AI could help food systems detect pathogens, fraud, and contamination faster

From detecting salmonella to warning about unsafe food suppliers, a new review shows how AI is advancing food safety research to faster, predictive monitoring

Review: Artificial Intelligence in Food Safety. Image credit: TSViPhoto / Shutterstock

In a recent systematic review published in the journal npj food scienceresearchers synthesized findings from 161 peer-reviewed publications, including journal articles and conference proceedings, to map the rapid growth in research on artificial intelligence (A.I.) Food safety. In this review, we analyzed how machine learning works (M.L.) and deep learning (DL) algorithms are increasingly being utilized to support clinical testing and analysis. Additionally, these models are currently being used to predict chemical contamination and help track foodborne outbreaks, thereby improving the safety and efficiency of the global food supply chain.

The results of this review document the rapid increase in AI research and reported use of AI in published food safety research, in parallel with the move to advanced deep learning models. Ultimately, this review highlights the potential of AI to actively support safer and more resilient global agricultural systems and enhance food safety decision-making.

background

Ensuring food safety is a major global challenge that impacts public health, economic stability, and food security. Traditionally, protecting food intended for human consumption has relied heavily on reactive measures, such as testing food samples after production or investigating outbreaks after consumers become ill.

However, modern agricultural networks generate large amounts of data that cannot be efficiently processed using traditional manual inspection methods. To address this traditional limitation and ensure a safe and verified food supply, experts are increasingly turning to artificial intelligence (AI).

While traditional statistics have long monitored risk, AI is introducing tools such as machine learning, which extracts features from data to predict outcomes, and deep learning, which automatically interprets raw datasets.

The recent surge in interest in leveraging AI in food science is best illustrated by the number of peer-reviewed studies published each year. In 2012, only one study focused on the application of AI in food safety. By 2023, that number has increased to 46. However, until now, a cohesive global map outlining exactly how these algorithms are studied and applied across food safety research has remained lacking.

About reviews

This review aimed to synthesize this expanding scientific landscape and provide a roadmap for future research in this field. This review first screened 783 candidate publications from the SCOPUS database.

To streamline the review process, the review introduced an active learning software tool called ‘ASReview’. ASReview is an ML tool that sequentially ranks papers based on their predicted relevance. In particular, the ML tool was able to reflect the opinions of each researcher to narrow down the choices. Using ASReview, researchers screened 434 records at the title and abstract level before performing full-text evaluation.

After full-text evaluation, 161 primary research articles and peer-reviewed conference proceedings published up to April 2024 were selected for final analysis. These publications were categorized based on research domain, implementation context, data collection methods, other methodologies, and the specific AI architecture used.

This review categorized studies across research areas, including microbiological hazards, chemical contaminants, food authenticity, foodborne illness outbreak surveillance, and broader food safety issues.

Review the findings

Analysis of the reviews revealed that AI is concentrated in specific areas, with microbiological hazards being the most common in 35% of the reviewed literature. In microbiology, 59% of studies used AI to enhance traditional clinical testing. For example, one study combined an “electronic nose” sensor with a classification algorithm to Salmonellaachieving an accuracy of 85% to 100%. In another example, a random forest model predicted a disease endpoint from no tags. Salmonella Analyzes gene sequences with 87% accuracy.

After microbiological hazards, chemical contaminants were the next most common research area, accounting for 25% of the reviewed literature. Here, AI was mainly used to non-destructively detect heavy metals and pesticides. Papers on food fraud and authenticity account for 17% of papers, highlighting applications such as scanning electronic invoices to alert suspicious oil manufacturers.

Apart from this, the application of AI in disease surveillance was also shown to offer potential advantages over traditional methodologies. One system uses anonymized smartphone searches and location data to identify contaminated venues, which the study found was more than three times more effective than traditional surveys.

Finally, the review found that the use of deep learning algorithms in recent research has increased significantly, increasing from 22% of papers in 2019 to 43% by 2023.

conclusion

This review highlights that while AI promises to improve food surveillance, significant hurdles must be overcome before its full potential can be realized. The main factor is severe class imbalance. This is because the majority of food safety data reflects a safe, low-contamination environment, making it difficult for algorithms to recognize rare, high-risk anomalies. Additionally, data privacy and unique restrictions often prevent open data sharing.

The authors also noted that many datasets contain few positive cases and sparse combinations of predictors, making it impossible to systematically compare model performance across studies. Additionally, this review is limited to literature indexed in Scopus, which may underrepresent commercial or manufacturing applications where research results have not been published in peer-reviewed journals.

The authors emphasize that the future will require new solutions such as explainable AI, which demystifies how models make decisions, and distributed federated learning. Adopting these innovations has the potential to move food safety from a primarily reactive system to a more predictive, transparent, and data-driven approach to surveillance.

Click here to download your PDF copy.