A New Infrared Thermography Based Approach to Ankylosing Spondylitis Assessment: A Preliminary Study

The clinical assessment of ankylosing spondylitis (AS) traditionally relies on hospital-based procedures, imposing a significant time and economic burden on patients. To address this challenge, this study proposes a non-contact, home-based method for assessing AS status by analyzing infrared thermal images of the knee. Specifically, we propose an Adaptive Weight U-Net (AWU) framework to compute the temperature difference ( $\Delta T$ ) between the knee joint and peripheral skin. This $\Delta T$ is used to classify healthy individuals, patients with inactive AS, and patients with active AS. We further investigate the correlation between knee inflammation and AS status by evaluating the relationship between $\Delta T$ and two key clinical inflammatory markers, C-reactive protein (CRP) and erythrocyte sedimentation rate (ESR). In a dataset of 221 participants (72 healthy, 77 inactive AS, 72 active AS), the framework achieves a classification accuracy of $77.38\%$ , with perfect discrimination between healthy and active AS cases. Kendall's correlation analysis reveals a moderate positive correlation between $\Delta T$ and CRP ( $\tau =0.5920$ , $p<0.001$ ) and a weak positive correlation with ESR ( $\tau =0.4236$ , $p<0.001$ ). The results indicate the potential of infrared thermography as a complementary tool for AS assessment and home monitoring.