The Learning Plateau and the Learning Rate for

The Learning Plateau and the Learning Rate for the VBLa. ST PT© compared to the FLS simulator Ganesh Sankaranarayanan Ph. D April 24, 2013 Orlando/ASE 2013

Introduction - The Virtual Basic Laparoscopic Skills Trainer (VBLa. ST©) is a virtual reality simulator that is capable of simulating the Fundamentals of Laparoscopic Surgery (FLS) tasks. - Has a custom interface with haptic (force) feedback capabilities. - Can compute scores automatically - No need for proctors - No need to replenish materials - Additional performance measures can be measured/coded any time cemsim. rpi. edu

VBLa. ST System VBLa. ST PC© FLS and VBLa. ST PT© system VBLa. ST LP©

VBLa. ST PT© § Can simulate the peg transfer task § The simulator has shown - Concurrent validity - Convergent validity

Learning Curve Study ( Convergent Validity) § Three groups - Control (no training) - VBLa. ST - FLS § 15 sessions (10 trials each session) - 5 days x 3 weeks - Pre-test, post-test, retention test (2 weeks after post test) § Normalized numerical score based on completion time and errors were calculated for both the systems § 18 medical students from the Tufts University School of Medicine were recruited in this IRB approved study. § Cumulative Summation Method (CUSUM) was used for assessing the learning curve of both VBLa. ST and the FLS systems.

Need for Learning Plateau and the Learning Rate § CUMSUM method is criterion based - Junior, intermediate, senior - MISTELS (Fraser et al. ) - VBLa. ST (Zhang et al. ) - Can track performance with every single trial § Learning curve has three distinct parameters (Cook et al. ) - Starting point ( where the performance starts) - The plateau ( where the performance flattens) - Learning rate ( how fast the performance level is reached) § The parameters are intuitive and easy to relate scores to performance cemsim. rpi. edu

Inverse Curve Fitting § § § Inverse curve Y = a – b/X a is theoretical maximum score b is the slope b/a is the rate 10 * b/a was defined as the number of trials to reach 90% of the asymptote § First defined and used for learning curve by Feldman et al. § Parameters computed using nonlinear regression § IBM PASW 18 was used for analysis

Results - Curve Fitting VBLa. ST PT© FLS

Results – Learning Curve Parameters Simulator Mean Starting Score Learning Plateau (a) (Mean ± Std) Learning Rate (10 * b/a) (Mean ± Std) VBLa. ST PT© 44. 5 ± 10. 51 94. 03 ± 3. 11 11 ± 3 FLS PT task 56. 42 ± 15. 11 94. 97 ± 1. 74 7± 3 • Both simulators achieved a stabilized higher scores by the end of 150 th trial

Learning in VBLa. ST P < 0. 00001 (pre and post test)

Discussion § Inverse curve fitting showed stable plateaus for both the simulators § Learning rate was lower in VBLa. ST compared to FLS - Similarly the CUSUM analysis also showed more number of trials to achieve the Junior, Intermediate and senior levels § VBLa. ST is a virtual reality simulator - Still requires some adaptation by users, especially when used for first time - Other solutions that are being currently implemented in the second generation of the VBLa. ST simulators are - Workspace matching - Tool peg interactions ( picking and transfer) as realistic to the FLS cemsim. rpi. edu

Acknowledgments § Funding from NIH, NIH/NIBIB 5 R 01 EB 010037 § Likun Zhang for conducting the study at the Tufts University School of Medicine cemsim. rpi. edu