Detection of Chicken Meat Adulteration in Cattle Meat Using BME688 Gas Sensor and Convolutional Neural Networks

Food fraud is a persistent shadow over the global food supply, where expensive ingredients are quietly swapped for cheaper ones to boost profits. This practice does more than just deceive consumers out of their money; it can trigger allergic reactions, violate religious dietary laws, and erode trust in what we eat. For decades, catching these substitutions has required sending samples to a laboratory, where scientists use complex, expensive machines to analyze DNA or proteins. While accurate, these methods are slow, require specialized training, and cannot be used right at the grocery store or slaughterhouse to check a batch of meat the moment it arrives. The challenge has been finding a way to detect fraud quickly and cheaply without sacrificing accuracy, ideally using a device small enough to hold in one hand.

A team of researchers at Bolu Abant Izzet Baysal University in Turkey has taken a significant step toward solving this problem by developing a system that "smells" meat to tell if it is pure or mixed. Their work focuses on a specific and common type of fraud: mixing cheaper chicken meat into more expensive beef. Instead of looking for DNA, their device detects the invisible cloud of gases that naturally rises from the meat. All fresh meat releases a unique mix of volatile organic compounds, which are essentially tiny chemical molecules that float into the air. Different animal species produce different chemical signatures, much like different people have different scents. The researchers built a compact electronic nose, a device equipped with a gas sensor that changes its electrical resistance when it encounters these specific airborne chemicals. By combining this sensor with a type of artificial intelligence known as a convolutional neural network, they created a system that can learn to recognize the distinct chemical fingerprint of pure beef versus beef that has been diluted with chicken.

To test their idea, the researchers prepared a series of meat mixtures in a controlled laboratory setting. They took ground beef and ground chicken, both of the same freshness, and blended them together in precise ratios. Some samples were pure beef, while others contained 10%, 20%, 30%, or 50% chicken. They placed these meat patties into sealed glass cups and positioned their sensor array just above the surface. The sensor used in this study, the BME688, is a small, low-cost chip capable of measuring gas, temperature, and humidity. To get the most detailed reading possible, the researchers did not just let the sensor sit still; they programmed it to heat up and cool down in four different patterns. This temperature modulation helps the sensor react differently to various chemicals, allowing it to capture a richer and more detailed picture of the gas cloud rising from the meat. They ran two sensors simultaneously: one measured the air above the meat, and the other measured the clean air of the room, allowing them to subtract the background noise and isolate the scent of the meat itself.

Once the data was collected, the researchers fed it into their artificial intelligence model. This model was designed to do two things at once. First, it acted as a classifier, trying to sort the samples into the correct category, such as "pure beef" or "beef with 20% chicken." Second, it acted as a regressor, attempting to guess the exact percentage of chicken in the mix. The results were remarkably precise. For the task of simply identifying which category a sample belonged to, the system achieved perfect accuracy on its test data, correctly distinguishing every single mixture regardless of which heating pattern the sensor used. This means the chemical differences between pure beef and beef mixed with chicken are distinct enough for the machine to spot them every time.

The system also proved highly effective at estimating exactly how much chicken was present. The best-performing model, which used a specific heating pattern called HP-412, could predict the amount of chicken with an average error of only about 1.5 percentage points. In practical terms, if the meat contained 20% chicken, the device would likely estimate it to be between 18.5% and 21.5%. This level of precision is significant because it moves beyond simply detecting that fraud exists to quantifying exactly how much of it is there, which is crucial for regulators and food safety inspectors. The study confirms that a low-cost sensor paired with smart software can perform complex chemical analysis without the need for a laboratory.

However, the researchers are careful to note the boundaries of their findings. The experiments were conducted under strict laboratory conditions using fresh meat that was prepared and tested immediately. The system has not yet been tested on meat from different breeds of cattle, meat that has been aged for weeks, or samples that have been frozen and thawed. These real-world variables could change the chemical signature of the meat and potentially confuse the sensor. The authors suggest that while their prototype is a powerful proof of concept, future work must involve testing on a much wider variety of samples collected over time to ensure the system works reliably in a busy market or processing plant. They also plan to explore integrating this technology into portable, handheld devices that could be used directly in the field.

This research highlights a shift in how food safety might be monitored in the future. Rather than relying solely on slow, expensive lab tests, the industry may soon have access to rapid, non-destructive tools that can screen meat for authenticity in real time. By proving that a small, inexpensive sensor can distinguish between pure beef and beef mixed with chicken with high accuracy, the study opens the door for a new generation of food safety technologies. These tools could eventually help protect consumers, ensure fair trade, and maintain the integrity of the food supply chain, all by listening to the subtle chemical whispers of the food itself.