AI-Integrated Conformal Cellulose Nanofiber-Reinforced Hydrogel Patch for Dual-Ratiometric Fluorescence Monitoring of Pyrimethanil Residues

Real-time and on-site monitoring of pesticide residues remains challenging due to the lack of plant-interfaced, actively enriching, and nondestructive analytical devices. Herein, we present an intelligent, plant-conformal cellulose nanofiber (CNF)-reinforced hydrogel sensor with dual-emission fluorescence for in situ quantification of the fungicide Pyrimethanil (PMT). Unlike passive substrates, the poly(vinyl alcohol) (PVA)/CNF hybrid matrix serves as a mechanically robust, active enrichment scaffold. It exhibits a highly interconnected porous architecture and excellent flexibility (fracture strain >370%), enabling stable, conformal adhesion to complex leaf surfaces. Crucially, the CNF network acts as an active enrichment scaffold, preconcentrating PMT via specific intermolecular interactions to enhance signal intensity and response kinetics. For signal transduction, in situ synthesized Eu-MOF and coordinated carbon dots (CDs) provide a stable dual-color emission, enabling a self-referencing ratiometric mechanism intrinsically resistant to light scattering and instrumental drift. Furthermore, to achieve high-fidelity quantification, a back-propagation artificial neural network (BP-ANN) is integrated and trained on multichannel optical (RGB) and compositional data sets. This AI-driven quantification accurately maps RGB features to PMT concentrations, reducing operator bias and mitigating illumination-related interference. The resulting hydrogel patches tightly conform to diverse plant surfaces, exhibit excellent mechanical robustness, and demonstrate a low detection limit of 33.46 μg·kg^-1 and satisfactory recovery rates ranging from 94.0% to 104.7% in real agricultural samples. Overall, this synergistic integration of CNF-assisted chemical enrichment, dual-emission ratiometric response, and data-driven analytics establishes a robust, field-deployable platform for real-time, high-fidelity monitoring of pesticide residues, offering a promising paradigm for sustainable and intelligent agricultural diagnostics.