Abstract: Cholesterol, a sterol derived from animal fat, is essential for and possesses important physiological functions in human body. However, unbalanced intake of cholesterol will induce serious cardiovascular and cerebrovascular diseases. A new, rapid and noninvasive method to quantitatively analyze cholesterol in foods of animal origin was developed by near-infrared reflectance spectroscopy (NIRS) within medium wavelength range, and its feasibility was examined. Original spectra of commercially available fresh beef high ribs were collected. The suspicious samples were analyzed and discriminated with Mahalanobis distance and studentized residual, and the outlier samples were determined by twicedetection diagnosis method and removed based on final judgment after secondary partial least squares (PLS) regression analysis. The optimal model for quantitative analysis of cholesterol in fresh beef was established after spectral pretreatment and model optimization. The experiment results indicated that the difference between the predicted values and reference values of the independent test set was not significant (P > 0.05). The overall prediction accuracy could meet the requirements of the Chinese national standard (≤ 10%).

Key words: near infrared reflectance spectroscopy (NIRS), medium wavelength range, fresh beef, cholesterol concentration, quantitative analysis