Optical tongue based on Mach–Zehnder interferometers and automated machine learning for spoilage microorganisms assessment in ricotta cheese

Ensuring the microbiological quality of fresh cheeses remains a critical challenge for the dairy industry due to the susceptibility of high-moisture products to spoilage microorganisms and the limited suitability of conventional analytical methods for rapid monitoring. This work proposes an optical sensing approach based on Mach–Zehnder interferometers implemented with micro-tapered long-period fiber gratings (MT-LPFGs) as an electronic tongue for assessing microbiological contamination in ricotta cheese. The sensing principle relies on monitoring variations in the surrounding refractive index (RI) induced by physicochemical changes resulting from microbial metabolic activity, which modulate the interferometric phase response and the subsequent spectral transmission profile. Samples were prepared with controlled concentrations of Staphylococcus spp. as a target indicator of microbial spoilage, obtained through serial dilutions. The framework predicts contamination levels ( CFU/g) by processing these complex, nonlinear spectral features using an Automated Machine Learning (AutoML) framework. Multiple learning algorithms, including neural networks, ensemble-based methods, and instance-based models, were systematically trained and evaluated using complementary error and correlation metrics. A multi-objective Pareto-based analysis was employed to jointly consider predictive accuracy and computational efficiency. The results demonstrate that the proposed framework enables accurate estimation of contamination levels, with the NeuralNetTorch model achieving a correlation coefficient ( ) of 0.976 and a limit of detection (LOD) of 1.62 CFU/g. Overall, the combination of interferometric fiber-optic sensing and AutoML offers a rapid, non-destructive, and robust solution for microbiological quality monitoring of fresh cheeses, with strong potential for real-time industrial applications.