Related Experiment Video
Updated: Jul 10, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.8K
Enchondroma Detection from Hand Radiographs with an Interactive Deep Learning Segmentation Tool-A Feasibility Study.
Turkka Tapio Anttila1, Samuli Aspinen1, Georgios Pierides1
1Musculoskeletal and Plastic Surgery, Department of Hand Surgery, University of Helsinki and Helsinki University Hospital, 00029 Helsinki, Finland.
Journal of Clinical Medicine
|November 25, 2023
Summary
This study developed a deep learning model to detect enchondromas, common benign bone tumors, in hand radiographs. The AI tool shows promise for improving early detection and diagnosis of these often-missed conditions.
Area of Science:
- Radiology
- Artificial Intelligence
- Oncology
Background:
- Enchondromas are common benign bone tumors, frequently affecting the hand.
- Symptoms include swelling and pain, but they are often asymptomatic and go unnoticed.
- Expansion can weaken bone, increasing fracture risk, and diagnosis can be challenging, especially in trauma cases.
Purpose of the Study:
- To evaluate the efficacy of a deep learning model in detecting enchondromas from hand radiographs.
- To assess the potential of AI in improving the diagnostic accuracy of enchondromas.
Main Methods:
- A deep learning model was trained using 414 hand radiographs featuring enchondromas.
- Performance was evaluated on a separate test set of 131 radiographs.
- Ground truth was established through expert annotation by three clinical specialists.
Main Results:
- The deep learning model successfully detected 56 out of 62 enchondromas in the test set.
- Achieved an area under the receiver operator curve (AUC) of 0.95.
- Obtained an F1 score of 69.5% for area statistical overlapping.
Conclusions:
- The developed deep learning model demonstrates significant potential as a screening tool for hand enchondromas.
- AI may assist radiologists in raising suspicion and improving the detection of these benign bone tumors.
- Further validation could integrate this tool into routine radiographic interpretation workflows.
