Deep learning-assisted surface-enhanced Raman spectroscopy detection of stimulants

The abuse of stimulants poses a significant threat to public health and the integrity of competitive sports, necessitating the development of highly sensitive and rapid-response detection technologies for effective regulation. This study integrates Surface-Enhanced Raman Spectroscopy (SERS) with deep learning algorithms to achieve high-sensitivity detection and identification of five stimulants, including clorprenaline, propranolol, terbutaline, tulobuterol, and cimaterol. First, intrinsic vibrational information of the molecules was obtained using conventional Raman spectroscopy, and the identification of characteristic peaks was achieved through Density Functional Theory calculations. Subsequently, gold nanoparticles (Au NPs) were employed as substrate materials, enabling the trace detection of the five stimulants using SERS. To validate the practical application potential, SERS detection was conducted on spiked blood samples for five stimulants, achieving trace detection of all five stimulants. Finally, four machine learning algorithms, support vector machine (SVM), deep neural network (DNN), recurrent neural network (RNN), long short-term memory network (LSTM), were constructed to classify and identify the obtained SERS spectral data, with the LSTM algorithm achieving an identification accuracy of 99.7%. This method holds promise for expansion into trace analysis scenarios such as food detection monitoring and biomarker screening.