AI-Powered “Electronic Nose” Sniffs Out Meat Freshness

A team of scientists led by Nanyang Technological University, Singapore (NTU Singapore) invented an artificial olfactory system that mimics the nose of mammals to accurately assess the freshness of meat.

The “electronic nose” (e-nose) consists of a “barcode” that changes color over time in response to the gases generated when meat decays, and a barcode “reader” in the form of a smartphone app that operated with artificial intelligence (AI). The e-nose has been trained to recognize and predict meat freshness using a large library of barcode colors.

When testing commercially packaged meat samples of chicken, fish, and beef that were left to age, the team found that the AI ​​algorithm of the deep neural convolutional network that powers the e-nose kept the freshness of meat at 98.5 percent predicted accuracy. For comparison, the research team evaluated the predictive accuracy of a commonly used algorithm to measure the response of sensors such as the barcode used in this E-Nose. This type of analysis gave an overall accuracy of 61.7 percent.

The e-nose, described in an article published in Advanced Materials journal in October, could help reduce food waste by confirming to consumers whether meat is safe to eat, more accurately than a best-before Label could be a team from NTU Singapore, who worked with scientists from Jiangnan University, China, and Monash University, Australia.

Scientists led by NTU Singapore have invented an artificial olfactory system that mimics the nose of mammals to accurately assess the freshness of meat. Photo credit: NTU Singapore

Co-lead author Professor Chen Xiaodong, director of the Innovative Center for Flexible Devices at NTU, said, “Our artificial olfactory system that we tested in real scenarios can be easily incorporated into packaging materials and yields in a short time without the bulky results Wiring used for electrical signal detection in some recently developed e-noses.

“These barcodes help consumers save money by making sure they don’t throw away products that are still fit for consumption, which is good for the environment too. The biodegradable and non-toxic nature of the barcodes also means that they can be safely applied in all parts of the food supply chain to ensure the freshness of the food. ”

A patent has been applied for this method of real-time food freshness monitoring and the team is now working with an agribusiness in Singapore to expand this concept to other types of perishable goods.

A nose for freshness

The e-nose developed by NTU scientists and their staff consists of two elements: a colored “barcode” that reacts with gases produced by the decay of meat; and a barcode reader that uses AI to interpret the combination of colors on the barcode. In order to make the e-nose portable, the scientists integrated it into a smartphone app, with which results can be obtained in 30 seconds.

The E-nose mimics how a mammalian nose works. When gases from the breakdown of meat bind to receptors in the mammalian nose, signals are generated and transmitted to the brain. The brain then collects these responses and organizes them into patterns so that the mammal can identify the smell that is present as meat age and putrefaction.

In the E-nose, the 20 bars in the barcode act as receptors. Each bar is made from chitosan (a natural sugar) embedded in a cellulose derivative and loaded with a different type of color. These dyes react with the gases produced when meat decays and change color in response to different types and concentrations of gases. This leads to a unique color combination that serves as a “scent fingerprint” for the condition of a meat.

For example, the first bar in the barcode contains a yellow dye that is slightly acidic. When this yellow dye is exposed to nitrogen-containing compounds made by the breakdown of meat (called bioamines), it turns blue when the dye reacts with these compounds. The intensity of the color changes with the increasing concentration of bioamines as the meat continues to break down.

For this study, the scientists first developed a classification system (fresh, less fresh or spoiled) using an international standard that determines the freshness of the meat. It does this by extracting and measuring the amount of ammonia and two other bioamines found in fish packaging wrapped in widely used transparent PVC (polyvinyl chloride) packaging films and stored at 4 ° C (39 ° C)Fahrenheit) over five days at different intervals.

At the same time, they monitored the freshness of this fish packaging with barcodes that were stuck on the inside of the PVC film without touching the fish. Images of these barcodes were taken at various intervals over five days.

The e-nose achieves an overall accuracy of 98.5 percent

One type of AI algorithm known as deep convolutional neural networks was then trained on images of various barcodes to identify patterns in the scented fingerprint that correspond to each category of freshness.

To measure the predictive accuracy of their e-nose, NTU scientists monitored the freshness of commercially packaged chicken, fish, and beef with barcodes taped onto the packaging film and stored at 25 ° C (77 ° Fahrenheit). Over 4,000 images of the barcodes of six meat packages were taken at different time intervals over 48 hours without opening the various meat packages.

The research team first trained its system to recognize patterns among the scent fingerprints captured in 3,475 barcode images before testing the system’s accuracy on the remaining images.

The results showed an overall accuracy of 98.5 percent – 100 percent accuracy in identifying spoiled meat and an accuracy of 96 to 99 percent for fresh and less fresh meat.

For comparison, the research team randomly selected 20 barcode images from each freshness category to assess the predictive accuracy of Euclidean distance analysis, a commonly used method of measuring the response of sensors like the barcode used in this E-Nose. This analysis showed an overall accuracy of 61.7 percent.

Prof. Chen, Chair of the President Professor of Materials Science and Engineering at NTU, said, “While e-noses have been extensively researched, there are still bottlenecks in their commercialization due to the problems of current prototypes with accurately detecting and identifying the odor. We need a system that has both a robust sensor setup and a data analysis method that can accurately predict odor fingerprints. This is what our e-nose offers.

“Its non-destructive, automated, and real-time monitoring functionality could also be used to detect the types of gases that other types of perishable food emit as they become less fresh. This provides a broadly applicable new platform for food quality control that we are now working towards. ”

Reference: “Portable platform for predicting the freshness of food based on colorimetric barcode combiners and deep convolutional neural networks” by Lingling Guo, Ting Wang, Zhonghua Wu, Jianwu Wang, Ming Wang, Zequn Cui, Shaobo Ji, Jianfei Cai, Chuanlai Xu and Xiaodong Chen, October 1, 2020, Advanced materials.
DOI: 10.1002 / adma.202004805