Prediction of pork belly cooking loss by hyperspectral imaging analysis of semispinalis capitis muscle

This study evaluated the performance of a partial least squares regression (PLSR) model in predicting the cooking loss of pork belly slices using hyperspectral images of the semispinalis capitis (SC) muscle from the pork shoulder butt. Pork shoulder butts and whole belly cuts were obtained from 70 half-right gilt carcasses at 24 h postmortem. Pork belly slices were obtained from the 6th and 11th thoracic vertebrae and the 4th lumbar vertebra. Reflectance hyperspectral images of the SC muscle were captured. Cooking loss was measured in SC muscle and pork belly slices. A significant correlation was observed between the cooking loss of pork belly slices and SC muscle. In the PLSR model for predicting the cooking loss of belly slices, standard normalized variate pre-processing and wavelength selection did not enhance the model accuracy. In contrast, first-derivative pre-processing showed the highest accuracy with R²v values ranging from 0.73 to 0.80. These results suggest that the cooking loss of pork belly slices can be nondestructively predicted using the hyperspectral image of SC muscle exposed on the half-carcass surface. Moreover, this study can provide a foundation for future research on nondestructive and real-time prediction of pork quality in production lines.