capture spatial features linked to glucose absorption. The second modality uses a compact photodiode voltage sensor and machine learning regressors (e.g., random forest) on normalized optical signals. Both approaches were evaluated on synthetic blood phantoms and skin-mimicking materials across physiological glucose levels (70 to 200 mg/dL). The CNN achieved a mean absolute percentage error (MAPE) of 4.82% at 650 nm with 100% Zone A coverage in the Clarke Error Grid, while the photodiode system reached 86.4% Zone A accuracy. This framework constitutes a state-of-the-art solution that balances clinical accuracy, cost efficiency, and wearable integration, paving the way for reliable continuous non-invasive glucose monitoring.