Abstract
With the growing global challenge of drug abuse, there is an urgent need for rapid, accurate, and cost-effective drug detection methods. This study introduces an innovative approach to drug abuse screening by quickly detecting ephedrine (EPH) in tears using drop coating deposition-surface enhanced Raman spectroscopy (DCD-SERS) combined with machine learning (ML). Using ultra performance liquid chromatography-tandem mass spectrometry (UPLC-MS/MS), the average concentration of EPH in tear fluid of Sprague-Dawley (SD) rats, measured over 3 h post-injection, was 1235 ng/mL. DCD-SERS effectively identified EPH in tear samples, with distinct Raman peaks observed at 1001 cm(-1) and 1242 cm(-1). To enable rapid analysis of complex SERS data, three ML algorithms-linear discriminant analysis (LDA), partial least squares discriminant analysis (PLS-DA), and random forest (RF)-were employed. These algorithms achieved over 90% accuracy in distinguishing between EPH-injected and non-injected SD rats, with area under the ROC curve (AUC) values ranging from 0.9821 to 0.9911. This approach offers significant potential for law enforcement by being easily accessible, non-invasive and ethically appropriate for examinees, while being rapid, accurate, and affordable for examiners.