Machine Learning-Assisted Carbon Quantum Dot-Enhanced Fluorescent Probe for the Detection of Zn2+ in Sweat.

Abstract

Zinc, an indispensable trace element for human body, plays a vital role in numerous physiological processes. While current methods for detecting Zn²⁺ exhibit high sensitivity and specificity, they typically rely on complex instrumentation and entail laborious sample preparations. This study synthesized highly selective fluorescent carbon quantum dots (CQDs) with microcrystalline cellulose extracted from biological waste as the raw material. The synthesized CQDs, leveraging their superior aggregation-induced emission (AIE) properties, enabled the detection of trace levels of Zn²⁺ in sweat and maintained stable fluorescence performance even in the presence of other chemical species. Furthermore, a machine learning-powered detection framework was developed, synergizing spectral feature clustering with a lightweight MobileViT architecture. This intelligent system boosted Zn²⁺ identification accuracy to 82.4% through automated analysis of 650 fluorescence profiles, while enabling real-time quantification. The machine learning-optimized workflow achieved exceptional performance (LOD: 0.17 μM) even in multi-interferent sweat matrices. This machine learning-enhanced CQD-based biosensing method establishes a transformative approach for next-generation trace element monitorin.