Determination of aflatoxin B1 in wheat using Raman spectroscopy combined with chemometrics

Abstract

Aflatoxin B1 (AFB1) is carcinogenic and highly susceptible to production in wheat. In this study, the quantitative detection of contaminant AFB1 in wheat was investigated by Raman spectroscopy combined with chemometric method realization. Firstly, Savitzky–Golay smoothing (SG) and baseline calibration methods were used to perform the necessary preprocessing of the collected raw Raman spectra. Then, three variable optimization methods, i.e., competitive adaptive reweighted sampling (CARS), iteratively variable subset optimization (IVSO), and bootstrap soft shrinkage (BOSS), were applied to the preprocessed wheat Raman spectra. Finally, partial least squares regression (PLSR) models were developed to determine AFB1 in wheat samples. The results showed that all three variable optimization algorithms significantly improved the predictive performance of the models. The BOSS-PLSR model has strong generalization performance and robustness. Its prediction coefficient of determination ($R_P^2$) was 0.9927, the root mean square error of prediction (RMSEP) was 2.4260 $\mu g$/kg, and the relative prediction deviation (RPD) was 11.5250, respectively. In conclusion, the combination of Raman spectroscopy and chemometrics can realize the rapid quantitative detection of AFB1 in wheat.