Eary Diagnosis of Ankylosing Spondylitis using MRI and Transfer Learning

Ankylosing spondylitis (AS) is a chronic inflammatory arthritis that mainly affects the spine and sacroiliac (SI) joints. AS causes severe pain and stiffness, which can greatly limit a person's ability to perform daily activities. Early diagnosis of AS is important for effective treatment and to prevent or delay irreversible damage. However, due to the rarity of the disease and the lack of available data, developing accurate diagnostic tools remains a challenge. AS patients are frequently misdiagnosed due to the lack of awareness about AS and the overlapping symptoms with other common conditions. In this research, we propose a method for the early diagnosis of ankylosing spondylitis using MRI scans of the SI joint. Given the limited dataset available for AS, we use transfer learning techniques to leverage knowledge from a large dataset of sacroiliitis MRI scans, which may not necessarily be caused by AS. Specifically, we fine-tune a pre-trained model, originally trained to classify a diverse set of arthritis cases as sacroiliitis or not, to better recognize AS-specific patterns in the SI joint MRI images. Our approach has achieved an accuracy of 96%. Being able to detect and differentiate AS from other similar conditions, among which it is often misdiagnosed, represents a significant advancement in the diagnosis of AS.