HARP Ratio Calculation

This study assessed the feasibility of using deep learning to automatically measure the hydronephrosis area to renal parenchyma (HARP) ratio from pediatric renal ultrasound images. Using 168 images from patients who underwent pyeloplasty, multiple segmentation architectures and ensemble models demonstrated strong agreement with manual tracings, achieving average Dice coefficients around 0.91. The results support accurate and reproducible AI-based assessment of hydronephrosis severity, though broader external validation is recommended.