Axial spondyloarthritis (SpA) is the most common inflammatory rheumatism in young men [1]. Axial SpA is still relatively rare, affecting approximately 0.6% of the population to a varying degree of severity. Nonetheless, treatments can improve symptoms and prevent worsening of axial SpA if an early correct diagnosis is obtained [2]. However, this early diagnosis of axial SpA requires extensive training and is challenging for both radiologists and rheumatologists.

Magnetic resonance imaging (MRI) of the sacroiliac joints (SIJ) is an essential tool for the diagnosis of SpA [3]. However, despite a clear definition of positive and negative findings developed by the ASAS/OMERACT (Assessment of SpondyloArthritis International Society/Outcome Measures in Rheumatology) working group [4], inter-reader agreement is limited even between experienced physicians [5]. Moreover, in daily practice, MRI examinations of SIJ are often interpreted by general radiologists with no specific training in musculoskeletal imaging.

In recent years, many studies have showed a growing interest for the application of artificial intelligence (AI) as a help to identify abnormal findings on imaging examinations [6]. In this regard, diagnosis through computerized image analysis using machine learning, and particularly deep learning is now possible [7,8]. These systems can recognize patterns and may improve diagnosis workflow efficiency in several fields of musculoskeletal imaging [9,10], Such an approach could be useful in the analysis of MRI examinations of SIJ in patients with axial SpA [11].

The purpose of this study was to develop and evaluate a deep learning model to detect bone marrow edema of SIJ and predict the MRI ASAS definition of active sacroiliitis in patients with chronic inflammatory back pain.