Automated chick gender determination using optical coherence tomography and deep learning

Abstract

Chick gender classification is crucial for optimizing poultry production, yet traditional methods such as vent sexing and ultrasound remain limited by human expertise, labor intensity, and insufficient resolution. This study introduces a novel approach that integrates Optical Coherence Tomography (OCT) and deep learning to enable high-resolution, non-invasive chick sexing. Unlike conventional imaging techniques, OCT provides micrometer-scale visualization of cloacal structures, allowing precise differentiation between male and female chicks based on internal anatomical markers. We developed a custom convolutional neural network (CNN) optimized for OCT data, incorporating asymmetric image resizing and enhanced feature extraction to improve classification accuracy. Our model achieved 79 % accuracy, outperforming conventional architectures such as Inception (63 %) and VGG-16 (74 %), highlighting the importance of a tailored, domain-specific model. This is the first study to integrate OCT with deep learning for automated chick sexing, demonstrating a scalable, real-time alternative to expert-dependent vent sexing. With further advancements in imaging and machine learning, our approach has the potential to transform chick sexing in commercial hatcheries, reducing reliance on skilled labor while enhancing classification efficiency and precision.