Anand Malpani1, S. Swaroop Vedula1, Henry C Lin, PhD2, Gregory D Hager, PhD1. 1Johns Hopkins University, 2Intuitive Surgical Inc.
Objectives:
We developed the concept of teaching cues to demonstrate errors and expert behavior, and deficit metrics to measure performance deviations from such ideal behavior with the goal of delivering automated virtual coaching for enhanced learning and skill acquisition.
Methods:
We used the da Vinci® Skills Simulator, a virtual reality (VR) trainer for robot-assisted surgery. We chose the needle passing (NP) task that involved eight NPs around a circle from inside to outside. We used visual feedback in the form of 3D graphics that overlay onto the console viewer of the system. Based on the ACS Surgery Residents Skills Curriculum Phase 1 modules, we targeted the learning elements of needle grasping, positioning and driving, and presented cues on:
- which instrument (left/right) to use for the current NP,
- where and at what angle should the needle be grasped, and
- how the needle should be driven through the tissue.
We measured deficit metrics associated to teaching cues (lower values indicate better performance).
- Grasp Position Deviation (GP) is absolute difference between participant’s grasp position along the needle curvature compared to the ideal position.
- Grasp Orientation Deviation (GO) is absolute difference between participant’s grasp angle and perpendicular direction of the needle’s plane.
- Ideal Path Deviation (in-plane (IPI) and out-of-plane (IPO)) is absolute difference between needle tip trajectory and the ideal needle curvature that should be followed through the tissue.
We conducted a pilot randomized controlled trial at Intuitive Surgical Inc. (Sunnyvale, California), to test the effect of our coaching framework versus independent learning on technical skill. We recruited clinical trainers who oversee the robotic training center and engineers. The experimental group performed three task repetitions along with our teaching cues and control group did so without any teaching cues, in addition to a baseline and final performance. We compared change from baseline for the deficit metrics between groups using a Mann-Whitney U test.
Results:
Of the 32 participants recruited, we excluded data from 2 participants due to technical glitches resulting in 14 (experimental) and 16 (control) subjects. We saw statistically significant difference between the groups in the GO metric (p-value = 0.04; experimental: -14.53; control: -4.22). We did not observe significant differences in other metrics (GP: experimental = -3.73 and control = 4.19; IPI: experimental = 0.00 and control = -0.01; IPO: experimental = -0.01 and control = -0.02).
Conclusions:
Coaching is critical for skill development and facilitates lifelong learning in surgery. Teaching errors and deficits (critique) and showing correct and expert behavior (demonstrate) are core coaching activities. Our study provides proof-of-concept that automated virtual coaching based on targeted teaching cues and objective assessment using deficit metrics is feasible in VR and may be effective in skill acquisition. A larger controlled study in a representative population of surgeons is required to validate our coaching framework.
Acknowledgments:
We are thankful to SenseGraphics AB (Sweden) and Link Foundation Fellowship for Advanced Training and Simulation.
Presented at the SAGES 2017 Annual Meeting in Houston, TX.
Abstract ID: 84306
Program Number: ETP720
Presentation Session: Emerging Technology Poster
Presentation Type: Poster