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

Area of Science:

• Animal Science
• Agricultural Technology
• Machine Learning Applications

Background:

• The laying hen industry routinely culls male chicks, raising ethical and animal welfare concerns.
• Current methods for determining chick embryo sex are invasive, costly, and time-consuming.
• There is a need for a humane and efficient alternative to identify chick embryo sex pre-hatching.

Purpose of the Study:

• To develop a low-cost, non-invasive method for predicting chick embryo sex using egg morphology.
• To assess the accuracy of machine learning models trained on egg imaging data.
• To provide a scalable solution for the ethical challenges in the poultry industry.

Main Methods:

• A custom imaging apparatus was constructed using a smartphone and light box for consistent chicken egg image capture.
• Morphological features of eggs (length, width, area, eccentricity, extent) were measured.
• Machine learning models, including a wide neural network, were trained to predict chick embryo sex based on egg morphology.

Main Results:

• The wide neural network model achieved a highest accuracy of 88.9% and a mean accuracy of 81.5% in predicting chick embryo sex.
• The custom imaging apparatus demonstrated comparable accuracy to high-cost industrial 3D scanners in capturing egg morphology.
• The developed method shows potential to prevent the annual culling of billions of male chicks.

Conclusions:

• This non-invasive egg imaging and machine learning approach offers a feasible and ethical alternative for sex determination in chick embryos.
• The method has the potential to significantly improve animal welfare and sustainability in the poultry industry.
• Further improvements in accuracy and adaptability to different industry settings are anticipated.