Qualitative analysis of wheat aflatoxin B1 using olfactory visualization technique based on natural anthocyanins

Abstract

Wheat is highly susceptible to aflatoxin B1 (AFB1) contamination, which affects food safety. This study proposed a new method that combines natural anthocyanin olfactory visualization technology with machine learning algorithms to detect the degree of wheat AFB1 contamination. The study used solvent extraction to extract anthocyanins from a variety of plant materials, and verified the effectiveness and applicability of the extraction by measuring the total anthocyanin content and UV-Vis spectroscopy. The pre-experiment identified nine appropriate anthocyanins as dyes, followed by the development of a sensor array to collect volatile odor data from wheat samples with differing AFB1 levels. The t-distributed Stochastic Neighbor Embedding (t-SNE) algorithm was used to downscale the color change data of the sensors before and after responding to different samples, and a support vector machine (SVM) classification model was constructed to identify the contamination degree of wheat samples. Particle Swarm Optimization (PSO) and Transient Trigonometric Harris Hawks Optimizer (TTHHO) are employed to optimize the SVM model. The findings indicated that the TTHHO-SVM model had superior performance in assessing the AFB1 contamination level in wheat, achieving an accuracy of 97.9 %. It was demonstrated that anthocyanin dye as a colorimetric sensor material could effectively and sensitively distinguish the degree of mold in wheat. This method effectively reduces the high cost and time consumption of traditional AFB1 detection methods and has potential applications.