# Pigment-Resistant, Portable Corneal Fluorescence Device for Non-Invasive AGEs Monitoring in Diabetes

## TL;DR

A new portable device uses corneal fluorescence to non-invasively monitor AGEs in diabetes, showing good accuracy and minimal interference from skin pigmentation.

## Contribution

A non-invasive, portable device for AGEs monitoring with strategies to reduce pigmentation interference and integration of BMI for improved accuracy.

## Key findings

- Purple and blue pigments caused more interference than green and pink in fluorescence measurements.
- Device-derived AGEs estimates showed less than 8% mean error compared to serum AGEs.
- Incorporating BMI into predictive models improved accuracy and reflected dietary and metabolic influences.

## Abstract

Advanced glycation end products (AGEs) are important biomarkers associated with diabetes and metabolic disorders; yet existing detection methods are invasive and unsuitable for frequent monitoring. This study aimed to develop a non-invasive and portable AGEs detection device, optimize strategies for mitigating pigmentation-related interference, and evaluate its feasibility for metabolic assessment. The proposed system employs a 365 nm ultraviolet LED excitation source, an optical filter assembly integrated into an ergonomic dark chamber, and an eyelid-signal-based algorithm to suppress ambient light and skin pigmentation interference. Simulation experiments were conducted to evaluate the influence of different pigment colors and skin tones on fluorescence measurements. A clinical study was performed in 200 participants, among whom 42 underwent concurrent serum AGEs measurement as the reference standard. Predictive models combining corneal fluorescence signals and body mass index (BMI) were constructed and evaluated. The results indicated that purple and blue pigments introduced greater interference, whereas green and pink pigments had minimal effects. Device-derived AGEs estimates demonstrated good agreement with serum AGEs, with a mean error below 8%. A hybrid model incorporating BMI achieved improved predictive accuracy compared with single-parameter models. Participants with high-AGE dietary habits exhibited elevated fluorescence signals and BMI. These findings suggest that the proposed device enables stable and accurate non-invasive AGEs assessment, with potential utility for metabolic monitoring. Incorporating lifestyle-related parameters may further enhance predictive performance and expand clinical applicability.