Genetics of digital phenotypes of keel bone in layer chickens and correlations with keel bone fractures and deviations

Abstract

BACKGROUND: Poultry is a global industry with laying hens that are genetically optimized for high egg yield. Keel bone fractures can affect up to 80% of laying hens, posing welfare and production problems. Therefore, genetic selection to reduce keel fractures is important. However, the lack of a reliable, automated, and heritable phenotypes for keel bones makes this a challenging task. The aim of this study was to (1) develop automated analyses of radiographic images to phenotype keel bones, and (2) investigate whether the proposed phenotypes are heritable and genetically correlated with the post-dissection scores of keel bone fractures and deviations. A total of 1051 laying hens (Bovans Brown and Lohmann Brown) from a commercial farm were x-rayed, followed by keel bone dissection and scoring for deviations and fractures. Furthermore, blood was sampled for genotyping using 50 K Illumina SNP chips. Keel bones were segmented (with ~ 0.90 accuracy) from the radiographic images using deep learning models, after which the images were automatically measured for general geometry and radiopacity. Multi-trait genomic restricted maximum likelihood was used to estimate genetic parameters. RESULTS: Heritability estimates ranged from 0.28 to 0.30 for both keel deviations and fractures observed post-dissection. The automated phenotypes had heritability estimates ranging from 0.07 to 0.10 for keel radiopacity and from 0.11 to 0.39 for keel geometry. Estimates of genetic correlations of keel geometry with keel deviation and fractures ranged from -0.57 to 0.72. CONCLUSIONS: Automated methods were developed for measuring keel bone radiopacity and geometry. Keel concave area was found to be a reliable and heritable phenotype that breeding companies can use to reduce keel deviations and fractures. These methods can also be adapted to measure other bones (e.g., tibiotarsal) or objects (e.g., eggs), allowing breeders to quickly compute phenotypes for keel, tibia, and egg size from the same radiographic image. The developed methods are well-suited for large-scale studies to assess different housing environments and nutrition strategies aimed at improving keel bone conditions.

Plant phenotyping relevance

ニワトリの骨を対象とするが、放射線画像と深層学習による骨形状・放射輝度の自動表現型測定法の開発が中心であり、植物フェノタイピングではないため対象外。

Code and data availability

The supplied blocks describe automated keel bone phenotyping from radiographic images (segmentation via deep learning, OpenCV-based measurements, GREML analysis), but contain no data availability statement, no public repository deposit, no author code URL, and no supplement explicitly containing the radiographic images

No evidence-backed public reproduction asset is currently recorded.