Morphology-Based In-Ovo Sexing of Chick Embryos Utilizing a Low-Cost Imaging Apparatus and Machine Learning

Abstract

The routine culling of male chicks in the laying hen industry raises significant ethical, animal welfare, and sustainability concerns. Current methods to determine chick embryo sex before hatching are costly, time-consuming, and invasive. This study aimed to develop a low-cost, non-invasive solution to predict chick embryo sex before hatching using the morphological features of eggs. A custom imaging apparatus was created using a smartphone and light box, enabling consistent image capture of chicken eggs. Egg length, width, area, eccentricity, and extent were measured, and machine learning models were trained to predict chick embryo sex. The wide neural network model achieved the highest accuracy of 88.9% with a mean accuracy of 81.5%. Comparison of the imaging apparatus to a high-cost industrial 3D scanner demonstrated comparable accuracy in capturing egg morphology. The findings suggest that this method can contribute to the prevention of up to 6.2 billion male chicks from being culled annually by destroying male embryos before they develop the capacity to feel pain. This approach offers a feasible, ethical, and scalable alternative to current practices, with potential for further improvements in accuracy and adaptability to different industry settings.