ABSTRACT
AIM: To evaluate the effect of subchondral oedema in T2-weighted Dixon magnetic resonance imaging (MRI) sequence evaluation of sacroiliac joint erosion in patients with axial spondyloarthropathy. MATERIALS AND METHODS: Twenty patients diagnosed with axial spondyloarthritis underwent MRI at a tertiary referral centre from December 2019 to March 2021 were included. In-phase, opposed-phase and fat-only images were scored by two musculoskeletal radiologists independently for the presence of erosions in eight sacroiliac joint quadrants. Sensitivity, specificity and areas under the curve (AUC) of the receiver operating characteristic curve were determined using T1W sequence as reference standard. Intra-observer and interobserver reliability were calculated using Cohen's kappa coefficient. RESULTS: The diagnostic performance of fat-only and in-phase images were similar (AUC 0.857-0.902 and 0.828-0.868) and better than opposed-phase images (AUC 0.613-0.658). The interobserver reliability of fat-only and in-phase images were substantial (k = 0.747 and 0.712), and moderate for opposed-phase images (k = 0.417). Intra-observer reliability was almost perfect for all the images. In the subgroup analysis, the specificity and AUC for oedema-positive group were lower than oedema-negative group in all image sets. Interobserver reliability was substantial for fat-only and in-phase images in both groups, but slight and moderate for the opposed-phase oedema-positive and negative groups, respectively. CONCLUSION: The presence of subchondral oedema in active sacroiliitis decreased the diagnostic accuracy of sacroiliac joint erosion detection on T2W Dixon MRI images.