Investigating Robot Control Policy Learning for Autonomous X-ray-guided Spine Procedures
Researchers have investigated the use of imitation learning for robot control policies in autonomous X-ray-guided spine procedures. They developed a realistic simulation sandbox to train policies for cannula insertion using only visual X-ray information. The trained policy demonstrated success in 68.5% of simulated cases, maintaining safe trajectories and showing potential for transfer to complex anatomies and even real X-ray data, though entry point precision remains a challenge. AI
IMPACT Demonstrates a novel application of imitation learning for autonomous robotic surgery, potentially reducing reliance on CT scans.