Force control for the robot-assisted laparoscopic ultrasound scanning system

Laparoscopic ultrasound scanning, an acritical diagnostic imaging method for liver diseases, necessitates precise contact force control that is typically dependent on skilled physicians. Given its high mobility and controllability, the robot is anticipated to supplant physicians in performing laparoscopic ultrasound scans. Consequently, this study introduces a laparoscopic ultrasound scanning system paired with a learning-based force controller. Compared with traditional methods, our approach offers precise control over the contact force between the ultrasound probe and human organ surfaces during scanning, reducing the learning curve for surgeons in laparoscopic liver ultrasound scanning. Acknowledging the complex nonlinear mechanical model involved when the laparoscopic ultrasound probe contacts human organs and tissues, we propose a learning-based method utilizing a long short-term memory network to estimate the contact force. This method allows for the precise estimation of contact force with human organ tissues by analyzing tendon force. The results indicate that the proposed LSTM neural network can accurately fit the mechanical model of the robotic catheter, with a root mean square error of 3.44 mN. Finally, we conducted a force control experiment for ultrasound scanning on a silicone liver model. The results indicate that the proposed method can achieve relatively stable force control for laparoscopic ultrasound scanning, with a maximum error of 2.41 mN during the transient phase of system control and a maximum absolute error of 1.8 mN in the steady state. • Contact force influenced by tendon force and robot movement history. • LSTM accurately models tendon force, positioning, and contact force. • LSTM and PD controller manage end contact force via tendon tension.