Digitally measuring meat

Researchers at Teagasc Food Research Centre, Ashtown, are investigating how effectively optical sensors and machine learning can be used to monitor the quality of processed meats and meat alternatives.

Processed meats represent a significant portion of the human diet. Numbers show that global meat consumption has risen steadily from 70.6 million tonnes in 1961 to 352.1 million tonnes in 2021.

Processed meats offer consumers a wide range of options, through a variety of processing and preparation methods, and even regional variances as seen with protected designation of origin (PDO) and protected geographical indication (PGI) varieties such as Spanish jamón ibérico and German Ammerländer schinken. All of these factors aim to offer a unique experience for consumers in terms of sensorial attributes like juiciness, texture and flavour.

Conversely, people are increasingly concerned with following healthy eating patterns. In this respect, consumers look for meat products with less fat, less salt, and fewer preservative chemicals such as sodium nitrite, without compromising on taste. Research Fellow Ahmed Rady explains: “In light of current enthusiasm to produce and eat more sustainable food, with a reduced carbon footprint associated with livestock and various stages in the meat processing chain, it is important to maintain the highest levels of quality assurance.”

These sustainability challenges have also led to huge investment, research and development efforts in the area of plant-based meat analogues (i.e. meat alternatives). In line with this, developing a deeper understanding of food quality properties will be important for future manufacturing.

Ahmed and the research team see digital tools as an effective way of measuring these properties.

“Digital tools nowadays are ubiquitous, relatively cheap, able to interactively connect with humans, and capable of generating enormous amounts of data that can be used to enhance the quality of products,” he explains.

“In the Digi-Meat project, we are investigating the feasibility of utilising different types of non-invasive sensors and machine learning algorithms to develop reproducible means for monitoring and predicting important quality aspects related to processed meats.”

Beefing up the data

The team applied a range of sensors to scan two major processed meat product typologies, beef burgers and cooked ham. These included optical sensors such as NIR range, RGB vision and hyperspectral imaging; ultrasonic 2D imaging; and dielectric microwave sensors.

Ahmed explains that several factors were deliberately altered to generate the research targets, in order to obtain predictive models that are inclusive and cover a wide range of variability observed in meat processing chains.

To begin with, the researchers are focussing on the primary results of the beef burger modelling, which was conducted during the outgoing phase of the research programme at the University of Nottingham. The studied factors for beef burgers included two mincing levels (coarse and fine), and six fat ratios (5-30%). Beef burgers were analysed to measure fat, moisture and crude protein, as well as water holding capacity, pH and water activity.  

“In addition, several machine learning algorithms were implemented, including one to create classification models for beef burgers based on fat levels,” explains Ahmed. “Some of the algorithms used include Linear Discriminant Analysis (LDA), k-Nearest Neighbour (kNN), Decision Trees (DT), Artificial Neural Networks (ANN) and ensemble methods.”

Initial results revealed that the hyperspectral imaging system tested was effective in differentiating between minced burgers with different fat levels (5-30% in addition to back fat as a reference). “It was clear that both ranges were able to identify different components in the raw burgers with distinct absorption, which are likely to be moisture, lipids and pigments including oxymyoglobin, deoxymyoglobin and metmyoglobin,” says Ahmed.

Coming off these initial positive findings, the team is excited to continue developing their methods. The next step, explains Ahmed, is to develop validated predicted models for the studied attributes, using a wider range of machine learning algorithms.  

“Developing a deeper understanding of food quality [...] will be important for future manufacturing”

Funding

This project was funded by the Research Leaders 2025 Fellowship under the Marie Skłodowska-Curie grant agreement number 754380.

Contributors

Ahmed Rady, Research Leaders 2025 and Marie Skłodowska-Curie Fellow, Teagasc Food Research Centre, Ashtown.

Ruth Hamill, Senior Research Officer, Teagasc Food Research Centre, Ashtown. Email: ruth.hamill@teagasc.ie

Nik Watson, Professor of Artificial Intelligence of Food, University of Leeds.

[pic caption] Optical sensors were used to scan processed meat such as beef burgers

[pic credit] Teagasc