Applying Raman Spectroscopy in Dairy Analysis

Advancements in technology can lead to increased accuracy in detecting food fraud, according to a recent study published in the Journal of Food Composition and Analysis (1). A research team, comprised of scientists from China University of Mining and Technology and led by Lina Zheng, proposed a new method for identifying fraudulent food products using Raman spectroscopic technology. The results they achieved indicated that this new method improved the detection of inauthentic food products.

Numerous studies have explored using spectroscopic techniques to detect food adulteration because these techniques are nondestructive and are able to capture the spectral information on the products under study quickly and efficiently (2,3). In this study, Lina Zheng from the China University of Mining and Technology in Xuzhou proposed and tested an alternative method for detecting food adulteration in dairy products. Their approach involved using an advanced fusion model based on Raman spectroscopy technology to quantitatively evaluate the nutritional components, particularly the fat content, of dairy products (1).

Various fresh dairy products | Image Credit: © baibaz - stock.adobe.com

The fusion model integrated four robust machine learning algorithms: light gradient boosting machine (LightGBM), support vector machine (SVM), random forest (RF), and extreme gradient boosting (XGBoost) (1). For each algorithm, Light GBM, SVM, RF, and XGBoost, they demonstrated a classification accuracy exceeding 90% for differentiating dairy brands and achieved a fat prediction normalized root mean square error (NRMSE) under 13% (1). The researchers combined these algorithms in their fusion model with an eye on improving overall performance.

The research did indeed accomplish that. According to the results published, the research team cited a classification accuracy of 99% (1). The fusion model was able to predict the fat content in dairy samples with an R² value of 0.98, a root mean square error (RMSE) of 0.2 g/100 mL, and an NRMSE of 6.2% (1).

The study also involved the classification of five different brands of liquid dairy products. Using principal component analysis (PCA), the researchers visualized differences in the Raman spectra of the various dairy products (1). Then, the competitive adaptive reweighted sampling (CARS) algorithm extracted Raman shifts with high classification contributions (1). The performance of the four basic models (LightGBM, SVM, XGBoost, and RF) and the fusion model was evaluated based on these characteristic Raman spectral bands (1).

Examining eight different types of dairy products, the research team showed through quantitative analysis that although individual models exhibited variability in their efficiency depending on the training objectives, the fusion model consistently outperformed them in all scenarios (1).

All these experiments revealed one thing: that the fusion model can consistently and accurately classify food products, which makes it a beneficial technological innovation for regulatory agencies and food manufacturers. For regulatory agencies, they can use this fusion model to enforce food safety regulations (1). Meanwhile, manufacturers can leverage this technology to maintain stringent quality control standards (1).

Zheng and the team strategically combined the four machine learning algorithms with Raman spectroscopy to produce a new technology capable of improving accuracy in detecting food adulteration.

References

(1) Feng, Z.; Liu, D.; Gu, J.; Zheng, L. Raman Spectroscopy and Fusion Machine Learning Algorithm: A Novel Approach to Identify Dairy Fraud. J. Food Comp. Anal. 2024, 129, 106090. DOI: 10.1016/j.jfca.2024.106090

(2) Saji, R.; Ramani, A.; Gandhi, K.; et al. Application of FTIR Spectroscopy in Dairy Products: A Systematic Review. Food and Humanity 2024, 2, 100239. DOI: 10.1016/j.foohum.2024.100239

(3) Fun, Y.; Ren, Y.; Sun, D.-W. Novel Analysis of Food Processes by Terahertz Spectral Imaging: A Review of Recent Research Findings. Trends Food Sci. Technol. 2024, 147, 104463. DOI: 10.1016/j.tifs.2024.104463