Automated detection of soil transmitted helminthes and schistosomiasis using yolo based deep learning models

Soil transmitted helminths (STHs) and schistosomiasis remain prevalent public health concerns in tropical and resource constrained regions necessitating accurate and timely diagnostic approaches for effective treatment and control. However, conventional microscopic examination of parasite eggs is laborious, time intensive and dependent on expert interpretation which can introduce variability and diagnostic errors. To address these challenges, this study proposes a real time deep learning-based detection and classification using yolo based object detection models for automated analysis of microscopy images. Specifically, yolov11 and yolov12 architectures were evaluated across three lightweights to medium model variants (Nano, Small and Medium) to assess tradeoffs between detection accuracy, inference speed and computational efficiency. A custom annotated dataset comprising 1713 images from Ethiopian health institute with four clinically relevant parasite classes ( Ascaris , hookworm, Schistosoma and Trichuris ) was used for independent training, validation and testing. Model performance was evaluated using standard object detection metrics including mean average precision (mAP), frames per second (FPS) and giga floating-point operations (GFLOPs). A 5-fold cross validation and statistical significance analysis ( p < 0.05) were conducted to ensure robustness and reproducibility. Experimental results indicate that yolov12m achieves the highest detection performance (mAP@0.5 = 94% ± 0.1%) and mAP@0.5–0.95 of 67.4% ± 0.3% while yolo11m scored the highest recall of 88.6% ± 0.4% indicating fewer false Negative. In contrast, yolov12n offers the lowest computational cost 6.3 giga floating-point operations (GFLOPs) by delivering the fast inference of 159 FPS with reduced sensitivity. These findings demonstrate the effectiveness of yolo based models for scalable, real time parasite detection and provide practical guidance for deployment in resource constrained healthcare settings.