Machine Learning Applied to the Detection of Mycotoxin in Food: A Review

TL;DR

This paper surveys how machine learning can enhance detection and prediction of mycotoxins in foods, addressing a major public health and economic problem. It synthesizes findings across data types, notably spectral/hyperspectral and imaging data, and stratifies methods into neural networks, random forests, gradient boosting, SVMs, and other techniques. The review highlights that neural networks dominate the literature, with strong performance reported, but reproducibility remains a critical issue due to sparse hyperparameter reporting and limited open-source code. It concludes that while ML offers significant potential for rapid, scalable screening and risk assessment, broader industry adoption requires standardized reporting, robust validation, and accessible data/code. The work also points toward interpretability enhancements (e.g., SHAP/LIME) and standardized benchmarks to enable reliable deployment in diverse agricultural settings.

Abstract

Mycotoxins, toxic secondary metabolites produced by certain fungi, pose significant threats to global food safety and public health. These compounds can contaminate a variety of crops, leading to economic losses and health risks to both humans and animals. Traditional lab analysis methods for mycotoxin detection can be time-consuming and may not always be suitable for large-scale screenings. However, in recent years, machine learning (ML) methods have gained popularity for use in the detection of mycotoxins and in the food safety industry in general, due to their accurate and timely predictions. We provide a systematic review on some of the recent ML applications for detecting/predicting the presence of mycotoxin on a variety of food ingredients, highlighting their advantages, challenges, and potential for future advancements. We address the need for reproducibility and transparency in ML research through open access to data and code. An observation from our findings is the frequent lack of detailed reporting on hyperparameters in many studies as well as a lack of open source code, which raises concerns about the reproducibility and optimisation of the ML models used. The findings reveal that while the majority of studies predominantly utilised neural networks for mycotoxin detection, there was a notable diversity in the types of neural network architectures employed, with convolutional neural networks being the most popular.

Figures (8)

• Figure 1: Number of publications between 2013 and 2023 found by our systematic search criteria in Scopus.
• Figure 2: Most popular machine learning methods reviewed in this work.
• Figure 3: Typical machine learning process.
• Figure 4: Basic neural network structure, showing an input layer, two hidden layers, and an output layer, where each circle represents a neuron and are interconnected by lines symbolising neural connections. The input layer receives the initial data, which is then processed through successive hidden layers using weights and activation functions, refining the information before it reaches the output layer.
• Figure 5: Decision tree process demonstrating the structure of a decision tree, including the root node, branching to decision nodes, and culminating in leaf/terminal nodes. The depth of the tree is indicated, showing the levels of decision-making from the root to the leaves.