Autonomous Soft Tissue Retraction Using Demonstration-Guided Reinforcement Learning

TL;DR

This paper develops a simulation environment and applies demonstration-guided reinforcement learning to enable autonomous soft tissue retraction with surgical robots, demonstrating feasibility and laying groundwork for future soft tissue manipulation research.

Contribution

It introduces a ROS-compatible physics simulation for soft tissue interaction and evaluates demonstration-guided RL for soft tissue retraction, a novel approach in surgical robotics.

Findings

Proof-of-concept for autonomous soft tissue retraction

Demonstration-guided RL outperforms traditional RL in this task

Simulation environment supports both rigid and soft body interactions

Abstract

In the context of surgery, robots can provide substantial assistance by performing small, repetitive tasks such as suturing, needle exchange, and tissue retraction, thereby enabling surgeons to concentrate on more complex aspects of the procedure. However, existing surgical task learning mainly pertains to rigid body interactions, whereas the advancement towards more sophisticated surgical robots necessitates the manipulation of soft bodies. Previous work focused on tissue phantoms for soft tissue task learning, which can be expensive and can be an entry barrier to research. Simulation environments present a safe and efficient way to learn surgical tasks before their application to actual tissue. In this study, we create a Robot Operating System (ROS)-compatible physics simulation environment with support for both rigid and soft body interactions within surgical tasks. Furthermore, we investigate the soft tissue interactions facilitated by the patient-side manipulator of the DaVinci surgical robot. Leveraging the pybullet physics engine, we simulate kinematics and establish anchor points to guide the robotic arm when manipulating soft tissue. Using demonstration-guided reinforcement learning (RL) algorithms, we investigate their performance in comparison to traditional reinforcement learning algorithms. Our in silico trials demonstrate a proof-of-concept for autonomous surgical soft tissue retraction. The results corroborate the feasibility of learning soft body manipulation through the application of reinforcement learning agents. This work lays the foundation for future research into the development and refinement of surgical robots capable of managing both rigid and soft tissue interactions. Code is available at https://github.com/amritpal-001/tissue_retract.