Autonomous Liver Ultrasound Examination Based on Imitation Learning and Stiffness Estimation

In the past few decades, the demand for assistive robots in the field of healthcare has steadily increased. We conducted research on an autonomous robot system based on imitation learning to assist in liver scans guided by ultrasound. Traditional procedures for liver examination using ultrasound imaging typically require a physician to slide the ultrasound probe along the patient's lower ribs for multi-angle observation. To automate this process, we employed imitation learning to capture the physician's manipulation techniques. However, the imitation generalization system often struggles to accurately extract the starting and ending points due to the obstruction of fat and skin, and the individual variations in patient anatomy lead to inconsistencies in the positions of lower ribcage. To improve the scanning process, we introduced stiffness estimation to differentiate between the abdomen and the rib, allowing real-time adjustments for optimal alignment of lower ribcage. Furthermore, due to the need for occasional physician intervention during the detection process, we implemented mechanisms for pausing and resuming the trajectory learning system to ensure continued operation. Ultimately, the proposed approach was evaluated using the Franka Panda robot system and demonstrated autonomous liver ultrasound examination in scenarios involving multiple human subjects.