Respiratory Motion-Robust Robotic Ultrasound Acquisitions via Vision-Haptic Fusion Control and 3-D Compensation

Abstract

Robotic three-dimensional (3D) ultrasound (US) has demonstrated potential for reliable organ volume acquisition. However, respiratory motion during scanning critically reduces volume measurements accuracy, presenting a clinical limitation that affects not only cardiovascular but also all abdominal and spinal US diagnoses. To address this, we develop a robotic system integrating vision-based motion monitoring and interaction control for respiratory motion-robust US acquisition. The motion monitoring module constructs a self-calibrating predictive model that dynamically estimates respiratory motion and predicts the desired contact position. We further propose a vision-haptic fusion motion control framework that integrates predicted position feedback into the force-sensing admittance controller, equipping it with model-predictive capabilities. Afterward, to ensure accurate volume measurements, a compensation framework is presented to reduce respiratory motion-induced distortions by reconstructing in a relatively static frame and refining results using anatomical priors. Finally, the whole system was validated using a vessel phantom mounted on a six-axis platform simulating respiratory motion. Experimental results demonstrate that the robotic system ensures interaction stability (0.295 N force tracking accuracy) for US acquisition. The compensation framework significantly improve volume measurements accuracy, achieving a mean spatial distortion of 0.48 mm. These improvements highlight the robotic system’s capacity for stable US acquisition and accurate volume measurements across US diagnoses, which are crucial for diagnoses affected by respiratory motion.