A systematic review and critical analysis of the combination of near-infrared spectroscopy and chemometrics for milk quality assessment

Tirado-Kulieva V.A.; Gonzales-Malca J.A.; Seminario-Sanz R.S.; Castro W. · Applied Food Research (2026) · DOI: 10.1016/j.afres.2026.102372

Resumen

NIR spectroscopy combined with chemometrics constitutes a rapid, non-destructive, and multiparametric alternative for milk quality assessment. This systematic review analyzes research trends in the application of NIR and chemometrics to bovine milk analysis. The review used a hybrid search strategy that combined an initial search in Scopus and Web of Science with backward snowballing. After applying the inclusion and exclusion criteria, a final corpus of 98 articles published between 1999 and 2026 was obtained. China led scientific output in absolute terms, followed by Italy, Spain, and the United States, although the ranking changed when output was normalized by population size, GDP, and bovine milk production. Regression models predominated, particularly PLSR, followed by classification models such as PLS-DA, together with a growing incorporation of machine learning algorithms such as SVM, RF, and ANN. The most frequent spectral preprocessing methods were SNV, DS, MSC, and SG, generally applied in hybrid strategies, whereas PCA and VIP were the most frequently used methods for dimensionality reduction. The main attributes evaluated were fat, protein, lactose, and water content, while classification tasks mainly focused on detecting adulteration with substances such as H2O2, water, NaClO, and urea. Major challenges and future research directions included comparisons between NIR and other spectroscopic techniques, the adoption of portable instruments, calibration transfer and maintenance, data fusion, and deep learning. Overall, the evidence confirms the high potential of NIR spectroscopy coupled with chemometrics for milk quality control; however, its robust implementation at real-world scale still requires further methodological and technological advances.

Instituciones: Universidad Nacional de Frontera