Small Ruminant Research 136 (2016) 173–178 Contents lists available at ScienceDirect Small Ruminant Research journal h om epa ge: www.elsevier.com/locate/smallrumres Short communication Ewe whole body composition predicted in vivo by real-time ultrasonography and image analysis S.R. Silva a,∗ , J. Afonso b , C.M. Guedes a , M.J. Gomes a , V.A. Santos a , J.M.T. Azevedo a , A. Dias-da-Silva a a CECAV, Universidade de Trás-os-Montes e Alto Douro, 5001-801 Vila Real, Portugal b CIISA, FMV, ULisboa, Avenida da Universidade Técnica, 1300-477 Lisboa, Portugal a r t i c l e i n f o Article history: Received 26 June 2015 Received in revised form 12 January 2016 Accepted 29 January 2016 Available online 1 February 2016 Keywords: Mature ewes Chemical body composition Real-time ultrasound a b s t r a c t The relationship between ultrasound measurements and the empty body chemical composition of mature ewes was studied in two breeds. The breeds were a milk-producing breed Churra da Terra Quente—CTQ (n = 33; live weight 42.0 ± 7.3 kg, mean ± SD), and a meat breed Ile de France—IF (n = 23; live weight 60.7 ± 9.1 kg, mean ± SD). Fat and muscle depths were measured in the live animals by real-time ultra- sound scanning (RTU; 7.5 MHz probe) over the 13th thoracic, and between the 3rd and 4th lumbar vertebrae, and total tissue depth over the 11th rib. Following slaughter, the carcass and non-carcass components of the empty body were combined and subjected to chemical analysis and assessment of energy value. Data obtained by RTU after image analysis was used to develop simple and multiple regres- sion models for each breed. The traits most accurately estimated from single RTU measurements were the absolute values for fat content and energy value of the empty body. The respective coefficients of determination (R 2 ) for IF and CTQ ewes using subcutaneous fat depth over the 13th thoracic vertebra were 0.768 and 0.908 for fat, 0.821 and 0.900 for energy; and for measurements between the 3rd and 4th lumbar vertebrae were 0.845 and 0.911 for fat, 0.852 and 0.906 for energy. All these coefficients were sig- nificant (P < 0.01). The best prediction models included one to three RTU measurements and were better for CTQ ewes, where adjusted R 2 values ranged from 0.923 (P < 0.001) for water to 0.969 (P < 0.001) for protein, than for IF ewes, where the range was 0.394 (P < 0.01) for protein to 0.940 (P < 0.001) for fat. The results revealed good estimates of the fat and energy content of the empty body of both breeds and also good estimates of protein content for CTQ ewes, but poor estimates of protein content for IF ewes, with the best prediction models for each body component being different from each breed. Consequently, it is concluded that predictive models that are specific to the breed and circumstances of the study in which they are to be used will have to be established to have a practical application. © 2016 Elsevier B.V. All rights reserved. 1. Introduction The knowledge of chemical body composition is fundamental to study the effect of nutritional and genetic factors on livestock pro- duction. Body composition studies in meat animals have focused primarily on those changes that occur during the growth of the ani- mal up to “market weight”, but also on changes that occur during pregnancy and lactation (Mitchell, 2007). Even today, dissection and chemical analysis represent the standard for body compo- sition determination of farm animals. However, such techniques raise several problems (e.g., ethical problems, the impossibility of ∗ Corresponding author. Fax: +351 259350482. E-mail address: ssilva@utad.pt (S.R. Silva). within animal comparisons and need of large numbers of animals, very intensive labor and high cost). To overcome these constraints several in vivo image methodologies have been proposed to get accurate estimates of chemical body composition, such as com- puter tomography (CT), magnetic resonance imaging, dual-energy X-ray absorptiometry or real-time ultrasound (RTU). These tech- niques have been reviewed by several authors (Silva and Cadavez, 2012; Scholz et al., 2015). The RTU scanning is already largely used in animal production and is considerably less expensive than other imaging techniques. Concerning the estimation of dissectible carcass components, many studies have been made based on ultra- sonic measurements and the effectiveness of regression models for prediction of carcass composition is now well established, partic- ularly for growing animals (Ripoll et al., 2009; Grill et al., 2015). Concerning the prediction of chemical body composition the data http://dx.doi.org/10.1016/j.smallrumres.2016.01.024 0921-4488/© 2016 Elsevier B.V. All rights reserved.