Estimating rheumatoid arthritis activity with infrared image analysis

Area of Science:

• Medical imaging
• Rheumatology
• Artificial intelligence in medicine

Background:

• Rheumatoid arthritis (RA) diagnosis and severity assessment are crucial for effective patient management.
• Previous research indicated temperature differences in hand and knee joints of RA patients compared to healthy individuals.

Purpose of the Study:

• To develop and validate a novel infrared imaging-based diagnostic tool for rheumatoid arthritis (RA) severity assessment.
• To identify new joint areas, specifically ankles, as significant indicators for RA assessment.
• To classify RA severity into High, Medium, and Low levels using machine learning.

Main Methods:

• Infrared image collection from RA patients and healthy subjects.
• Statistical analysis of joint temperature distribution.
• Classification of RA severity using a C5.0 decision tree algorithm.
• Evaluation of classifier performance using sensitivity and specificity metrics.

Main Results:

• Ankle joint temperature distribution was found to be statistically significant for RA assessment.
• The C5.0 classifier achieved high accuracy in categorizing RA severity.
• Achieved 96% sensitivity and 92% specificity in classifying RA severity levels.

Conclusions:

• Infrared thermography combined with machine learning offers a promising, non-invasive method for RA severity assessment.
• The developed tool demonstrates high diagnostic accuracy, potentially improving clinical decision-making for rheumatoid arthritis.
• Future research will focus on automating image analysis and clinical validation against MRI as the gold standard.