Computerised tomography of chickens

1. The use of computerised tomography (CT) to predict body composition of chickens has been evaluated. Prediction equations were constructed by regression analyses with CT-observations as independent variables. The accuracy of prediction was evaluated on separate sets of birds, not included in the regression analyses. 2. Different methods of generating CT-observations were evaluated. The most robust CT-variables for prediction purposes were obtained from a principal component analysis of the distribution of CT-values. The accuracy of prediction depended on the traits and the material analysed. 3. The correlation between observed and predicted values for the amount of abdominal fat or breast cut in grams were in the range of 0.70 and 0.90 where the principal component method was applied. The bias was 1 to 2% of the standard deviation for the trait. 4. The amount of abdominal fat relative to carcase weight was predicted with similar accuracy, whereas the relative amount of breast cut seemed to be more difficult to predict. The increase in accuracy of prediction for amount of abdominal fat was considerable when compared to an equation including only body weight and sex. 5. The results from a small scale experiment, where a wide range of compositional traits were determined by more laborious and precise methods, indicated that the accuracy of prediction may be increased further and that CT-observations may in addition provide important information about the chemical composition of the tissues analysed.