Prediction of starch content in meatballs using near infrared spectroscopy (NIRS)

Meatballs are a popular food in Asian countries. A good quality consists of low starch. In this study, the quality of meatballs was evaluated by starch content using short and long-wavelength near infrared spectroscopy (NIRS). The result found that long-wavelength NIRS can predict starch contents in all kinds of meatballs. The model of beef meatballs showed a high coefficient of multiple determination of validation set (R2-val) of 0.97 and a low standard error of crossvalidation (SECV) of 2.64%; the chicken meatballs model had an R2-val of 0.97 and a SECV of 2.63%; and the pork meatballs model had an R2-val of 0.98 and a SECV of 2.37%. In addition, a universal model was created by combining the spectra of all meatballs. The universal model had an coefficient of multiple determination of calibration set (R2-cal) of 0.98, standard error of calibration (SEC) of 2.22%, R2-val of 0.97, standard error of prediction (SEP) of 2.67% and Bias of 0.05%. The results indicated that NIRS can predict starch contents with high accuracy and could apply for quality classification via rapid analysis.