Hydronephrosis Surgical Intervention

This study developed machine learning models using kidney ultrasound images from multiple visits to predict which patients with hydronephrosis may require pyeloplasty. Although multi-visit models incorporated more longitudinal data, they did not perform better than models using a single ultrasound, based on AUROC and AUPRC comparisons across multi-institutional datasets. The findings suggest that a single early ultrasound may be sufficient for accurate risk stratification in clinical practice.