Recovery of AFB1 intrinsic fluorescence in vegetable oils with continuous variations in matrices by theoretical analysis and experiments

Abstract

The fluorescence characteristics of aflatoxin B1 (AFB1) have enabled the development of an effective and non-destructive screening method for AFB1 in agro-products using fluorescence spectroscopy. However, the complex and varied matrices present in most foodstuffs can significantly distort the intrinsic fluorescence of AFB1. In this study, the absorption and scattering properties of vegetable oils were obtained using double integrating spheres (DIS), and fluorescence intensity was obtained using laser-induced fluorescence (LIF) technique. A six-parameter analytical model has been developed to recover AFB1 intrinsic fluorescence based on the absorption and scattering features at excitation (375 nm) and emission (424 nm) wavelengths employing a one-dimensional convolutional neural network (1D-CNN). Prediction models for AFB1 concentration in vegetable oils with gradient variations of matrices were calibrated using the disturbed and recovered intrinsic fluorescence intensity, respectively. The models were validated and compared to demonstrate the feasibility and superiority of the proposed method.