Abstract: 【Objective】Endpoint temperature (EPT) of heat treated meat products is a determinant factor to control food-borne diseases. EPT of cooked meat and meat products has been determined using different methods including determination of enzyme activity, coagulation test and Sodium Dodecyl Sulfate-Polyacrylamide Gel Electrophoresis (SDS-PAGE) etc. However, these methods are time consuming, destructive and labor intensive. The objective of this study was to investigate if near-infrared spectroscopy (NIR) in combination with partial least squares (PLS) data analysis can be used to determine the EPT in cooked meat.【Method】The meat samples (chicken and fish) were heat treated until they reached a final internal temperature of 50 to 90℃ (5℃ intervals) at the rate of 1℃·min^-1. The cooked samples were removed from the waterbath and cooled immediately in ice water to 4℃. The 144 samples in total (the number of chicken and fish was 77 and 67 respectively) were homogenized for 1 minute and deposited at 4℃ before measurement. Then NIR reflectance spectra were collected using a scanning monochromator equipped with a rotating sample cup with a quartz window between 11 000 and 4 000 cm^-1 at 8 cm^-1 intervals. Each sample was scanned for three times. Each spectrum, an average of 64 scans, was recorded over the selected wavenumber range. Randomly selected 108 spectra from chicken and fish samples were assigned to a calibration set while remaining 36 spectra were used for a validation set. The calibration sets were used for establishing the calibration model and the validation sets were used to test model prediction ability.The spectra was processed using the method of Standard Normal Variate, first-order differential and Norris Derivative. The principal component number was determined by the cross-validation mean square error (RMSECV). The prediction mean square error (RMSEP), correlation index (r) and the standard deviation (σ) were used to evaluate prediction ability of the calibration model.【Result】By RMSECV, the principal components of the chicken and fish were 9 and 11 ascertained through RMSECV value (1.59% and 0.96%) and r value (0.9844 and 0.9936). The lower the RMSECV and lower the r values, the more accurate is the calibration model. As can be observed in the study, prediction of EPT using the NIR spectra showed very high correlation (r=0.9966 for chicken and r=0.9832 for fish) with a prediction error (RMSECV) of 3.02% and 2.94%, respectively.【Conclusion】The results showed that NIR spectroscopy has the potential for use in the food processing industry and food inspection to ensure that all meat products have reached the recommended EPT, keeping it safe. And the model established in this paper has a good prediction ability of EPT for meat products, which is a non-destructive, simple and reliable technique to monitor the ?nal quality of a product.

Key words: chicken, fish, near-infrared spectroscopy, partial least squares, endpoint temperature