A study on egg sex-related associations based on PP-HGNet and morphological parameters

• A large, industry-relevant dataset was constructed using 1,008 eggs from two commercial laying hen breeds, Xiaoyou 818 and Datu2. • Egg sex identification was investigated using a dual-validation framework, combining statistical analysis with PP-HGNet–based deep learning on morphological features. • Demonstrated that neither single nor composite morphological parameters (e.g., EI, L) can reliably determine egg sex in commercial laying hens, defining clear limits for industrial application. In the poultry industry, accurately determining the sex of eggs before hatching is crucial due to the low economic value of male chicks. Previous studies on the application of the Egg Shape Index (EI) for sex identification have yielded conflicting results, with many studies relying on small or non-commercial populations, lacking sufficient industry relevance. To address these issues, this study systematically analyzed 1008 eggs from two commercial layer breeds (Xiaoyou 818 and Datu2), providing a larger and more industry-relevant dataset compared to previous research. We evaluated the potential of external morphological features for sex identification using both statistical analysis and deep neural network (PP-HGNet). The results showed no significant differences between male and female eggs, and the model's predictive accuracy approached random chance. These findings suggest that while EI and related morphological features alone are insufficient for reliable early sex identification, the large sample size, multi-breed design, and dual-method approach ensure the robustness and transparency of the results, offering valuable insights for industrial applications.