Rapid quantitative detection of Ractopamine using Raman scattering features combining with deep learning

Abstract

Establishing a universal and efficient method for determining ractopamine residues in pork is of paramount importance for ensuring food safety. However, the main challenge lies in achieving accurate quantitative detection of complex samples using Surface-Enhanced Raman Scattering (SERS), as it requires overcoming interference from substrate-sample mixing and variations in hotspot distribution. This study introduces an innovative approach to address this challenge by proposing a breakthrough interference factor removal network based on deep learning, termed SERSNet. By enhancing the depth of SERS spectroscopy, SERSNet establishes a correlation between the spectra of pork samples with varying concentrations of ractopamine. A multilayer convolution module is developed to effectively extract the spectral features of ractopamine. The Mean Absolute Error (MAE), root mean square error (RMSE), and Mean Absolute Percentage Error (MAPE) of the proposed model in this paper are 0.90, 0.48, and 80.48, respectively. The performance of the SERSNet model surpasses that of the Multiple Linear Regression (MLR) model. The SERSNet algorithm proposed in this paper demonstrates competitiveness and yields superior results.