Related Experiment Video
Updated: Aug 20, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.9K
Automated Detection of Surgical Implants on Plain Knee Radiographs Using a Deep Learning Algorithm.
Back Kim1, Do Weon Lee2, Sanggyu Lee3
1College of Medicine, Seoul National University, Seoul 03080, Republic of Korea.
Medicina (Kaunas, Lithuania)
|November 24, 2022
Summary
A new deep learning algorithm can automatically detect 17 types of knee surgical implants on X-rays. This AI tool shows high accuracy, aiding clinical decisions for patients with multiple knee operations.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Orthopedics
Background:
- Increasing number of patients require multiple knee surgeries.
- Need for efficient methods to identify surgical implants on radiographs.
Purpose of the Study:
- Develop a deep learning algorithm for detecting 17 types of knee surgical implants.
- Evaluate the algorithm's performance on plain knee radiographs.
Main Methods:
- Utilized a dataset of 5206 internal and 238 external knee X-rays.
- Employed a You Only Look Once (YOLO) deep learning network.
- Trained and tested the model on various implant types, including total knee arthroplasty.
Main Results:
- Achieved high accuracy (up to 0.978) and specificity (up to 0.999) across internal and external test sets.
- Demonstrated a high true positive rate (>0.99 internally, 0.96 externally) for total knee arthroplasty detection.
- The algorithm effectively handled overlapping implant cases.
Conclusions:
- Deep learning can automate surgical implant identification on knee radiographs.
- The developed algorithm shows promise for clinical applications and future AI-driven image analysis.
- This technology can support faster clinical decision-making in orthopedic care.
