New Fluorescence Model Enhances Aflatoxin Detection in Vegetable Oils

Different types of vegetable oils © alex9500-chronicles-stock.adobe.com

Aflatoxin B1 (AFB1) is a potent fungal toxin that frequently contaminates vegetable oils, posing significant health risks (1,2). While fluorescence spectroscopy offers a promising and sensitive non-invasive detection method, complex food matrices often distort AFB1’s intrinsic fluorescence, making detection challenging (1,3). A research team led by Meng Wang, Xiaoqi Zhao, Xiaoyun Yang, and Xueming He has developed an innovative model to overcome these challenges. Their study, published in Food Chemistry, introduces a neural network-based approach to recover AFB1 fluorescence, improving detection accuracy in vegetable oils with varying compositions (1).

The Challenge: Optical Interference in Aflatoxin Detection

Traditional methods for detecting AFB1 in vegetable oils rely on fluorescence spectroscopy, which utilizes AFB1’s natural emission of blue light when exposed to ultraviolet (UV) radiation. However, the presence of various natural compounds in vegetable oils, such as polyphenols, vitamin E, chlorophyll, and carotenoids, can significantly interfere with fluorescence signals. These compounds absorb and scatter light, distorting the emitted fluorescence and leading to inaccurate AFB1 concentration estimates. To address this, the researchers explored a new theoretical and experimental framework to isolate and recover AFB1’s true fluorescence signal (1,3).

Developing a Fluorescence Recovery Model

The team utilized double integrating spheres (DIS) to measure the absorption and scattering properties of vegetable oils and applied laser-induced fluorescence (LIF) to obtain fluorescence intensity readings. By incorporating a one-dimensional convolutional neural network (1D-CNN), they created a six-parameter analytical model capable of recovering AFB1’s intrinsic fluorescence. The model considered variations in absorption and scattering at both excitation (375 nm) and emission (424 nm) wavelengths, allowing for more precise detection of AFB1 regardless of the oil’s composition (1).

The researchers tested their model by comparing fluorescence-based AFB1 concentration predictions before and after applying their recovery algorithm. The results demonstrated that the recovered fluorescence data led to significantly improved detection accuracy, confirming the feasibility and superiority of their method (1).

Advancing Food Safety with Artificial Intelligence (AI)-Driven Spectroscopy

The study marks an advancement in food safety testing. Current methods for AFB1 detection often require complex sample preparation and expensive instrumentation, limiting their practicality for large-scale screening. The newly developed fluorescence recovery model provides a rapid, cost-effective, and non-destructive alternative, making it particularly suitable for routine food safety monitoring (1).

The application of deep learning techniques, such as the 1D-CNN model used in this study, represents a growing trend in analytical chemistry and food science. By leveraging AI-driven spectroscopy, researchers can enhance the accuracy of fluorescence-based detection methods, making them more robust against the interference caused by complex food matrices (1).

Implications and Future Directions

The findings of this research have significant implications for the food industry. By improving the reliability of AFB1 screening in vegetable oils, the model can help regulatory agencies and food manufacturers ensure compliance with safety standards. The study’s approach may also be extended to other food products affected by optical interference, such as grains, nuts, and dairy products (1).

Future research will focus on refining the model further by incorporating additional food matrices and exploring its applicability to real-world industrial settings. With ongoing advancements in AI and spectroscopic techniques, this study paves the way for more sophisticated, accessible, and efficient methods for detecting harmful contaminants in food.

The research showcases the potential of interdisciplinary approaches combining spectroscopy, machine learning, and food science to tackle long-standing challenges in food safety (1).

References

(1) Wang, M.; Zhao, X.; Yang, X.; He, X. Recovery of AFB1 Intrinsic Fluorescence in Vegetable Oils with Continuous Variations in Matrices by Theoretical Analysis and Experiments. Food Chem. 2025, 467, 142305. DOI: 10.1016/j.foodchem.2024.142305

(2) International Agency for Research on Cancer (IARC). Aflatoxin Page. Available at: https://www.iarc.who.int/news-events/mycotoxin-exposure-and-human-cancer-risk-a-systematic-review-of-epidemiological-studies/ (accessed March 4, 2025).

(3) Magnus, I.; Abbasi, F.; Thienpont, H.; Smeesters, L. Laser-Induced Fluorescence Spectroscopy Enhancing Pistachio Nut Quality Screening. Food Control 2024, 158, 110192. DOI: 10.1016/j.foodcont.2023.110192