Classification of Contaminated Dried Figs According to Their Aflatoxin B1 Concentrations by Deep Learning-Based VGG16 Model

Aflatoxins are toxic secondary metabolites; among them, aflatoxin B1 is classified as a Group 1 carcinogen by the International Agency for Research on Cancer. Aflatoxins exhibit fluorescence under ultraviolet light. Kojic acid, another substance produced by Aspergillus species, has also been reported to exhibit fluorescence. These compounds are commonly detected using chromatographic methods and rapid test kits. This study aimed to classify aflatoxin B1 concentration in dried figs by analysing UV images using a convolutional neural network-based image processing approach, thereby reducing dependence on chromatographic methods. Fresh figs were sliced, dried, and inoculated with Aspergillus flavus (A. flavus) Changes in aflatoxin and kojic acid levels were monitored using chromatographic analyses over a 21-day period, at 7-day intervals, while images were acquired concurrently. Training, validation, and test datasets were generated using UV images of different aflatoxin B1 standard solutions at concentrations of 5 to 250 μg/L. The VGG16-based model achieved an overall precision, recall, and F1-score values of 0.96, 0.96, and 0.95, respectively. The model accuracy for fig samples ranged from 0.64 to 1.00, highlighting the potential of image-based analysis as a rapid screening approach. On day 0, no aflatoxin B1 was detected by HPLC, and no fluorescence was observed in the fig samples. Accordingly, the model classified all samples as aflatoxin B1-negative. In conclusion, the developed model demonstrated its ability to classify predefined aflatoxin B1 concentration levels, including the lowest class of 5 μg/L. The model has the potential to contribute significantly to the development of fast, low-cost, image-based analysis systems.