Researchers at the University of Connecticut (UConn) have developed low-cost, artificial intelligence-based methods that can quickly and easily detect microbial contamination and spoilage in milk and meat.

Notably, the method for detecting bacteria in milk was able to successfully identify eight pathogenic and spoilage microorganisms with greater than 98 percent accuracy, within two hours. The findings were published in Food Chemistry.

The development of the detection methods was led by Yangchao Luo, Ph.D.’s research group in the UConn College of Agriculture, Health, and Natural Resources, in collaboration with Zhenlei Xiao, Ph.D. Drs. Luo and Xiao are both professors in UConn’s Department of Nutritional Sciences.

Rapid, Simultaneous Detection of Milk Contaminants

The system for detecting microbial contaminants in milk uses a 96-well plate paired with an array of 12 sensors that react to bacteria based on molecular structure, producing distinct signal patterns. These patterns are analyzed using a machine learning algorithm trained to recognize specific pathogens.

The researchers tested the method against five pathogenic bacteria, including Listeria, Escherichia coli, and Salmonella, along with three spoilage organisms, representing common microbial milk contaminants.

Traditional microbiological methods typically detect one pathogen at a time and can require several days, specialized laboratory infrastructure, and trained personnel. In contrast, the new approach enables simultaneous detection of multiple organisms in a shorter timeframe without requiring advanced technical training.

The novel sensor array’s accuracy, low cost, and ease of use offer a promising alternative to traditional plate counting and ELISA methods.

“We hope to develop a technology that can simultaneously detect as many species as possible so that we can easily trace back the original source of contamination,” said Dr. Luo.

Potential for Onsite and Consumer-Level Testing

Given that performing the milk test does not require specialized training, the researchers hope their method could also support the development of simplified, point-of-use testing solutions.

At present, the group is working to create a smartphone app to interpret fluorescence signals generated by the sensors. They are also working to eliminate sample preparation steps, such as protein removal, that can affect test accuracy.

Sensor Technology Enables Non-Invasive Meat Quality Monitoring 

In related work published in Food Frontiers, the researchers developed sensors to detect volatile organic compounds (VOCs) associated with spoilage bacteria activity in meat.

The VOC-based system identifies spoilage and potential contamination by detecting gases emitted by bacteria, producing a measurable color change. Machine learning models were trained to associate VOC patterns with specific bacterial types.

Unlike traditional sampling methods, VOC detection does not require direct contact with the product, which the researchers noted may simplify testing for intact meat cuts. The team indicated that the technology could also be integrated into packaging to provide a visual indicator of freshness or contamination risk.

The researchers said the approach may offer a non-invasive method for monitoring meat quality throughout storage and distribution.