CLASSIFICATION OF TEXTURIZED VEGETABLE PROTEIN INGREDIENTS USING HYPERSPECTRAL AND MULTISPECTRAL IMAGING AND CHEMOMETRICS: DETECTION OF ALLERGENIC SUNFLOWER PROTEINS

In recent years, plant-based meat-like products (PBM) have gained prominence as sustainable alternatives to animal meat. PBM main ingredients are texturized vegetable proteins (TVP) that can derive from different sources. Some of them, like soy proteins have allergenic potential. In this study, two imaging systems were employed: a hyperspectral camera (900–1600 nm), which provides a detailed characterization of the optical and chemical properties of the samples, and a multispectral camera (340–1100 nm), which enables faster and more efficient data acquisition but with lower spectral resolution. Classification models were developed using Soft Independent Modeling of Class Analogy (SIMCA) and Partial Least Squares Discriminant Analysis (PLS-DA) to differentiate different types of TVP, with an emphasis on those made from sunflower seed protein due to its potential allergenicity (Taylor et al., 2021). Two classes of TVP were considered: (1) sunflower and (2) a combination of soy and pea, both dry and rehydrated. SIMCA and PLSDA models, built using hyperspectral data, showed error rates of 0% and 4-5%, respectively. In validation, SIMCA demonstrated high sensitivity, specificity, and precision, achieving 100% accuracy in the classification of both dry and wet TVP using multispectral images. These results highlight the potential of spectroscopy, combined with chemometric tools, for the efficient classification of TVP. The application of this approach contributes to the development of quality control and food safety systems, providing key information for the industry and consumer protection. Taylor, S. L., Marsh, J. T., Koppelman, S. J., Kabourek, J. L., Johnson, P. E., & Baumert, J. L. (2021). A perspective on pea allergy and pea allergens. Trends in Food Science & Technology, 116, 186–198. https://doi.org/10.1016/j.tifs.2021.07.017