Linear and nonlinear models for assessing carcass composition using dual X-ray absorptiometry in egg- and meat-type chickens

Abstract

The objective of this study was to develop appropriate correction equations for dual energy X-ray absorptiometry (DXA) for total carcass composition of live meat- and egg-type chickens. Linear (bivariate linear and multivariate linear) and nonlinear (polynomial, multivariate polynomial, broken-line and Gompertz) equations were used to estimate carcass composition of DXA-scanned birds based on chemical proximate analysis. A total of 288 laying females (10-30 wk of age) and 305 broiler breeder females (4-32 wk of age) were used. The same birds scanned by DXA were dissected and utilized for whole-body proximate chemical analysis for body lean, fat, and mineral content (ash). As indicators of carcass fat and lean, abdominal fat pad and breast muscle weights were also recorded. Models were evaluated using root mean square error (RMSE), Bayesian Information Criterion (BIC), coefficient of determination (R2), Durbin Watson test for autocorrelation (DW), and residuals observation (RES). Model estimations were done separately by strain or combined. Estimations of composition responses fit at least 1 of each linear and nonlinear models for the egg- and meat-type chickens on all parameters estimated (P < 0.05). In the egg-type chickens, multivariate linear regression was the best fit for body lean with the lowest RMSE and BIC, and highest R2 whereas body fat, body ash, and breast muscle were best predicted by the multivariate polynomial model. In the meat-type chickens, body lean was best predicted by the multivariate linear model with the lowest RMSE and BIC, and the highest R2 whereas the multivariate polynomial was the most parsimonious model for body fat, body ash, and abdominal fat. Positive autocorrelations were observed in several models tested for body fat, body ash, breast muscle, and abdominal fat pad when both strains were analyzed combined (P < 0.05). In summary, a strain-based correction is recommended to all the parameters, with exception of the BW estimation. Correction equations developed in this study demonstrated that the DXA technique is a reliable alternative to proximate chemical analysis in egg- and meat-type chickens.