Related Experiment Video
Updated: Aug 14, 2026
08:21
Utilizing a 3D Printed Laparoscopic Nissen Fundoplication Model to Shorten a Resident's Learning Curve
Published on: August 15, 2025
Machine Learning-Based Automated Assessment of Intracorporeal Suturing in Laparoscopic Fundoplication
Shekhar Madhav Khairnar1, Huu Phong Nguyen1, Alexis Desir1
1Department of Surgery, University of Texas Southwestern Medical Center, Dallas, Texas, USA.
Summary
This study introduces an AI-powered tool tracking system for surgical skill assessment, eliminating manual annotation. The AI model accurately classifies surgical performance in laparoscopic procedures, improving feedback for trainees.
Area of Science:
- Artificial Intelligence in Medicine
- Surgical Education Technology
- Computational Anatomy
Background:
- Automated assessment of surgical skills using artificial intelligence (AI) provides valuable, instantaneous feedback for trainees.
- Kinematic metrics derived from bimanual tool motions reliably predict performance in laparoscopic tasks.
- Current automated tool tracking requires time-intensive human annotation, hindering widespread implementation.
Purpose of the Study:
- To develop and evaluate an AI-based tool tracking model using the Segment Anything Model (SAM) to automate surgical skill assessment.
- To eliminate the need for human annotators in the process of automated tool tracking.
- To assess the model's usefulness in evaluating performance during laparoscopic suturing in fundoplication procedures.
Main Methods:
- An AI tool tracking model was applied to videos of Nissen fundoplication performed on porcine models.
- Surgeons were categorized into novice and expert groups.
- Both supervised (using kinematic features and machine learning classifiers) and unsupervised (1-D Convolutional Neural Network) models were implemented and compared.
Main Results:
- The unsupervised learning approach using a 1-D Convolutional Neural Network achieved the highest accuracy (0.817 ± 0.108) and F1 score (0.806 ± 0.110).
- This unsupervised method eliminates the need for manual computation of kinematic features.
- The supervised learning model with Principal Component Analysis and Random Forest achieved an accuracy of 0.795 ± 0.065 and an F1 score of 0.778 ± 0.071.
Conclusions:
- An AI model was successfully developed for automated performance classification from surgical videos, independent of human annotation.
- This AI-driven approach enhances surgical skill evaluation by providing objective and efficient feedback.
- The developed model holds significant potential for improving surgical training and assessment.