Precise Detection of Eimeria Oocysts in Sheep: A Deep Learning Model Based on Microscopic Images

Background Parasitic infections remain a significant contributor to productivity losses in global livestock production systems. Conventional diagnostic approaches rely on specialized microscopic equipment and veterinary pathology expertise. Contemporary automated detection systems face implementation barriers stemming from scarcity of annotated microscopy datasets and persistent challenges in discriminative feature extraction. Methods This study develops YOLO-GA, a deep learning-enhanced object detection framework for precise identification of Eimeria oocysts in ovine specimens. Building upon the YOLOv5 architecture, we integrate two novel components: 1) Contextual Transformer (CoT) modules for localized contextual feature enhancement, and 2) Normalized Attention Mechanisms (NAM) for adaptive feature recalibration. Architectural optimizations including skip connections and hierarchical feature fusion were incorporated into the Backbone and Neck networks. Results Experimental validation using a curated dataset of 1,500 Eimeria oocyst microscopy images (200× optical magnification) demonstrated model efficacy, achieving a mean average precision (mAP) of 98.9% with 95.2% precision. Comparative analyses revealed YOLO-GA's superior performance across evaluation metrics while maintaining a computationally efficient framework relative to existing detection models. Conclusions The YOLO-GA framework successfully achieves automated Eimeria oocyst detection with high diagnostic accuracy and operational robustness. This advancement establishes a foundation for intelligent veterinary parasitology diagnostics and scalable livestock health monitoring solutions.