Radiofrequency Wave Sensing for Rapid Animal Health Monitoring: A Proof-of-Concept Study

Anemia caused by gastrointestinal parasitism is a major constraint to small ruminant productivity, particularly in low-resource production systems where diagnostic tools and veterinary access are limited, with use of FAMACHA as a biological reference This study evaluated the potential of radio-frequency non-destructive technique (RF-NDT) wave-derived features as non-invasive biomarkers for anemia detection in goats, using FAMACHA© scores as a biological reference. Variable clustering of the top ten frequencies revealed distinct patterns across health states. Healthy (FAMACHA© 1) animals were characterized by a single frequency cluster centered at 8.43 GHz, which explained 93.7% of variation, whereas moderately affected animals (FAMACHA© 2) shifted to 9.33 GHz with reduced uniformity (88.7%). Borderline animals (FAMACHA© 3) required two clusters (9.89 and 8.23 GHz), explaining 91.0% of variation, indicating increasing tissue heterogeneity with anemia progression. Regression analysis demonstrated strong predictive power, with Linear Regression achieving R² = 1.00 and Random Forest R² = 0.79 (RMSE = 0.07), Support Vector Regression underperformed (R² = 0.31). Classification models confirmed the feasibility of categorical anemia detection. The Multilayer Perceptron achieved the highest accuracy (0.84), F1-score (0.83), and ROC-AUC (0.94), outperforming Support Vector Machine (accuracy 0.67, F1 = 0.67) and K-Nearest Neighbors (accuracy 0.60, F1 = 0.61). These findings establish proof-of-concept that RF waves capture physiologically meaningful dielectric signatures linked to anemia, reflecting hemoglobin concentration, hydration, and microcirculatory function. The integration of RF sensing with machine learning offers a rapid, and non-invasive scalable diagnostic approach. Future work should expand validation across breeds and environments, optimize sensor design, and embed neural classifiers for field-ready deployment.