AI-Enhanced Rapid SERS Screening of Trace Quinolone Antibiotics across the Source-Pathway-Sink Ecosystem

The overused quinolone antibiotics in animal husbandry and clinical medicine pose a growing threat for global health as they enter ecosystem via agricultural discharge and medical wastewater. Consequently, risk assessment for environmental and human exposure is highly demanded, which calls for an accurate, reliable and rapid qualitative and quantitative analysis of trace antibiotics in different matrices. Surface-enhanced Raman spectroscopy (SERS), providing fingerprint chemical information with near-single-molecule sensitivity, has received considerable attention. Herein, considering the common physicochemical properties of various quinolone antibiotics, we first developed sample pretreatment strategy consisting of solid-phase extraction (SPE), liquid-liquid extraction (LLE), salting-out assisted liquid-liquid extraction (SALLE), and NPs purification (NPs (p)). The modular SPE-LLE-SALLE-NPs (p) coupling with spectral separate network (SSNet) (a deep-learning based spectral unmixing algorithm) enabled a rapid (<8 min per sample), sensitive (μg L^-1), on-site, and intelligent SERS analysis of 19 quinolone antibiotics in food, biological, and environmental matrices. Besides, the identification of each QN in a five-target mixture sample enabled SPE-LLE-SALLE-NPs (p)-SSNet-SERS to be a highly promising rapid detection strategy in blind sample screening qualitatively and quantitatively. The integrated SPE-LLE-SALLE-NPs (p)-SSNet-SERS offers a cross-media, full-process, and big data-enhanced rapid monitoring solution for antibiotic pollution within the "source-pathway-sink" framework.