Force-guided autonomous robotic ultrasound scanning control method for soft uncertain environment

Area of Science:

• Robotics
• Medical Imaging
• Artificial Intelligence

Background:

• Autonomous ultrasound imaging is crucial for therapeutic assistance but challenging in complex, uncertain clinical environments.
• Existing methods often require prior environmental or target information, limiting their adaptability.

Purpose of the Study:

• To develop an autonomous robotic ultrasound (US) control method using reinforcement learning (RL).
• To enable robot control of probe posture and contact force using only contact force as input, without a priori information.

Main Methods:

• An RL agent was trained using a policy gradient theorem-based model to learn the relationship between the US probe's 6-degree-of-freedom contact force and output force.
• A force control strategy based on an admittance controller was implemented for synchronous control of force, orientation, and position.

Main Results:

• The method achieved stable scanning on an unknown soft phantom, producing US images with minimal differences (3 ± 0.4%) compared to free-hand scans.
• Contact force and trajectory analyses confirmed stable scanning on uncertain soft surfaces.

Conclusions:

• The proposed force-guided US robot scanning control method is effective for soft, uncertain environments.
• Experimental validation demonstrates feasibility for complex skin surface scanning, highlighting potential in clinical robotic US imaging, especially with limited visual feedback.