Deep learning models excel in complex classification tasks but often lack interpretability, limiting their adoption in domains where explainability is critical, such as medicine and veterinary science. This work presents a hybrid approach that combines deep learning and symbolic reasoning to classify bull spermatozoa morphology in an explainable manner. We utilise YOLOv8 for object detection and morphological and viability classification of bull’s spermatozoa from microscope-acquired images, achieving high accuracy. To tackle explainability, FastLAS was employed to learn human-readable classification rules. These rules, coupled with the xASP2 framework, enable traceable justifications for each classification, addressing the black-box nature of deep learning. Experimental evaluation demonstrates that, while FastLAS does not match YOLO’s performance, it outperforms traditional machine learning models and offers significant benefits in explainability. This approach provides a practical solution for integrating explainable AI in reproductive biology, with implications for medical AI systems where transparency is essential.
