US-VLA: An Ultrasound Vision-Language-Action Model for Embodied Abdomina
2026-08-17 • Robotics
RoboticsComputer Vision and Pattern Recognition
AI summaryⓘ
The authors developed a new AI system called US-VLA that helps guide ultrasound scanning by understanding clinical goals and controlling the probe based on real-time ultrasound images. Their model combines ultrasound images with other contextual information to better follow the scanning process. They also created a large dataset of expert ultrasound scans for liver and kidney exams to train and test their system. Their experiments show that US-VLA works well in guiding ultrasound probe movements and can generalize within abdominal ultrasound tasks.
Ultrasound imagingArtificial intelligenceReinforcement learningProbe manipulationVision-language modelMedical imaging datasetsSemantic feedbackAbdominal ultrasoundExpert trajectoriesReal-time guidance
Authors
Cheng Zhang, Xingzheng Wu, Guihao Yan, Xifeng Hu, Zhi Liu, Mei Wu, Qing Cai
Abstract
Artificial intelligence-assisted ultrasound scanning enhances diagnostic reliability and efficiency by providing real-time guidance for standardized image acquisition and reducing operator dependence. However, existing reinforcement learning and learning-assisted ultrasound scanning methods typically rely on carefully designed reward functions or extensive interaction data, which limits their generalization ability and stability across different devices, patient populations, and complex clinical scenarios. To address these challenges, we propose an ultrasound vision-language-action model (US-VLA) for automated ultrasound scanning that explicitly encodes clinical semantic goals and generates sequential probe manipulation actions under real-time ultrasound feedback. In particular, we first design an ultrasound-aware expert fusion module to jointly integrate ultrasound observations with auxiliary contextual information, enabling semantic ultrasound feedback to effectively guide the scanning process. Then, we construct US-VLA-Data, a real-world dataset covering liver and kidney examinations, which includes five clinically defined standard planes and comprises 320 expert scanning trajectories with approximately 80,000 synchronized timesteps. Extensive experiments demonstrate that US-VLA achieves competitive performance in ultrasound probe manipulation tasks, indicating its effectiveness and promising generalization within the evaluated abdominal ultrasound setting. The source code is available at https://github.com/VMVLab/US-VLA.