Automatic detection and classification of knee osteoarthritis using deep learning approach.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Orthopedics

Background:

• Automated analysis of knee osteoarthritis (OA) from digital X-rays remains a challenge.
• Existing diagnostic methods often lack transparency and efficiency.
• Deep learning offers potential for objective OA assessment.

Purpose of the Study:

• To develop and evaluate an AI tool for locating and grading knee OA severity using digital X-ray images.
• To assess the efficacy of deep learning in predicting knee OA based on the Kellgren-Lawrence (KL) grading system.
• To improve the automation and accuracy of knee OA diagnosis.

Main Methods:

• Utilized 3172 anterior-posterior view knee X-ray images from patients over 50 with OA symptoms.
• Employed Faster R-CNN with ResNet-50 for joint space width localization and feature extraction.
• Applied AlexNet with transfer learning for classifying OA severity according to KL grades.

Main Results:

• The AI model achieved a maximum accuracy of 98.516% in identifying minimal knee joint space width (JSW).
• The overall accuracy for classifying knee OA severity reached 98.90%.
• The model's performance improved with network fine-tuning, surpassing existing methods.

Conclusions:

• The developed AI tool demonstrates high accuracy and potential for automated knee OA diagnosis.
• Fine-tuning the deep learning network enhances diagnostic performance.
• Future work will involve extending the model for OA grading in MRI data.