Boriceanu, Ioana-Roxana (National University of Science and Technology Politehnica Buchar), Popescu, Dan (National University of Science and Technology Politehnica Buchar), Ichim, Loretta (Politehnica University of Bucharest)

Kidney Stone Detection and Segmentation Using a YOLO V11 + U-Net Pipeline

Scheduled for presentation during the Invited Session "Intelligent Data processing" (WeBB), Wednesday, June 11, 2025, 14:00−14:20, Room A

33rd Mediterranean Conference on Control and Automation, June 10-13, 2025, Tangier, Morocco

This information is tentative and subject to change. Compiled on May 9, 2025

Abstract

Kidney stones, or renal calculi, affect approximately 10% of the global population and can lead to severe complications such as urinary obstruction and kidney failure. We propose a two-stage pipeline for automated detection and segmentation of kidney stones in CT images. The first stage uses a YOLOv11 network trained on a public dataset to identify regions of interest (ROIs). High-quality segmentation masks, generated using the Segment Anything Model (SAM), were added to the same dataset and used to train a U-Net for detailed segmentation. The final pipeline combines YOLO for ROI detection and U-Net for segmentation, achieving strong performance with metrics such as a Dice coefficient of 0.93 and IoU of 0.88. This approach contributes to kidney stone research and provides a scalable framework for nephrolithiasis assessment in clinical applications.