E-nose and AI combined to detect mycotoxins in pet food

Poisons produced by certain fungi, or mycotoxins, have serious ramifications for the pet food industry. For example, mycotoxin-contaminated cat food ingredients may have caused the 2021 wave of feline pancytopenia that killed more than 350 cats in the United Kingdom. Dangerous mycotoxins, such as aflatoxins, zearalenone, deoxynivalenol and fumonisin B1, have been found in pet food and ingredients around the world. These mycotoxins can cause both acute poisoning and chronic health issues, including immune suppression, liver damage, kidney toxicity and cancer.

Early and accurate detection of mycotoxins in the pet food supply chain can prevent these health threats. However, traditional detection methods, such as chromatographic techniques and immunoassays often require complex sample preparation and expensive instrumentation.

Mycotoxin detection using AI and E-noses

A team of scientists at the Chinese Academy of Agricultural Sciences set out to develop a cost-effective rapid means of detecting zearalenone in pet foods. The researchers experimented with a combination of electronic nose (E-nose) technology and machine learning algorithms, a type of artificial intelligence that allows a computer to analyze data and teach itself without explicit programming. The scientists’ approach detected volatile compounds associated with mycotoxin contamination, potentially providing a non-invasive, rapid, and cost-effective alternative to traditional methods. They published their results in the journal Toxins.

In their study, 142 pet food samples collected in China between 2021 and 2023 were analyzed for zearalenone contamination. Using E-nose technology, researchers recorded volatile compound profiles to classify contamination levels. Machine learning models trained on this data proved accurate, with an ensemble model achieving 90.1% classification accuracy. This approach effectively identified zearalenone contamination.

The integration of E-nose technology with machine learning offers several key advantages:

1. Speed and Efficiency: The E-nose can analyze multiple samples rapidly, facilitating real-time decision-making during production.
2. Cost-Effectiveness: Compared to traditional detection methods, this approach reduces reliance on expensive equipment and labor-intensive processes.
3. High Accuracy: The ensemble model's 90.1% accuracy underscores the reliability of this method, outperforming standalone models.

This study demonstrated the potential of machine learning-enhanced E-nose technology as a tool for the pet food industry. By enabling rapid, accurate and cost-effective screening, this approach may equip manufacturers with the means to ensure product safety and meet regulatory standards. As mycotoxin contamination continues to pose risks, innovative solutions like this will play a vital role in safeguarding pet health and maintaining consumer trust.

The adoption of advanced technologies in quality assurance processes is not just a necessity but an opportunity for the pet food industry to set new benchmarks in product safety and innovation. With further refinement and industrial-scale implementation, E-nose technology could become a cornerstone of mycotoxin management, enhancing the safety and quality of pet food worldwide.