Related Experiment Video
Updated: Jul 11, 2025
05:57
A Teleoperated Robotic System-Assisted Percutaneous Transiliac-Transsacral Screw Fixation Technique
Published on: January 6, 2023
2.3K
ASAP-CORPS: A Semi-Autonomous Platform for COntact-Rich Precision Surgery
Mythra V Balakuntala1, Glebys T Gonzalez2, Juan P Wachs1
1School of Engineering Technology, Purdue University, West Lafayette, IN 47906, USA.
Military Medicine
|November 10, 2023
Summary
This study introduces a novel reinforcement learning (RL) method for surgical robots, enabling them to learn complex procedures from limited demonstrations. This advance improves robotic surgery in remote, austere environments by enhancing skill acquisition and task success rates.
Area of Science:
- Robotics
- Artificial Intelligence
- Surgical Technology
Background:
- Remote military operations demand rapid medical response in austere, communication-limited environments.
- Semi-autonomous teleoperated systems are crucial for critical care but require advanced AI for complex procedures.
- Training surgical robots often necessitates extensive programming or large demonstration datasets, which are impractical for complex maneuvers.
Purpose of the Study:
- To develop a method for robots to learn surgical skills from limited demonstrations, reducing data requirements.
- To enhance the generalizability and robustness of robotic surgical policies in dynamic environments.
- To enable autonomous execution of complex surgical tasks, like cricothyroidotomy, in challenging conditions.
Main Methods:
- Collected a demonstration dataset of the cricothyroidotomy task using a teleoperated robotic platform.
- Developed models for automatic segmentation and classification of surgical subtasks (surgemes).
- Implemented a multimodal off-policy reinforcement learning approach with rewards derived from expert demonstrations.
Main Results:
- Achieved 98.2% accuracy in task segmentation and 96.25% accuracy in surgeme classification.
- The robot execution achieved a 93.5% task success rate, outperforming behavioral cloning and shaped-reward RL.
- Proposed interaction features significantly improved classification accuracy for surgical tasks.
Conclusions:
- The developed method effectively learns surgical skills from demonstrations, surpassing existing methods.
- The approach enhances classification accuracy for surgical task segmentation and recognition.
- This technology shows significant potential for advancing remote telemedicine in battlefield scenarios.
