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

Area of Science:

• Food Science
• Computational Biology
• Public Health

Background:

• Mycotoxins, toxic fungal metabolites, contaminate food crops, posing risks to human and animal health.
• Traditional detection methods are slow and impractical for large-scale food safety screening.
• Machine learning (ML) shows promise for rapid and accurate mycotoxin detection in the food industry.

Approach:

• Systematic review of recent ML applications for mycotoxin detection in food ingredients.
• Analysis of ML model advantages, challenges, and future potential.
• Emphasis on the importance of data and code transparency for research reproducibility.

Key Points:

• Neural networks, particularly convolutional neural networks (CNNs), are widely used for mycotoxin detection.
• Many studies lack detailed reporting of hyperparameters and open-source code, hindering reproducibility.
• ML offers significant potential for enhancing global food safety and public health surveillance.

Conclusions:

• ML advancements are crucial for improving food safety and mitigating mycotoxin risks.
• Standardized reporting and open data practices are essential for advancing ML in mycotoxin research.
• Future research should focus on addressing reproducibility concerns and exploring diverse ML architectures.