Prediction of carcass composition in Atlantic salmon by computerized tomography

Abstract

The prediction of carcass composition of Atlantic salmon (Salmo salar) using computerized tomography (CT) data was examined. 174 fish were included in the study, and prediction equations derived by multiple linear regression (MLR) and principal component regression (PCR) were evaluated. Validation of alternative equations was done according to the relative standard error of prediction (SEP/SD), relative mean prediction error (BIAS/SD) and the correlation between chemically determined and predicted values. The potential of the CT technique for accurate and fairly unbiased predictions of percentage fat and dry matter in salmon carcasses was clearly demonstrated. SEP/SD of about 0.40, BIAS/SD less than 0.10 percentage units and a correlation between observed and predicted values of 0.92 for both compositional components were obtained for the best prediction equations. A significant prediction of protein content was not obtained.The accuracy of predicting fat and dry matter content was improved by 38% by using CT data instead of the simple carcass measurements gutted body weight and condition factor.The impact of the results on a breeding program for quality-related traits in farmed Atlantic salmon is discussed.