Deep learning-based joint detection in Rheumatoid arthritis hand radiographs

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Rheumatology

Background:

• Rheumatoid arthritis is an autoimmune disease impacting multiple joints.
• Early and accurate diagnosis is crucial for managing rheumatoid arthritis.
• Technological advancements offer new tools for disease detection.

Purpose of the Study:

• To evaluate the efficacy of YOLOv5l6 for detecting rheumatoid arthritis in radiograph images.
• To assess the model's performance with limited training samples.
• To improve the generalization of deep learning models for rheumatoid arthritis detection.

Main Methods:

• Utilized the YOLOv5l6 deep learning model for joint detection in X-ray images.
• Trained the YOLOv5l6 model on radiograph images of healthy individuals.
• Implemented data augmentation techniques to enhance model generalization.

Main Results:

• YOLOv5l6 achieved high performance in detecting rheumatoid arthritis joints.
• The model demonstrated effectiveness even with a limited number of training samples.
• Data augmentation further improved the model's ability to generalize.

Conclusions:

• YOLOv5l6 is a promising tool for the efficient detection of rheumatoid arthritis in radiograph images.
• Training on healthy joint images can yield effective diagnostic performance for rheumatoid arthritis.
• Deep learning models, enhanced by data augmentation, show potential in rheumatology diagnostics.