Event

Anatomy-anchored imitation learning enables autonomous ultrasound scanning

1 reports · 1 sources

Yuan Bi and co-authors propose the "anatomy-anchored imitation learning" framework, which combines anatomy-guided probe placement with learned ultrasound navigation for autonomous robotic ultrasound scanning. The authors report that, on 16 unseen participants, the task success rates for standard-view acquisition of the kidney, gallbladder, and liver and for radial artery scanning were 96%, 94%, 88%, and 94%, respectively. The framework collects multimodal demonstrations during natural freehand scanning via a handheld acquisition platform, enabling data collection without a robot. The authors also report that end-to-end trials with automated placement achieved a success rate of 90%–100%, and that the policy transferred to a second robot without retraining, with a success rate of 83%–100%.