Related Experiment Video
Updated: Feb 20, 2026
Utilizing a 3D Printed Laparoscopic Nissen Fundoplication Model to Shorten a Resident's Learning Curve
Published on: August 15, 2025
Automatically rating trainee skill at a pediatric laparoscopic suturing task
Yousi A Oquendo1,2, Elijah W Riddle3, Dennis Hiller3
1Department of Mechanical Engineering & Applied Mechanics, University of Pennsylvania, Philadelphia, USA.
Background:
Minimally invasive surgeons must acquire complex technical skills while minimizing patient risk, a challenge that is magnified in pediatric surgery. Trainees need realistic practice with frequent detailed feedback, but human grading is tedious and subjective. We aim to validate a novel motion-tracking system and algorithms that automatically evaluate trainee performance of a pediatric laparoscopic suturing task.
Methods:
Subjects (n = 32) ranging from medical students to fellows performed two trials of intracorporeal suturing in a custom pediatric laparoscopic box trainer after watching a video of ideal performance. The motions of the tools and endoscope were recorded over time using a magnetic sensing system, and both tool grip angles were recorded using handle-mounted flex sensors. An expert rated the 63 trial videos on five domains from the Objective Structured Assessment of Technical Skill (OSATS), yielding summed scores from 5 to 20. Motion data from each trial were processed to calculate 280 features. We used regularized least squares regression to identify the most predictive features from different subsets of the motion data and then built six regression tree models that predict summed OSATS score. Model accuracy was evaluated via leave-one-subject-out cross-validation.
Results:
The model that used all sensor data streams performed best, achieving 71% accuracy at predicting summed scores within 2 points, 89% accuracy within 4, and a correlation of 0.85 with human ratings. 59% of the rounded average OSATS score predictions were perfect, and 100% were within 1 point. This model employed 87 features, including none based on completion time, 77 from tool tip motion, 3 from tool tip visibility, and 7 from grip angle.
Conclusions:
Our novel hardware and software automatically rated previously unseen trials with summed OSATS scores that closely match human expert ratings. Such a system facilitates more feedback-intensive surgical training and may yield insights into the fundamental components of surgical skill.
