Modular Robotic System With AI-Enhanced Human-Robot Collaboration for Disposable Ureteroscopy in Kidney Stone Surgery

Flexible ureteroscopy is a cornerstone of kidney stone management. However, conventional methods pose occupational risks such as musculoskeletal strain and radiation exposure. This study presents a modular robotic system with AI-enhanced human-robot collaboration, designed to address these challenges while maintaining clinical efficacy. The system features a disposable ureteroscope to minimize infection risks and a modular architecture for rapid assembly. It integrates dual control frameworks: force-feedback-based robotic arm positioning and adaptive dual-mode leader-follower teleoperation. Computer vision enables autonomous endoscopic delivery through image classification based on ResNet-18 (precision: 99.58%, recall: 99.79%) and visual guidance within real-time stone detection via YOLOv7 (96.69% mAP@0.5). The validation experiments for autonomous delivery demonstrated a mean axial positioning error of 1.246 mm (<inline-formula> <tex-math notation="LaTeX">$\sigma \,\, {=} \,\, 0.965$ </tex-math></inline-formula>), which meets stringent clinical standards. Crucially, the proposed leader-follower robotic control architecture demonstrated non-inferior operational efficiency compared to manual manipulation (manual: <inline-formula> <tex-math notation="LaTeX">$204.7~{\pm }~71$ </tex-math></inline-formula>.5 s vs. robotic: <inline-formula> <tex-math notation="LaTeX">$202.1~{\pm }~64$ </tex-math></inline-formula>.9 s), while the vision-guided function establishes a technical pathway toward autonomous surgical execution. These results validate the system’s clinical equivalence to manual techniques, augmented by enhanced precision, safety, and adaptability, positioning it as a novel solution for modern urological interventions.