PS12-6. Hematological and Rheological Determinants of Species-specific Blood Dispersion Patterns for Automated Anemia Detection in Livestock.

Abstract

Anemia alters hemoglobin concentration, erythrocyte distribution, and blood rheology, leading to measurable changes in fluid dispersion behavior. This study investigated the influence of species-specific hematological and rheological properties on blood-spot patterns captured by an image-based biosensor for automated anemia detection in sheep, goats, and cattle. Whole blood samples were applied to a glycerol-treated substrate, generating structured dispersion gradients indicative of viscosity, pigment concentration, and clotting dynamics. Sheep displayed highly uniform, centrally distributed patterns characterized by stable color gradients and minimal peripheral irregularities. These consistent rheological properties enabled near-perfect discrimination between anemia classes with 98.8% accuracy, reflecting tightly clustered hematocrit distributions and predictable erythrocyte aggregation behavior. Goats exhibited greater heterogeneity in boundary gradients and internal clotting textures, consistent with documented caprine hematological diversity and changes in erythrocyte morphology and plasma composition. This increased dispersion complexity reduced class separability, resulting in 89.3% accuracy and increased sensitivity to borderline anemia states. Cattle showed consistent outer-ring formation with moderate internal gradient transitions, indicating relatively stable but more heterogeneous rheological properties compared to sheep. Classification accuracy reached 90.6%, demonstrating reliable yet multifactorial feature contributions during prediction. Explainable AI visualizations confirmed biologically meaningful interpretation patterns: sheep classifications were primarily influenced by centralized pigment density; cattle classifications required integration of boundary morphology and mid-zone gradients; and goat classifications depended on mixed internal texture and peripheral cues. These results demonstrate that biosensor-derived blood dispersion images encode physiologically relevant species-level hematological signals. The integration of rheological insights with automated image analysis strategies can advance precision livestock health monitoring and improve anemia management across diverse production systems.