SutureFormer learns surgical trajectories using goal-conditioned offline RL

Researchers have developed SutureFormer, a novel framework for learning surgical trajectories from endoscopic video using goal-conditioned offline reinforcement learning in pixel space. This approach treats the needle tip as an agent that moves step-by-step, naturally capturing the continuity of motion and enabling the explicit modeling of plausible state transitions. SutureFormer utilizes Conservative Q-Learning with Behavioral Cloning regularization to optimize policies from expert demonstrations, effectively converting sparse annotations into dense reward signals. Experiments on a kidney wound suturing dataset demonstrated a significant reduction in Average Displacement Error compared to existing methods. AI

IMPACT This research could lead to more precise and autonomous robot-assisted surgery by improving the accuracy of needle trajectory prediction.

RANK_REASON The cluster contains a research paper detailing a new method for surgical trajectory learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• Average Displacement Error
• behavioral cloning
• Conservative Q-Learning for Offline Reinforcement Learning
• Huanrong Liu
• kidney wound suturing dataset
• Pixel Space Battles
• reinforcement learning
• SutureFormer

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →