Method: Selection of anatomical sites for predicting in vivo body composition in pig and goat using tissue thicknesses measured from computed tomography images

Precise phenotyping of body composition and temporal changes is essential in animal research and breeding, both for genetic selection through better characterisation of complex traits and for adapting breeding strategies to production goals and societal expectations. However, phenotyping in vivo body composition remains a challenge. The aim of this study was to use computed tomography ( CT ) image analysis to identify anatomical proxy sites for predicting in vivo whole-body fat and muscle content and therefore whole-body composition of pigs and goats. Twenty-two crossbred Pietrain × (Large white × Landrace) gilts (99.2 ± 7.5 kg BW, 152.5 ± 0.9 days old) and 20 dairy goats (54.7 ± 6.6 kg BW, 3 ± 0.6 years old) were scanned after anaesthesia using CT. The fat ( FT ) and lean ( LT ) thicknesses were calculated on an axial image within a 15 × 15 cm area on the back of the animals, centred on the spine and in a coronal orientation. The first cranial vertebra was used as a landmark for each vertebra. Linear regressions were calculated to predict CT fat and lean volumes of the whole animal, excluding offal, from either 2 (BW and FT) or 3 (BW, FT, LT) predictive variables. The best anatomical site to predict whole-body fat volume with 2 or 3 predictive variables was the 13th thoracic vertebra (adjusted R 2 : 0.70, coefficient of variation of the RMSE ( CV RMSE ): 6.6%) in pigs and the 1st thoracic vertebra (adjusted R 2 : 0.57, CV RMSE : 22.7%) in goats. The 1st lumbar vertebra when using 3 predictive variables (adjusted R 2 : 0.59, CV RMSE : 3.7%) or the 5th thoracic vertebra when using 2 predictive variables (adjusted R 2 : 0.45, CV RMSE : 4.3%) were considered as the best anatomical sites to predict whole-body lean volume in pigs. In goats, the best anatomical sites were the 5th lumbar vertebra when using 3 predictive variables (adjusted R 2 : 0.30, CV RMSE : 9.1%) or the 8th thoracic vertebra when using 2 predictive variables (adjusted R 2 : 0.22, CV RMSE : 10.0%). The identification of anatomical proxy sites that are easily accessible in live animals and can be routinely measured under farm conditions without the need for restraint or anaesthesia provides a basis for estimating body composition using existing non-invasive, field-applicable methods, and emerging technologies, such as radar-based techniques. The use of these proxy sites opens new opportunities for precision livestock phenotyping in both monogastric mammals and ruminants.