Evaluation of carcass-grading systems in swine
Abstract
Five carcass-grading methodologies were examined on 154 pigs representing 7 genotypes and 2 sexes to determine the accuracy in estimating carcass lean content. Fat depth measurements taken laterally to the dorsal midline were more accurate predictors of carcass lean weight than those taken at the dorsal midline. Equations including 10th rib loin muscle area, 10th rib fat depth and carcass weight were subject to genotype bias in predicting carcass lean content. A combination of dissected ham lean, carcass weight and 10th rib fat depth improved predictive accuracy of carcass fat-free lean (R$\sp2$ = 0.91, RSD = 1.64 kg). Dissected Boston butt and loin leans produced the least-biased estimates. Hennessy and Destron probe measurements at the 3/4 last rib performed best in predicting fat-free lean with R$\sp2$ values of 0.85, and 0.83 and RSD values of 2.18 kg and 2.26 kg, respectively. The least-biased equation was obtained at the 3/4 last rib using the Hennessy probe or at the 10th rib with the Destron probe. Bioelectrical impedance measurements derived from warm and chilled carcasses were subject to sex and genotype biases. The best 4-variable model including carcass weight, resistance, reactance and Vol$\sb1$ predicted carcass fat-free lean with a R$\sp2$ of 0.77 and RSD of 2.69 kg. TOBEC index from various carcass formats had high correlations with carcass lean (coefficients ranged from 0.83 to 0.94). TOBEC index from slow conveyor speed (2.1 ft/sec) had higher predictive accuracy than from fast conveyor speed (3.8 ft/sec). Inclusion of fat depth measurement and carcass weight, in addition to TOBEC index, alleviated some biasing effects. TOBEC index had identical accuracy as the best parameter(s) from phase curve in predicting dissected lean in the primal cuts. Combining fat depth measurement with TOBEC index further reduced prediction biases. Neural networks based on ham phase reading, ham weight, and fat depth accurately classified hams into 4 categories. It also categorized 86 carcass sides into corresponding classes expressed as percentage of carcass lean. Integration of neural networks with carcass-grading systems showed strong potential in improving classification accuracy.
Degree
Ph.D.
Advisors
Forrest, Purdue University.
Subject Area
Agriculture|Food science|Mathematics
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.