Machine learning based on behavioral, fecal and oocyst analyses for chicken coccidiosis diagnosis: A systematic review with exploratory extension

In the context of large-scale and intensive poultry farming, avian coccidiosis causes substantial economic losses. Timely and accurate diagnosis is essential for effective control and prevention. However, conventional diagnostic methods, including microscopic examination and molecular assays, are constrained by limited efficiency and a heavy reliance on specialized expertise, rendering them inadequate for real-time, high-throughput diagnosis in modern farming systems. Recent advances in machine learning have demonstrated significant potential in image recognition and early disease detection, particularly for diagnosing coccidiosis based on fecal and oocyst images, positioning this technology as a promising tool for future diagnosis of avian coccidiosis. Despite considerable progress in this area, approaches that rely solely on fecal and oocyst recognition face challenges in early screening. Moreover, limitations in current datasets and models often lead to insufficient generalization under practical farming conditions. Furthermore, although monitoring behavioral and physiological abnormalities could facilitate early detection, research on this aspect remains largely unexplored for avian coccidiosis. To address these gaps, this systematic review with exploratory extension adopts a three-dimensional framework, "behavioral physiology, fecal characteristics, and oocyst morphology" to systematically summarize the current applications and challenges of machine learning in identifying avian coccidiosis, with the aim of clarifying existing research gaps and technical bottlenecks. Finally, to systematically meet the demand for early and efficient diagnosis, this review proposes a conceptual framework for an AI-based surveillance system for chicken coccidiosis that integrates multimodal data across behavioral screening, fecal validation, and oocyst-based analysis encompassing species identification, oocyst counting, sporulation-state classification and viability assessment. This framework will serve as a theoretical foundation for moving the field from laboratory research toward field applications, thereby enabling early, precise, efficient, and intelligent prevention and control of avian coccidiosis.