Continuous renal calculi tracking for autonomous robotic ureteroscopic lithotripsy

Quan Zhang, Xiaohang Nie, Jingqian Sun, Ruichao Tang, Yichao Tang

Abstract: Renal calculi, while not inherently life-threatening, can induce excruciating pain during acute episodes. The predominant clinical treatment – ureteroscopic lithotripsy (URS) –currently faces challenges including restricted maneuverability, frequent manual adjustments during dynamic calculi movement, and tissue damage risk, highlighting the need for robotic assistance. This study proposes an autonomous lithotripsy system through three integrated technological advancements: 1) a robotic ureteroscope with sub-millimeter-scale positioning accuracy; 2) a concatenated Quenching-net semantic segmentation visual processing framework based on convolutional neural networks, achieving a segmentation accuracy of 98.5% (validated on 12,768 endoscopic images); 3) a control strategy developed through collaborative deep reinforcement learning(DRL), enabling a 93% success rate in tracking randomly moving calculi. This system's autonomous calculi localization capability reduces operator fatigue and may mitigate cognitive bias in calculi targeting. It demonstrates how embodied AI enhances medical procedural precision while preserving human oversight in critical decisions.