Realization of High-Accuracy Prediction of Metmyoglobin Content in Frozen Pork by VIS–NIR Spectroscopy Detection Method

• Published in: Food analytical methods 2024-12, Vol.17 (12), p.1668-1677
• Main Authors: Wang, Yi Ming, Cai, Hong Xing, Ren, Yu, Wang, Ting Ting, Wu, Hong Zhang, Hua, Yang Yang, Li, Dong Liang, Liu, Jian Guo, Li, Teng
• Format: Article
• Language: English
• Subjects: Accuracy; Algorithms; Analytical Chemistry; Chemistry; Chemistry and Materials Science; Chemistry/Food Science; Error analysis; Food quality; Food safety; Food Science; Freezing; Frozen food; Frozen foods; Infrared spectroscopy; Microbiology; Modelling; Near infrared radiation; Nondestructive testing; Oxidation; Pork; Prediction models; Public health; Root-mean-square errors; Spectroscopy; Spectrum analysis; Wavelengths
• ISSN: 1936-9751; 1936-976X
• DOI: 10.1007/s12161-024-02686-7

Freezing is a common method to maintain pork quality. However, prolonged frozen storage can cause oxidation reactions of metmyoglobin in pork, resulting in meat quality deterioration. Therefore, it is significant to detect frozen pork quality rapidly and non-destructively for public health and food safety. Metmyoglobin content is considered a critical indicator for evaluating the quality of frozen pork. In this paper, a rapid non-destructive method combining visible and near-infrared (VIS–NIR) spectroscopy technology with chemometrics was applied for the high accuracy ion of metmyoglobin content. First, VIS–NIR spectral data were collected on the pork samples with different freezing times. The raw spectral data were pre-processed using six methods: 1 st derivative, 2 nd derivative, Savitzky-Golay convolutional smoothing, vector normalization, standard normal variate, and multiple scattering corrections. Then, partial least squares (PLS) and random forest (RF) algorithms were applied to establish the prediction models of metmyoglobin content respectively, while the characteristic wavelengths were extracted by combining with the successive projections algorithm (SPA). The results showed significant effects on the prediction accuracy by using different modeling combinations. The MSC-RF-SPA model performed best in prediction, with a coefficient of determination ( R 2 ) of 0.901 and a root mean square error ( RMSE ) of 0.0216, which confirmed the ability to evaluate metmyoglobin content in frozen pork with high accuracy. The results of this study indicated that Vis–NIR spectroscopy technology coupled with MSC-RF-SPA modeling is a promising method, which provided a new way to accurately detect metmyoglobin content in frozen pork.