Detection of Narcotic Drugs in Urine Samples Using Attenuated Total Reflectance–Fourier Transform Infrared Spectroscopy and Machine Learning Algorithms

Drug abuse is a substantial problem worldwide, with a prevalence of >2% in Southeast Asia. However, the detection of illegal drugs in urine is time-consuming and requires expert interpretation. Therefore, this study aimed to establish attenuated total reflectance-Fourier transform infrared (ATR-FTIR) spectroscopy and machine learning algorithms to screen and analyze urine samples of narcotic drug users for methamphetamines and tetrahydrocannabinol. Urine supernatant samples provided clearer absorption peaks compared to the whole urine samples, allowing for better discrimination between nondrug users and drug users. Principal component analysis revealed key spectral regions (3500-3000 cm^-1 and 1700-1400 cm^-1) for model development. The predictive model was successfully developed using a support vector machine to classify nondrug and drug users' urine samples in the single spectral range of 3500-3000 cm^-1 (<i>F1</i> score of 81.9 ± 12.5%, with 80.8% accuracy, 80.8% sensitivity, and 85.2% specificity). Therefore, the SVM model was considered suitable for classifying drug and nondrug users' urine samples. These findings underscore the promise of ATR-FTIR spectroscopy combined with machine learning for the rapid detection of drugs in urine samples. This innovative technique potentially improves the drug screening process. However, we recommend that the current findings be interpreted as preliminary. Future studies are necessary to increase the sample size, create a comprehensive library of standard spectral data for various narcotics to aid in identification, and optimize the models to enhance the accuracy of classifying drug users' urine samples.