Related Experiment Video
Updated: Oct 1, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.0K
Determining the anatomical site in knee radiographs using deep learning.
Anton S Quinsten1, Lale Umutlu2, Michael Forsting2
1Institute of Diagnostic and Interventional Radiology and Neuroradiology, University Hospital Essen, Hufelandstr. 55, 45147, Essen, Germany. anton.quinsten@uk-essen.de.
Scientific Reports
|March 8, 2022
Summary
Deep learning accurately identifies the anatomical side in knee radiographs. This artificial intelligence approach matches the high accuracy of human radiographers in marking knee sides on X-rays.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Radiology
Background:
- Correct anatomical side marking is crucial for radiograph quality.
- Ensuring accurate side identification in knee radiographs is essential for diagnosis and treatment planning.
Purpose of the Study:
- To evaluate a deep neural network for predicting the anatomical side in anterior-posterior knee radiographs.
- To assess the accuracy of deep learning in identifying left or right knee sides from X-ray images.
Main Methods:
- A ResNet-34 deep neural network was trained on 2892 knee radiographs.
- The model was validated on an internal cohort (932 radiographs) and an external cohort (490 radiographs).
Main Results:
- The deep learning network achieved 99.8% accuracy on the internal validation cohort.
- The network demonstrated 99.9% accuracy on the external validation cohort.
- Performance was comparable to that of experienced radiographers.
Conclusions:
- Deep learning effectively predicts the anatomical side in anterior-posterior knee radiographs.
- AI-driven analysis offers a highly accurate method for anatomical side determination in knee X-rays.
