Determination of crude protein and fat in carcass and breast muscle samples of poultry by near infrared reflectance spectroscopy.

Near infrared reflectance spectroscopy (NIR) equipment used consisted of an InfraAlyzer 400+ fitted with 19 discrete filters and an HP-85 microcomputer. Calibrations were developed with chicken broiler carcass samples. Samples were divided into two sets: a calibration (CAL) from which equations were developed and a prediction (PRE) set used to validate the equations. The CAL and PRE sets in the carcass calibration consisted of 45 and 54 samples, respectively. Two equations were developed for the prediction of fat in carcass samples (FAT1 and FAT2). Calibration for breast muscle samples included 30 and 27 samples for CAL and PRE sets, respectively. The accuracy of the calibrations was measured by the coefficient of determination (r2), standard error of the estimate (SEE), coefficient of variation (CV), and the mean difference between NIR and chemical values (d). Repeatabilities of chemical and NIR procedures were determined by the F-ratio of the variance of the differences between replicates. The result of the regression analysis between chemical and NIR values indicated good predictions for crude protein (CP) and fat in carcass samples. The r2 and SEE for CP, FAT1, and FAT2 were .98, .94; .91, 2.48; and .91, 2.51 in CAL set; and .91, 1.03; .68, 2.32; and .69, 2.30 in PRE set samples, respectively. No differences were found between FAT1 and FAT2 equations. The carcass calibration equations were tested with samples from laying hens. Higher SEE were found for CP (1.78), but SEE for fat were similar to broiler carcasses. The calibration for breast samples showed higher SEE for CP and fat when compared with carcass samples.(ABSTRACT TRUNCATED AT 250 WORDS)