Comparison of data analytics strategies in computer vision systems to predict pig body composition traits from 3D images.

Computer vision systems (CVS) have been shown to be a powerful tool for the measurement of live pig body weight (BW) with no animal stress. With advances in precision farming, it is now possible to evaluate the growth performance of individual pigs more accurately. However, important traits such as muscle and fat deposition can still be evaluated only via ultrasound, computed tomography, or dual-energy x-ray absorptiometry. Therefore, the objectives of this study were: 1) to develop a CVS for prediction of live BW, muscle depth (MD), and back fat (BF) from top view 3D images of finishing pigs and 2) to compare the predictive ability of different approaches, such as traditional multiple linear regression, partial least squares, and machine learning techniques, including elastic networks, artificial neural networks, and deep learning (DL). A dataset containing over 12,000 images from 557 finishing pigs (average BW of 120 ± 12 kg) was split into training and testing sets using a 5-fold cross-validation (CV) technique so that 80% and 20% of the dataset were used for training and testing in each fold. Several image features, such as volume, area, length, widths, heights, polar image descriptors, and polar Fourier transforms, were extracted from the images and used as predictor variables in the different approaches evaluated. In addition, DL image encoders that take raw 3D images as input were also tested. This latter method achieved the best overall performance, with the lowest mean absolute scaled error (MASE) and root mean square error for all traits, and the highest predictive squared correlation (R2). The median predicted MASE achieved by this method was 2.69, 5.02, and 13.56, and R2 of 0.86, 0.50, and 0.45, for BW, MD, and BF, respectively. In conclusion, it was demonstrated that it is possible to successfully predict BW, MD, and BF via CVS on a fully automated setting using 3D images collected in farm conditions. Moreover, DL algorithms simplified and optimized the data analytics workflow, with raw 3D images used as direct inputs, without requiring prior image processing.

Mixed effects structural equation models and phenotypic causal networks.

Complex networks with causal relationships among variables are pervasive in biology. Their study, however, requires special modeling approaches. Structural equation models (SEM) allow the representation of causal mechanisms among phenotypic traits and inferring the magnitude of causal relationships. This information is important not only in understanding how variables relate to each other in a biological system, but also to predict how this system reacts under external interventions which are common in fields related to health and food production. Nevertheless, fitting a SEM requires defining a priori the causal structure among traits, which is the qualitative information that describes how traits are causally related to each other. Here, we present directions for the applications of SEM to investigate a system of phenotypic traits after searching for causal structures among them. The search may be performed under confounding effects exerted by genetic correlations.

Predição do peso vivo a partir de medidas corporais em animais mestiços Holandês/Gir / Prediction of live weight based on body measurements in crossbred animals Holstein/Gir

O objetivo deste estudo foi investigar as relações entre o peso corporal e as medidas corporais altura de garupa (ag), comprimento de garupa (cg), comprimento corporal (cc) e perímetro torácico (pt), em bovinos oriundos principalmente do cruzamento das raças Holandês e Gir. Foram utilizados dados de 483 vacas, 469 novilhas e 62 machos, de três rebanhos distintos, analisados separadamente para cada categoria a fim de estabelecer equações polinomiais dos pesos em relação às medidas corporais. As correlações simples do peso corporal com pt, cc, cg e ag foram respectivamente 0,807; 0,440; 0,187 e 0,504 para vacas; 0,928; 0,735; 0,819 e 0,880 machos, e 0,942; 0,748; 0,902 e 0,573 para novilhas. Embora as regressões de peso corporal em relação ao cc e cg tenham sido significativas (P<0,05), o aumento da porcentagem de explicação das variações do peso corporal obtido com a inclusão destas medidas, em adição ao pt, não parece justificar o custo das medições. As equações de predição do peso corporal em função do pt foram as seguintes: para vacas, peso = 12.174 - 187,410 pt + 0,97196960 pt² - 0,00162382 pt³, para novilhas, peso= 1.717-35,167 pt + 0,238978 pt² - 0,00046260 pt³ e, para machos, peso = -3.862+76,014 pt-0,488837 pt²+ 0,00109755 pt³.

The objective of this study was to investigate the relationship between hip height (ag), rump length (cg), body length (cc) and heart girth (pt) with live body weight of crossbred animals, mainly from the cross between Holstein and Gir breeds. Data on 483 cows, 469 heifers and 62 males in three herds were analyzed for each category using polynomial regression equations of body weight on measurements. The correlations between body weight and pt, cc, cg and ag were, respectively 0.807, 0.440, 0.187 and 0.504 for cows, 0.928, 0.735, 0.819 and 0.880 for males and 0.942, 0.748, 0.902 and 0.573 for heifers. Although the regressions of body weights on cc and cg were significant (P<0.05), the additional goodness of fit of a model that includes these two traits in addition to heart girth does not justify the extra cost for recording these traits. The prediction equations were: for cows, body weight = 12174-187.410 pt+0.97196960 pt²-0.00162382 pt³, for heifers, body weight = 1717-35.167pt+0.238978pt²-0.00046260pt³ and for males, body weight = -3862+76.014pt-0.488837pt²+0.00109755pt³.