Early Failure Detection in Autonomous Surgical Soft-Tissue Manipulation via Uncertainty Quantification

2025-01-17

AI Summary

Autonomous surgical robots frequently suffer from sudden failures during soft-tissue manipulation due to unpredictable tissue deformation and mechanical heterogeneity. Method: We propose the first uncertainty quantification framework tailored for learned soft-tissue manipulation policies. Our approach integrates deep ensembles with Monte Carlo Dropout to build a zero-shot sim2real failure prediction model, enabling generalization across tissue types and bimanual coordination. Coupled with a reinforcement learning policy network and the dVRK physical platform, it enables real-time task-risk assessment and human-in-the-loop takeover. Results: Evaluated on the real dVRK platform, our system achieves a 47.5% improvement in zero-shot transfer performance over baselines. It reliably predicts impending failures—triggering timely human intervention—while sustaining high autonomous task success rates. This significantly enhances robustness and safety of surgical robots operating on previously unseen soft tissues.

Technology Category

• Intelligent Robots: Manipulation
• Machine Learning: Calibration & Uncertainty Quantification
• Natural Language Processing: Safety and Robustness

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI

Abstract

Autonomous surgical robots are a promising solution to the increasing demand for surgery amid a shortage of surgeons. Recent work has proposed learning-based approaches for the autonomous manipulation of soft tissue. However, due to variability in tissue geometries and stiffnesses, these methods do not always perform optimally, especially in out-of-distribution settings. We propose, develop, and test the first application of uncertainty quantification to learned surgical soft-tissue manipulation policies as an early identification system for task failures. We analyze two different methods of uncertainty quantification, deep ensembles and Monte Carlo dropout, and find that deep ensembles provide a stronger signal of future task success or failure. We validate our approach using the physical daVinci Research Kit (dVRK) surgical robot to perform physical soft-tissue manipulation. We show that we are able to successfully detect task failure and request human intervention when necessary while still enabling autonomous manipulation when possible. Our learned tissue manipulation policy with uncertainty-based early failure detection achieves a zero-shot sim2real performance improvement of 47.5% over the prior state of the art in learned soft-tissue manipulation. We also show that our method generalizes well to new types of tissue as well as to a bimanual soft tissue manipulation task.

Problem

• Robotics Surgery
• Uncertainty Prediction
• Soft Tissue

Innovation

• Predictive Uncertainty
• Deep Ensembles Technique
• Adaptive Bimanual Robotic Surgery