From Freehand Demonstrations to Autonomous Robotic Ultrasound: An Anatomy-Anchored Learning Framework
From Freehand Demonstrations to Autonomous Robotic Ultrasound: An Anatomy-Anchored Learning Framework
This research introduces an anatomy-anchored imitation learning framework for robotic ultrasound scanning. The framework combines anatomy-guided probe placement with learned ultrasound navigation, enabling autonomous scanning with high task success rates. The approach reduces reliance on third-person vision and achieves 96%, 94%, and 88% success in kidney, gallbladder, and liver standard-plane acquisitions, respectively, with significant improvements over baseline models.
This paper presents an anatomy-anchored imitation learning framework for robotic ultrasound scanning, demonstrating improved task success rates and reduced reliance on third-person vision. The approach combines anatomical guidance with egocentric vision and force feedback, achieving high success rates in standard-plane acquisitions and radial artery sweeping.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
Reported numbers
time-to-target
0
View original evidence
hand scanning without a robot. On 16 unseen participants, task success reaches 96%, 94%, 88%, and 94% for kidney, gallbladder, and liver standard-plane acquisition and radial artery sweeping, respectively. Compared with a fine-tuned pi_0 model, success on the abdominal tasks improves by 10-42 percentage points, with lower time-to-target, peak contact force, and trajectory jerk. Ablations support the complementary rol
Open source S5peak contact force
0
View original evidence
hand scanning without a robot. On 16 unseen participants, task success reaches 96%, 94%, 88%, and 94% for kidney, gallbladder, and liver standard-plane acquisition and radial artery sweeping, respectively. Compared with a fine-tuned pi_0 model, success on the abdominal tasks improves by 10-42 percentage points, with lower time-to-target, peak contact force, and trajectory jerk. Ablations support the complementary rol
Open source S5
- paper: not_reported
- code: not_reported
- weights: not_reported
- dataset: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
hand scanning without a robot. On 16 unseen participants, task success reaches 96%, 94%, 88%, and 94% for kidney, gallbladder, and liver standard-plane acquisition and radial artery sweeping, respectively. Compared with a fine-tuned pi_0 model, success on the abdominal tasks improves by 10-42 percentage points, with lower time-to-target, peak contact force, and trajectory jerk. Ablations support the complementary rol
Open source S5
Source：arXiv Robotics — research abstracts · arxiv.org