AI-assisted two-step enhanced SERS platform for rapid and ultra-sensitive detection of toxic molecules in biofluids

Rapid detection of toxic molecules in biological fluids is crucial for clinical diagnosis, emergency toxicology, and forensic investigations. However, traditional methods such as liquid chromatography and gas chromatography are limited of complex sample preparation, time - consuming detection, and lack a basis for detecting different matrices and multiple toxicants. Herein, a two-step enhancement platform Ag@BOCPs based on surface-enhanced Raman spectroscopy (SERS) was developed, overcoming the limitations of the poor universality of the conventional silver nanoparticle. Plasma resonance-induced enhancement and dynamic surface cleaning enable highly sensitive, rapid and reproducible detection of toxic substances in biofluids. This SERS platform can rapidly analyze toxic substances such as prescription drugs, food additives, and pesticides in a wide range of biological fluids including blood, urine and breast milk. The detection limit for Hexadecyl trimethyl ammonium Bromide is 200 ng/mL, and 10 ng/mL for Thiabendazole, demonstrating extremely high sensitivity. By integrating machine learning algorithms such as Uniform Manifold Approximation and Projection (UMAP) and Support Vector Machine (SVM), the system can automatically classify toxic substances with high accuracy of 94 %. Clinical validation demonstrated excellent reproducibility (with RSD lower than 5 %) and strong linearity (coefficient of determination higher than 0.99) of patient samples in quantitative analysis. This AI-assisted SERS platform provides a fast, reliable and versatile method for toxicology screening, with a wide range of applications in fields such as emergency medicine, forensic science and food safety.