Related Experiment Video
Updated: Aug 30, 2025
06:31
Author Spotlight: Enhancing Rheumatoid Arthritis Research Through HR-pQCT Imaging Analysis
Published on: October 6, 2023
2.3K
A Crowdsourcing Approach to Develop Machine Learning Models to Quantify Radiographic Joint Damage in Rheumatoid
Dongmei Sun1,2, Thanh M Nguyen1, Robert J Allaway3
1University of Alabama at Birmingham, Birmingham.
JAMA Network Open
|August 29, 2022
Summary
The RA2-DREAM Challenge developed AI algorithms to accurately quantify rheumatoid arthritis (RA) joint damage from X-rays. These tools can aid clinical trials and patient treatment decisions by providing rapid, reliable assessments.
Area of Science:
- Rheumatology
- Medical Imaging
- Artificial Intelligence
Background:
- Accurate quantification of joint damage in rheumatoid arthritis (RA) is crucial for clinical trials and treatment decisions.
- Current methods for assessing joint space narrowing and erosions are often manual and time-consuming.
- Automated, unbiased assessment of radiographic damage is needed for efficient patient management and research.
Purpose of the Study:
- To develop machine learning methods for quantifying radiographic damage in rheumatoid arthritis (RA).
- To catalyze the development of automated assessment tools through an international crowdsourcing competition.
- To improve the accuracy and efficiency of joint damage quantification in RA.
Main Methods:
- The Rheumatoid Arthritis 2-Dialogue for Reverse Engineering Assessment and Methods (RA2-DREAM) Challenge was designed as an international crowdsourcing competition.
- Participants used existing radiographic images and expert-curated Sharp-van der Heijde (SvH) scores for training and evaluation.
- Algorithms were developed to automatically quantify overall damage, joint space narrowing, and erosions.
Main Results:
- The RA2-DREAM Challenge attracted 173 submissions from 26 international teams.
- Winning algorithms demonstrated high accuracy, with scores closely matching expert-curated Sharp-van der Heijde scores.
- Reproducibility and concordance were confirmed through bootstrapping, Bayes factors, and independent validation, with concordance indices ranging from 0.71 to 0.82.
Conclusions:
- The RA2-DREAM Challenge successfully developed feasible, quick, and accurate algorithms for quantifying RA joint damage.
- These automated methods have the potential to significantly benefit RA research and clinical practice.
- Integration into electronic health records can provide clinicians with timely, quantitative data for improved treatment decisions.
