Ratiometric Determination and Discrimination of Oxicams via Dual-Excitation Carbon Dots Assisted by Machine Learning

Oxicams, a major category of nonsteroidal anti-inflammatory drugs, are widely used in daily life. However, excessive consumption of oxicams can pose significant risks to human health. Herein, we introduce an innovative and highly sensitive fluorescent approach for the detection and discrimination of oxicams. This proposed method utilizes the fluorine and nitrogen codoped carbon dots, which display a bright blue emission with two excitation peaks at 280 and 340 nm, synthesized via a hydrothermal method. The variation in the ratio of excitation intensities, measured at a constant emission wavelength, exhibits a linear relationship with the concentration of oxicams. To validate the feasibility of this approach, meloxicam (MLX) is selected as the model compound. The method demonstrates high sensitivity, achieving a low limit of detection (LOD) of 97 nM across a broad concentration span from 0.097 to 25 μM. Moreover, after comparing various machine learning algorithms, the XGBoost algorithm is identified as the optimal choice for discriminating oxicams at ultralow concentrations (0-3.5 μM), with 100% accuracy for unknown samples outside the data sets. Finally, a convolutional neural network (CNN) algorithm-assisted sensing platform has been successfully implemented for the accurate prediction of oxicams in real samples. In summary, this study broadens the application horizon of carbon dots in sensing technologies and offers a viable strategy for real-time monitoring of oxicams, ultimately benefiting public health.