A New Tensioning Method using Deep Reinforcement Learning for Surgical Pattern Cutting

Abstract

Surgeons normally need surgical scissors and tissue grippers to cut through a deformable surgical tissue. The cutting accuracy depends on the skills to manipulate these two tools. Such skills are part of basic surgical skills training as in the Fundamentals of Laparoscopic Surgery. The gripper is used to pinch a point on the surgical sheet and pull the tissue to a certain direction to maintain the tension while the scissors cut through a trajectory. As the surgical materials are deformable, it requires a comprehensive tensioning policy to yield appropriate tensioning direction at each step of the cutting process. Automating a tensioning policy for a given cutting trajectory will support not only the human surgeons but also the surgical robots to improve the cutting accuracy and reliability. This paper presents a multiple pinch point approach to modelling an autonomous tensioning planner based on a deep reinforcement learning algorithm. Experiments on a simulator show that the proposed method is superior to existing methods in terms of both performance and robustness.

Performance and Capability Assessment in Surgical Subtask Automation

“…Attanasio et al [ 21 ] utilized visible area as outcome metric in autonomous retraction. Nguyen et al [ 30 ] measured the accuracy of pattern cutting next to autonomous tensioning. In the study of Shademan et al [ 14 ], autonomous end-to-end anastomosis is presented and validated in vivo on porcine, where number of sutures , number of suturing mistakes , leak pressure , luminal diameter reduction , weight at surgery , and weight at sacrifice are measured and compared to manual execution.…”

“…Currently, many research groups are working on this problem [ 12 , 13 ]; some groups chose to work in ex vivo (or rarely in vivo) [ 14 , 15 ] or realistic phantom environments [ 16 ], but simplified silicone phantoms are utilized mostly [ 15 , 17 , 18 , 19 , 20 , 21 ]. In the most recent years, the automation of simple surgical training exercises on rigid [ 22 , 23 , 24 , 25 , 26 , 27 , 28 ] or deformable [ 29 , 30 ] phantoms tends to receive increasing attention. Among all the training exercises, the automation of different versions of peg transfer is presented in the most significant number of studies [ 22 , 23 , 24 , 25 , 26 , 27 , 28 , 31 ], probably due to its simplicity, enabling to elaborate the basic principles and best algorithms for automation.…”

A Review on Deep Learning in Minimally Invasive Surgery

“…They implemented this approach using a dVRK in both simulated and physical scenarios. In later works, this method is improved with a multiple pinch point Deep RL algorithm that exhibits better results [93]- [94].…”

Real-to-Sim Registration of Deformable Soft Tissue with Position-Based Dynamics for Surgical Robot Autonomy

“…However, FEM as a general strategy has a significant problem in that one cannot explicitly apply position constraints on the simulation easily, so the registration between real world and simulation cannot be explicitly defined. Works in autonomous debridement [12] and tissue tensioning [13], [14] used learning method to identify proper tissue properties from visual input. However, none of these works considered the physical dynamics explicitly.…”