Visible near-infrared hyperspectral imaging as a tool to characterise chicken breasts with myopathies and their durability

Abstract

The surgency of myopathies has posed challenges for the industry as well as researchers, making relevant the use of objective and non-destructive technologies to inspect and discriminate these disorders. In this context, hyperspectral imaging (HSI) provides special properties that allow for an accurate selection of the affected region(s). Two experiments were conducted to evaluate the feasibility of visible and near-infrared (VIS-NIR) HSI to (1) discriminate between myopathies and (2) assess their evolution during refrigerated storage. Hyperspectral images of 98 and 77 chicken breasts, for experiment (1) and (2), respectively, were analysed dividing the breast in 3 regions to precisely assign each one a myopathy or the absence of one. Support vector machine models were employed for classification. Differences between myoglobin content and water binding detected in the VIS-NIR range (386–1016 nm) were relevant enough to accurately discriminate between myopathies (76.1 % accuracy), especially spaghetti meat (94.0 % balanced accuracy). Discrimination was also successful for storage days, detecting spoilage through spectral myoglobin isoform fingerprints (99.3 % accuracy) in the short-wave NIR region (800–1015 nm). These findings suggest a potential industrial use of hyperspectral systems to sort chicken breasts based on myopathy presence by region, and to predict the evolution of their quality traits during refrigerated storage, ultimately tailoring the breast destination for each case and avoiding food waste.