MRI and DECT images were separately anonymized and scored using Horos (The Horos Project, Version 3.3.6, Pureview, MD, USA) by two readers: reader 1 (TD), a musculoskeletal radiologist with 13 years of experience and reader 2 (DD), a research student with 3 years of experience in musculoskeletal imaging. The readers were blinded to identifying information and the images and results of the other imaging modality.
For DECT, the readers had access to 120 kV-equivalent and VNCa images and were allowed to freely adjust the window levelling and to invert the images. For MRI, T1-weighted and STIR images were scored. Only images of the sacroiliac joints were shown, while spinal images were not included in the image stack. The two readers used an established 24-region scoring model of the SIJs, which divides each joint into 4 anterior, 4 middle, and 4 posterior quadrants [6 (link)]. Bone marrow oedema and fatty bone marrow deposition were scored as followed: 0: absent; 1: <33% of the quadrant, 33–66% of the quadrant; ≥66% of the quadrant. On VNCa images, osteitis was defined as a hyperdense subchondral lesion of the bone marrow adjacent to the cartilaginous joint surface. In contrast, fat lesions were defined as hypodense compared with normal bone marrow attenuation. Sclerosis was assessed on a three-point scale: 0: no sclerosis; 1: possible/little sclerosis; 2: marked sclerosis. Disagreements between the two readers were solved by two different experts: an expert with 7 years of experience in musculoskeletal imaging solved disagreements in the interpretation of DECT images and another expert with 18 years of experience in musculoskeletal imaging disagreements in MRI interpretation.
Free full text: Click here