Classification of shear force and cooking loss of pork tenderloin using the hyperspectral imaging analysis

This study was conducted to investigate whether the shear force (SF) and cooking loss (CL) of pork tenderloin can be non-destructively predicted using hyperspectral image (HSI) analysis. A total of 100 tenderloins from different pork carcasses were obtained at 24 h postmortem. Each tenderloin was sliced into 3 slices of 6 cm thickness, resulting in 300 slices in total. HSI analysis was performed in reflectance mode within the spectral rage of 402-1002 nm (151 bands). The SF and CL of each slice were subsequently measured. The tenderloin samples were categorized into three SF and CL classes (low, middle, and high SF or CL). The 70% and 30% of the HSI data were randomly assigned to the calibration set and the prediction set, respectively. Partial least squares discriminant analysis (PLS-DA) was employed to classify the sample into the three SF and CL categories. The highest accuracy of the calibration models of SF and CL was 74.29% and 70.48%, respectively. In contrast, the highest accuracy of prediction models of SF and CL was 53.33% and 55.56%, respectively. Therefore, it is considered that there are limitations in predicting SF and CL of tenderloin using HSI in the Vis/NIR range employed in this study.