Introduction

This study presents a two-stage deep learning–based approach designed to automate joint localization and damage classification in hand X-rays of patients with rheumatoid arthritis (RA). We developed an innovative two-stage method in which the You Only Look Once v7 (YOLOv7) model is employed for joint localization. This stage incorporates performance enhancements through window-level transformation and the Distance Intersection over Union (DIoU) algorithm. For the second stage, joint damage classification is addressed using a modified EfficientNetV2 model integrated with an attention mechanism to effectively manage data imbalance.

• The effectiveness of our approaches has been demonstrated through the following two publications:
  • Deep Learning with an Attention Mechanism for Enhancing Automated Modified Total Sharp/van der Heijde Scoring of Hand X-ray Images in Rheumatoid Arthritis, Journal of Medical and Biological Engineering (JMBE), 2025.
  • Deep Learning-Based Computer-Aided Diagnosis of Rheumatoid Arthritis with Hand X-ray Images Conforming to Modified Total Sharp/van der Heijde Score, Biomedicines 2022, 10, 1355. https://doi.org/10.3390/biomedicines10061355

Acknowledgments

• This work was primarily supported by the Ministry of Science and Technology, Taiwan.