Phenomics

Genetic Parameters for Image-based, Growth and Carcass Traits in Fattening Pigs

---

Abstract

Improving carcass yield and prime cut weights is crucial for the profitability and sustainability of the pork industry. Accordingly, modern breeding programs have placed emphasis on carcass and meat quality traits. However, assessing carcass performance typically requires slaughter, restricting phenotypic data to progeny and lowering the accuracy of breeding value estimation for selection candidates. To address this challenge, this study extracted and evaluated digital morphological traits of pigs from multiple imaging sources. Data were collected from 1,248 pigs aged 169-176 days at a commercial finishing farm in North Carolina. During pre-slaughter management, an Intel RealSense D435 Depth Camera connected to edge computing device was used to automatically collect images over an electronic scale, where animals were weighed, IDs captured via RFID ear tags, and tissue samples taken for genotyping using the Illumina 50k SNP chip. Hot carcass weight (HCW), Belly weight (BLW), and Loin weight (LW) were collected at the associated packing plant. YOLOv8 deep learning models were trained to remove low-quality images and produce 2D segmentation masks, while point clouds were built from raw depth matrices. The final dataset included 9,936 depth images, their corresponding masks, and point clouds, from which 154 statistical and morphological features were extracted. Four digital morphometric features (DMF) from depth images, three from 2D masks, and three from point clouds were selected based on low feature collinearity and high Pearson correlation with True Live Body Weight (TBW). Predicted Live Body Weight (PBW) values were obtained from a linear regression model trained with selected DMF. A Leave-One-Out cross-validation scheme was used to avoid including overfitted PBW values in the genetic analyses. The regression model for TBW achieved R² of 0.528 and mean absolute error of 6.53 kg, while the correlation between TBW and PBW was 0.728. Genetic parameters between traits were estimated using single-step GBLUP bi-trait models in Blupf90+ software. TBW and carcass traits showed heritability (h²) consistent with the literature (TBW: 0.355 ±0.052; HCW: 0.216 ±0.067; BLW: 0.62 ±0.098; LW: 0.39 ±0.12). Image-based traits exhibited moderate potential for direct genetic selection, with h² ranging from 0.185 ±0.059 to 0.3 ±0.07 for DMF and from 0.215 ±0.06 to 0.34 ±0.07 for PBW. TBW showed a high genetic correlation (rg) with PBW (0.70 ±0.1) and DMF (−0.57 to 0.696) (Table 1). Similar patterns were observed for the rg estimated between DMF and HCW (−0.59 ±0.2 to 0.545 ±0.21) or BLW (−0.79 ±0.13 to 0.046 ±0.25) (Table 1). Image-based traits showed moderate heritability and strong genetic associations with TBW and carcass weights, supporting their use to support indirect selection for carcass merit. These traits can be collected automatically and noninvasively, capturing morphological variation beyond body mass and enabling large-scale, high-frequency phenotyping in commercial settings.

Keywords: 2026

How to Cite:

Medeiros, B., Saha, C., Abdulrahman, T., Costa, R., Leach, R., Huang, Y., Botero, M. & Alves, A., (2026) “Genetic Parameters for Image-based, Growth and Carcass Traits in Fattening Pigs”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2286130. doi: https://doi.org/10.31274/wcgalp.23858

Rights: 1