Phenomics

Integrative analysis of leg health in turkeys using meta-analysis and deep learning approaches

---

Abstract

Genetic selection and advancements in breeding programs have significantly improved turkey growth rates, muscle yield and overall profitability. However, these gains have been accompanied by an increased incidence of leg health disorders, notably tibial dyschondroplasia (TD), a skeletal condition associated with lameness, angular bone deformities, fractures and elevated pre-slaughter mortality. The first objective of this project included a meta-analysis of 62 experiments to identify factors influencing TD incidence in meat-type poultry. Results from linear mixed models revealed a significant relationship (0.13 ± 0.028; P m) and walking score (WS; scored from 1 = clear abnormalities to 6 = no abnormalities). Additionally, CT scans were used to estimate breast muscle yield (BMY) and CT-derived body weight (BWCT). For model training purposes, manual annotation of 64 CT scan slices of the left (LL) and right (RL) tibiotarsi were conducted on a random subset of 15 birds. A panoptic segmentation model, developed in Python (v3.1) and trained using Detectron2 with a transfer learning approach, quantified white (cortical bone) and black (medullary cavity) pixel counts to calculate cortical bone area (Ct.Ar). Due to the panoptic segmentation model achieving a higher evaluation performance for RL, subsequent analyses and interpretations focused exclusively on this leg. A multi-trait animal model was used to estimate heritability of Ct.Ar RL at 0.24 ± 0.04 (line A) and 0.33 ± 0.04 (line B). Favourable genetic correlations between Ct.Ar RL and BW traits ranged from 0.14 ± 0.06 to 0.34 ± 0.06 across both lines. However, unfavorable genetic correlations were observed between Ct.Ar RL and BMY (line A: -0.22 ± 0.10; line B: -0.13 ± 0.09). These results highlight the potential of Ct.Ar as a selection criterion for enhancing leg health in turkeys, and emphasize the value of incorporating in vivo CT phenotyping into genetic improvement programs for more accelerated selection.

Keywords: 2026

How to Cite:

Amini, B., Baes, C., Bai, X., Barbut, S., Ellis, J., Harlander, A., Jahnel, R., Leishman, E., Makanjuola, B., Miglior, F., Tulpan, D. & Vanderhout, R., (2026) “Integrative analysis of leg health in turkeys using meta-analysis and deep learning approaches”, World Congress on Genetics Applied to Livestock Production Digital Archive 2026(1): 2286219. doi: https://doi.org/10.31274/wcgalp.23881

Rights: 1