This work presents the development and evaluation of an Automatic Robotic Ultrasound Scanning (ARUS) system capable of real-time segmentation and 3D reconstruction of muscles and bones. The ARUS system integrates a robotic arm, ultrasound imaging, and stereo vision to achieve fully automated scanning, eliminating the need for operator expertise.
The system employs hybrid position-force control, advanced calibration techniques, and a U-Net-based segmentation model to ensure precision in dynamic environments. Experimental evaluations were conducted on simple and realistic phantoms, as well as in vivo, demonstrating robust reconstruction accuracy with deviations below 1 mm under optimal conditions. Visual servoing was incorporated to enhance reconstruction quality, though its effectiveness was limited by processing delays and challenges in dynamic and non-uniform environments. Force regulation exhibited consistent performance across all experiments, with mean forces maintained within safe limits. Heat maps provided insights into force distribution, highlighting areas for system improvement.
This study underscores the ARUS system’s potential as a cost-effective alternative to MRI for muscle analysis, offering real-time insights into muscle dynamics during movement. Future work includes developing interactive 3D muscle reconstructions, optimizing computational efficiency, and exploring innovative control and reconstruction strategies to enhance clinical applicability.