Machine learning-enhanced colorimetric and surface-enhanced Raman scattering-based bimodal lateral flow immunoassay for sensitive and simultaneous detection of the residues of lomefloxacin, ofloxacin, and pefloxacin

Residual quinolone antibiotics (lomefloxacin [LOM], ofloxacin [OFL], pefloxacin [PEF]) in food matrices present significant public health risks due to bioaccumulation and toxicity, necessitating field-deployable methods with superior sensitivity over conventional laboratory techniques. We developed a dual-mode lateral flow immunoassay (LFIA) integrating antibody-conjugated Au@4-ATP@AgNPs nanoprobes with machine learning (ML)-enhanced multi-peak surface-enhanced Raman scattering (SERS) analysis (1071, 1142, 1392, 1442, 1576 cm -1 ) for simultaneous quantification in milk. Results showed the dual-mode SERS-LFIA platform achieved limits of detection (LODs) of 2.94 pg/mL (LOM), 264 pg/mL (OFL), and 1.05 ng/mL (PEF)-representing 330-, 946-, and 476-fold improvements over conventional colorimetric LFIA (CM-LFIA), respectively, with recovery rates of 83.2-118.1% and reproducibility (CV < 13%) in complex milk matrices. ML-based multi-peak analysis reduced quantification error from 21-fold to 13-fold overestimation, compared to single-peak calibration. Overall, this ML-integrated SERS-LFIA delivers unprecedented sensitivity, accuracy, and field-deployability for quinolone residue monitoring in dairy products.