A Smartphone-Read, Machine Learning-Enhanced Lanthanide Hydrogel Sensor for Multimodal and On-Site Detection of Kanamycin in Food Safety

Residual kanamycin from veterinary use poses a persistent threat to environmental safety and public health, as its accumulation in water systems and animal-derived foods can foster antibiotic resistance and induce toxic effects. To address the urgent need for on-site monitoring, we developed mechanically robust, self-healing, dual-network hydrogels (Gel-1 and Gel-2) coordinated with lanthanide ions (Eu³⁺ and Tb³⁺). The hydrogels exhibit tunable porosity, mechanical robustness, and autonomous damage recovery, making them ideal for field use. Upon interaction with kanamycin, the tailored ligands trigger distinct fluorescence responses via ligand-to-lanthanide energy transfer, as confirmed by experimental and theoretical studies. We designed a hierarchical detection strategy integrating four complementary modes: (1) single-emission turn-on fluorescence for coarse screening; (2) self-calibrated ratiometric sensing for improved accuracy; (3) machine learning-assisted full-spectrum analysis for high-precision quantification; and (4) smartphone-based colorimetric detection for equipment-free on-site use. This multimodal sensor achieves a remarkable detection limit of 0.950 ppm and a broad dynamic range of 1-5000 μM. The synergy between material design and data science underscores the potential of intelligent hydrogel sensors for monitoring antibiotic residues in food and environmental safety.