Overview
Researchers from Vanderbilt University and Université Grenoble Alpes have introduced RAUL, a reference-assisted reconstruction framework that recovers ureteroscope trajectories from endoscopic video alone—no external tracking hardware required. The work, submitted to arXiv on 16 September 2026 under arXiv:2609.19236 [cs.CV], targets a persistent gap in surgical training: the absence of objective, quantitative metrics for evaluating ureteroscopic navigation skill.
The Clinical Problem
During ureteroscopic kidney stone surgery, incomplete navigation of the anatomy can leave stones behind, contributing to repeat interventions. While experienced surgeons achieve notably lower reintervention rates, no standardized objective metric currently exists to determine when a trainee has reached a skilled level of scope navigation. Existing approaches typically depend on electromagnetic tracking systems—equipment that is expensive and difficult to scale across training programs.
Proposed Method: Reference-Assisted Localization
RAUL sidesteps the need for external sensors by exploiting a simple but effective strategy. For each phantom, a slow, high-quality reference exploration video is captured and used to generate a reference reconstruction. Subsequent exploration videos are then localized against this reference, yielding full trajectory estimates from video alone.
The approach is evaluated against electromagnetically tracked scope pose as ground truth, and navigation metrics are computed from the resulting phantom exploration trajectories to compare surgical residents across experience levels.
Key Results
- Localization accuracy: Mean translation root mean square error of 0.5 ± 0.1 mm across 9 phantoms.
- Coverage improvement: Frame-wise localization coverage rises from 50.5 ± 14.9% with standard Structure-from-Motion (SfM) to 86.1 ± 7.2% with the proposed reference-assisted pipeline.
- Skill discrimination: Reconstructed trajectories revealed statistically significant differences between high- and low-experience trainees across established navigation metrics.
Significance
According to the authors, this is the first demonstration of video-only recovery of ureteroscope trajectories—without external tracking sensors—for surgical skill assessment. By removing the dependency on dedicated tracking equipment, RAUL opens the door to scalable, automated evaluation of ureteroscopy navigation skill, with the potential to standardize how trainees are assessed and certified.
Paper Details
- Title: RAUL: Reference-Assisted Ureteroscopy Localization for Skill Assessment
- Authors: Fangjie Li, Mai Bui, Charan Mohan, Michael Miga, Matthieu Chabanas, Nicholas Kavoussi, Jie Ying Wu
- arXiv: 2609.19236 [cs.CV]
- Submitted: 16 September 2026
- Subject: Computer Vision and Pattern Recognition (cs.CV)
via ArXiv CV