Anatomical Prior-Driven Framework for Autonomous Robotic Cardiac Ultrasound Standard View Acquisition
Zhiyan Cao, Zhengxi Wu, Yiwei Wang, Pei-Hsuan Lin, Li Zhang, ZHEN XIE, Huan Zhao, Han Ding
AI summary
Problem
Standard cardiac ultrasound view acquisition is highly operator-dependent and inconsistent. Current models lack anatomical constraints for robust segmentation and rely on fragile heuristics or black-box learning for autonomous probe control.
Approach
The method embeds spatial-topological constraints into a YOLO segmentation model via a graph module, extracts quantifiable anatomical features, and formulates robotic probe adjustment as a reinforcement learning problem guided by probabilistic anatomical benchmarks.
Key results
- SRG-YOLOv11s improves mAP50 by 11.3% and mIoU by 6.8% on the Special Case dataset
- RL agent achieves 92.5% success rate in simulation
- RL agent achieves 86.7% success rate in phantom experiments
- Validated zero-shot deployment on a custom robotic ultrasound platform
Why it matters
Enables reproducible, expert-level cardiac ultrasound acquisition without heavy operator dependency, advancing autonomous medical imaging.
Abstract
Cardiac ultrasound diagnosis is critical for car- diovascular disease assessment, but acquiring standard views remains highly operator-dependent. Existing medical segmenta- tion models often yield anatomically inconsistent results in im- ages with poor textural differentiation between distinct feature classes, while autonomous probe adjustment methods either rely on simplistic heuristic rules or black-box learning. To address these issues, our study proposed an anatomical prior (AP)- driven framework integrating cardiac structure segmentation and autonomous probe adjustment for standard view acqui- sition. A YOLO-based multi-class segmentation model aug- mented by a spatial-relation graph (SRG) module is designed to embed AP into the feature pyramid. Quantifiable anatomical features of standard views are extracted. Their priors are fitted to Gaussian distributions to construct probabilistic APs. The probe adjustment process of robotic ultrasound scanning is formalized as a reinforcement learning (RL) problem, with the RL state built from real-time anatomical features and the reward reflecting the AP matching. Experiments validate the efficacy of the framework. The SRG-YOLOv11s improves mAP50 by 11.3% and mIoU by 6.8% on the Special Case dataset, while the RL agent achieves a 92.5% success rate in simulation and 86.7% in phantom experiments.