BackgroundFindings of fatty lesions in the context of other imaging manifestations, especially bone marrow edema and erosions can effectively assist in the diagnosis of axSpA. Chemical shift-encoded MRI is a sequence which allows for the quantification of fat signal and has been applied in the imaging evaluation of the SIJ in axSpA. The objective of this study was to investigate the diagnostic performance of morphological features of fatty lesions visualized by CSE-MRI in the imaging evaluation of SIJ in axSpA.MethodsFatty lesions with morphological features (subchondral, homogeneity and distinct border) were assessed and recorded as a binary variable in each quadrant of the SIJ. Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were calculated for different morphological features as well as the anatomical distribution in patients with nr-axSpA and r-axSpA. T1-weighted images and CSE-MRI fat fraction maps were directly compared in the recognition of different morphological features.ResultsEighty-two patients [non-SpA (n = 21), nr-axSpA (n = 23), r-axSpA (n = 38)] with lower back pain (LBP) were enrolled. Presence of the three morphological features of fatty lesions had a specificity of 90.48% in axSpA. The sensitivities of being subchondral, homogeneity and distinct border were 52.17, 39.13 and 39.13% in nr-axSpA on T1-weighted images. For patients with r-axSpA, the sensitivities reached 86.84, 76.32 and 57.89%. No significant difference was found in the distribution of fatty lesions between T1-weighted images and CSE-MRI. However, CSE-MRI fat fraction maps could detect significantly more fatty lesions with homogeneity (p = 0.0412) and distinct border (p = 0.0159) than T1-weighted images in the sacroiliac joint, but not subchondral lesions (p = 0.6831).ConclusionThe homogeneity and distinct border are more relevant for the diagnosis of axSpA. Moreover, CSE-MRI could detect more typical morphological features of fatty lesions than T1-weighted images in showing these two features. The presence of all three features was more likely to be indicative of axSpA.