Submitted to: Southern Poultry Science Society Meeting Abstracts
Publication Type: Abstract Only
Publication Acceptance Date: January 17, 2000
Publication Date: January 17, 2000
Citation: WINDHAM, W.R., LAWRENCE, K.C., FELDNER, P.W. PREDICTION OF FAT CONTENT IN POULTRY LEAN MEAT BY NEAR INFRARED TRANSMISSION ANALYSIS. SOUTHERN POULTRY SCIENCE SOCIETY MEETING ABSTRACT. 2000. Technical Abstract: Meat processors need an accurate and reliable means of measuring fat content of deboned raw materials and final products on a production line. This quality control check is needed to verify the lean content of the final product to ensure compliance with consumer specifications. Near infrared transmission (NIT) spectroscopy has been used to analyze the fat composition of beef, pork and lamb. The NIT technique to analyze fat content in poultry lean meat was investigated. Deboned poultry breast muscle, trimmings, and finished products (i.e. chicken nuggets with spices and additives) were collected from a local processing plant over a 4-month period. Fat was measured by the Soxtec solvent extraction method. Fat values ranged from 1.0 to 12.3%. Meat samples were scanned from 850 to 1050 nm with an Infratec 1265 transmission monochromator. Three fat prediction models were developed with 1) local processor samples (N=50), 2) instrument manufacturer supplied poultry meat database (N=87) and 3) the combination of local processor samples and instrument manufacturer supplied database (N= 137). The standard error of cross validation was 0.39%, 0.78% and 0.70% fat for local processor samples, instrument supplied database, and combined data sets, respectively. For validation, an additional 25 local processor samples were predicted with the 3 fat prediction models. The standard error of performance was 0.40%, 0.68%, and 0.50% fat, for equations developed with local processor samples, instrument supplied database, and combined data sets, respectively. This study shows that NIT can be used to predict fat in poultry lean meat and that fat calibrations developed with samples collected from a processing line are more accurate.