Related Experiment Video
Updated: Dec 25, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.3K
Evaluation of Deep Learning Models for Identifying Surgical Actions and Measuring Performance
Shuja Khalid1, Mitchell Goldenberg1, Teodor Grantcharov1
1Surgical Safety Technologies, Toronto, Ontario, Canada.
JAMA Network Open
|April 1, 2020
Summary
Deep learning models accurately assess surgical skills from video clips, categorizing actions and competence levels. This technology offers a new way to provide surgeons with objective feedback for skill improvement.
Area of Science:
- Medical technology
- Artificial intelligence
- Surgical education
Background:
- Surgical performance evaluation relies on subjective video review, leading to data loss and potential errors.
- Objective assessment of surgical skills is crucial for improving patient safety and training efficacy.
Purpose of the Study:
- To evaluate a deep learning framework for assessing surgical video clips.
- To categorize videos by surgical step and surgeon competence level.
Main Methods:
- Trained deep learning models on 103 video clips of surgeons performing knot tying, suturing, and needle passing.
- Utilized precision, recall, and accuracy to measure model performance in estimating surgical actions and skill levels.
Main Results:
- Models achieved high accuracy in identifying surgical actions (mean precision 0.97, mean recall 0.98).
- Performance level estimation showed mean precision of 0.77 and mean recall of 0.78.
- The framework demonstrated effectiveness using only video input.
Conclusions:
- Deep machine learning can effectively analyze surgical video clips to identify skill-related patterns.
- This approach represents a significant step towards automated, objective feedback systems for surgical training.
- The technology has the potential to enhance surgical skill refinement and patient safety.
