Surgeon Training in Telerobotic Surgery via a Hardware-in-the-Loop Simulator

Xiao Li, Homa Alemzadeh, Daniel Chen, Zbigniew Kalbarczyk, Ravishankar K. Iyer, Thenkurussi KesavadasJournal of Healthcare Engineering 2017 · DOI

TL;DR

This paper introduces a hardware-in-the-loop simulator for telerobotic surgery training that goes beyond basic motor skills. It uniquely integrates safety hazard simulation with haptic feedback and an optimal motion planning tool, aiming to better prepare surgeons for real-world critical events and improve surgical efficiency.

What Problem Does This Paper Address?

The increasing adoption of robot-assisted surgery worldwide has highlighted the need for effective training. While existing simulators offer basic motor skill practice, they often lack crucial features for safety training, particularly in simulating and responding to safety-critical events. These events, ranging from system malfunctions to potential cyber-attacks, can have severe consequences for patient safety. Furthermore, current training often focuses on achieving goals rather than optimizing instrument motion, which is vital for efficiency and precision. This paper addresses this gap by developing a simulator that not only trains motor skills but also actively exposes trainees to simulated safety hazards and guides them towards optimal movement patterns.

Methods

The core of this work is a hardware-in-the-loop (HIL) simulator built around the open-source Raven-II surgical robot. Key components and techniques include:

  • Hardware-in-the-Loop (HIL) Integration: The Raven-II robot hardware is integrated into the simulation loop, allowing for a more realistic training experience.
  • Physics Engine (Open Dynamics Engine - ODE): A physics engine is used to simulate the robot's dynamics and its interactions with the environment, providing a virtual representation of the robot's state.
  • Safety Hazard Injection Engine: This novel engine intentionally injects faults into the robot's control software after its automatic safety checks. This is designed to mimic real-world adverse events, including those reported to the FDA MAUDE database, and train surgeons to respond to unexpected system behavior.
  • Haptic Feedback Strategy: A crucial element for safety training, this strategy uses the difference between the simulated robot's state (from the physics engine) and the actual robot's state to generate haptic feedback. This alerts the operator to discrepancies and potential hazards. A spring-damper model is employed for this feedback.
  • Fast Marching Tree (FMT*) Based Motion Planner: This algorithm is used to generate semi-optimal, collision-free motion plans for the robot arms. It helps trainees learn efficient instrument movement patterns, especially in constrained environments. The FMT* algorithm is chosen for its asymptotic optimality and efficiency in high-dimensional spaces.

Datasets

The paper mentions using data from the U.S. Food and Drug Administration (FDA) Manufacturer and User Facility Experience (MAUDE) database to identify common safety hazard scenarios. These scenarios are then used to inform the design of the safety hazard injection engine. The training scenarios themselves are based on models used by the Fundamentals of Robotic Surgery (FRS) organization. No new datasets were introduced for evaluation; rather, existing surgical training models and simulated adverse events were utilized.

Key Insights and Ideas

  • Haptic Feedback for Safety Awareness: The most significant insight is the effective use of haptic feedback not just for palpation, but as a critical safety alert system. By providing tactile cues when the simulated robot's state deviates from the real robot's, the simulator can make trainees aware of impending hazards before they become critical.
  • Simulating "Unseen" Failures: Injecting faults after the robot's built-in safety checks is a clever way to simulate scenarios that might bypass standard safety mechanisms, thus preparing surgeons for more insidious or complex failures.
  • Bridging the Gap Between Skill and Safety: The simulator effectively combines motor skill training with safety hazard response, addressing a critical deficiency in current robotic surgery training programs.
  • Optimal Motion Planning for Efficiency: The integration of FMT* provides a valuable tool for teaching surgeons to move instruments efficiently and precisely, which can lead to improved surgical outcomes and reduced procedure times.

Key Results

  • Successful Hazard Simulation: The simulator successfully reproduced safety hazard events, specifically demonstrating a "sudden jump" scenario by injecting faults into motor commands.
  • Effective Haptic Feedback: Haptic feedback was shown to be triggered precisely when faults were injected, providing timely alerts to the operator. The force feedback was designed to be noticeable but not interfere with normal operation.
  • Motion Planning Performance: The FMT* planner demonstrated its ability to generate collision-free paths. Performance was evaluated based on planning time, path tracking time, number of waypoints, and success rate. For instance, with 1000 samples, planning time was around 47 seconds, path tracking time was about 4 seconds, and the success rate was 85%. Increasing samples improved path quality but increased planning time.
  • Model Verification: The physics engine model was validated against real robot data, showing good agreement in joint profiles and end-effector position errors.

Limitations and Future Work

  • Limitations:
    • The paper acknowledges that perfect transparency in haptic feedback is not possible, especially with complex robot dynamics.
  • The ODE simulation models only 7 degrees of freedom and assumes decoupled joint motion, ignoring cable coupling complexities.
  • The current FMT* implementation assumes uniform sampling in feasible joint space and doesn't explicitly maximize obstacle clearance, which could be improved.
  • The haptic feedback saturation range needs careful tuning to avoid interfering with normal operation.
  • Future Work:
    • Providing haptic feedback to guide the operator along pre-planned optimal paths for specific tasks (e.g., grasping).
  • Incorporating visual or audio cues to alert the operator when approaching workspace boundaries of the haptic device.
  • Further research into human factors to optimize the delivery and interpretation of haptic safety alarms.
  • Exploring additional cost function terms for the path planner, such as maximizing obstacle clearance.

Summary updated

Abstract

This work presents a software and hardware framework for a telerobotic surgery safety and motor skill training simulator. The aims are at providing trainees a comprehensive simulator for acquiring essential skills to perform telerobotic surgery. Existing commercial robotic surgery simulators lack features for safety training and optimal motion planning, which are critical factors in ensuring patient safety and efficiency in operation. In this work, we propose a hardware-in-the-loop simulator directly introducing these two features. The proposed simulator is built upon the Raven-II™ open source surgical robot, integrated with a physics engine and a safety hazard injection engine. Also, a Fast Marching Tree-based motion planning algorithm is used to help trainee learn the optimal instrument motion patterns. The main contributions of this work are (1) reproducing safety hazards events, related to da Vinci™ system, reported to the FDA MAUDE database, with a novel haptic feedback strategy to provide feedback to the operator when the underlying dynamics differ from the real robot's states so that the operator will be aware and can mitigate the negative impact of the safety-critical events, and (2) using motion planner to generate semioptimal path in an interactive robotic surgery training environment.
BibTeX
@article{li2017surgeon,
  author = {Xiao Li and Homa Alemzadeh and Daniel Chen and Zbigniew Kalbarczyk and Ravishankar K. Iyer and Thenkurussi Kesavadas},
  title = {{Surgeon Training in Telerobotic Surgery via a Hardware-in-the-Loop Simulator}},
  journal = {Journal of Healthcare Engineering},
  year = {2017},
  doi = {10.1155/2017/6702919},
}
Figure