Using Artificial Intelligence and FT-IR Spectroscopy to Detect Buffalo Milk Adulteration

A recent study conducted by researchers from Zonguldak Bulent Ecevit University in Türkiye aimed to improve food and beverage adulteration analysis methods. This study, published in the Journal of Food Composition and Analysis, investigates how integrating artificial intelligence (AI) algorithms with Fourier Transform Infrared (FT-IR) spectroscopy can improve the detection of adulteration in buffalo milk (1). This study indicates the type of role spectroscopy can play in improving the authenticity of milk products in the global food industry (1).

Water buffalo milk, Manda sütü. Generated with AI. | Image Credit: © mabaci - stock.adobe.com

What is Buffalo Milk and Where is it Produced?

Buffalo milk comes from the mammary glands of buffaloes (2). It is a vital ingredient in high-value dairy products such as buffalo yogurt, and it also has several probiotic effects and other natural health benefits (1,3). Currently, India and Pakistan produce most of the world’s buffalo milk (~80%) (2). Other countries that also produce buffalo milk include China, Egypt, and Nepal (2).

Why is Buffalo Milk Susceptible to Adulteration?

Buffalo milk is especially susceptible to adulteration because of limited availability and higher production costs. In regions where buffalo breeding is not widespread, milk is often sourced from multiple farmers, increasing the risk of unintentional or intentional mixing with lower-cost cow milk (1). Such adulteration not only compromises product quality and consumer trust but also presents economic and health-related concerns (1).

What did the Researchers do to Address this Problem?

In their study, the research team collected a series of buffalo and cow milk samples and created mixtures containing cow milk in concentrations ranging from 0.2% to 10% (v/v) (1). Using FT-IR spectroscopy, they analyzed the mixtures. FT-IR spectroscopy was chosen for this analysis because it is a rapid, non-destructive analytical technique that can analyze chemical compositions (1). Then, the researchers applied six different algorithms. These algorithms include decision trees (DT), k-nearest neighbors (k-NN), support vector machines (SVM), and ensemble bagged trees, to identify adulteration levels based on the spectral data.

What Algorithm Performed the Best?

Out of all six algorithms used, the ensemble bagged trees algorithm performed the best. It achieved a 90.38% accuracy (ACC) when it came to detecting even low-level adulteration (1). The research further compared AI models with traditional chemometric methods, specifically soft independent modeling of class analogy (SIMCA) and data-driven SIMCA (DD-SIMCA), which were also evaluated for classification performance (1).

What Makes This Study Unique?

One of the most unique aspects of the study was the use of particle swarm optimization (PSO). PSO is a nature-inspired computational method that mimics the social behavior of birds or fish to find optimal solutions (1). By applying PSO, the researchers were able to reduce the number of FT-IR measurements needed without negatively impacting accuracy (1). This reduction in measurement requirements significantly boosts the efficiency of the detection process, making it more feasible for real-time industrial applications.

According to the authors, their study demonstrates that buffalo milk adulteration can be adequately analyzed using a combination of FT-IR spectroscopy and AI algorithms (1). Given that existing detection methods are often labor-intensive and expensive, this AI-driven strategy marks a major advancement in food quality assurance (1). The system not only identifies adulteration more accurately than conventional techniques, but it also does so much more quickly than traditional methods.

How Can Future Studies Build On This Study?

Future studies can build on this work. The research team mentioned in their conclusion that they want to expand their data set to include milk samples from a broader range of buffaloes and farming operations. By incorporating more diverse sample sources, especially from different farmers, they hope that they would be able to build an AI-based model that is can be more widely implemented (1).

References

1. Colak, S.; Uzunsoy, I.; Narin, A.; Duran, U. Adulteration Detection of Cow Milk in Buffalo Milk Using Fourier-transform Infrared Spectroscopy and Artificial Intelligence-based Techniques. J. Food Comp. Anal. 2025, 140, 107203. DOI: 10.1016/j.jfca.2025.107203
2. Lang, A. All You Need to Know About Buffalo Milk. Healthline.com. Available at: https://www.healthline.com/nutrition/buffalo-milk (accessed 2025-05-12).
3. Vargas-Ramella, M.; Pateiro, M.; Maggiolino, A.; et al. Buffalo Milk as a Source of Probiotic Functional Products. Microorganisms 2021, 9 (11), 2303. DOI: 10.3390/microorganisms9112303