Quality-Driven Adaptive Control Framework for Robotic Ultrasound Imaging of Vascular Anatomies

This paper proposes a quality-driven adaptive control framework for robotic vascular anatomies scanning to facilitate the acquisition of high-quality ultrasound (US) images. Specifically, a novel probability-based US image quality evaluation metric for vascular anatomies is introduced, leveraging an image segmentation network to establish a mapping between the controlled variables of the robot (e.g., pose and force) and US image quality. Furthermore, an adaptive US probe control strategy driven by US image quality is developed to optimize real-time image acquisition, with its stability rigorously proven. To assess the effectiveness of the proposed framework, two experiments were conducted on a human tissue-mimicking phantom, encompassing both static and dynamic scenarios. The experimental results demonstrate that the proposed framework ensures stable contact force and significantly enhances US image quality for robot-assisted vascular anatomy imaging, even in the presence of external disturbances.