Quantifying bone marrow inflammatory edema in the spine and sacroiliac joints with thresholding
Journal article, Peer reviewed
MetadataShow full item record
Original versionBMC Musculoskeletal Disorders. 2017, 18 (1), . 10.1186/s12891-017-1861-1
Background Psoriatic Arthritis (PsA) is a chronic inflammatory arthritis that develops in patients with psoriasis. Inflammatory edema in the spine may reflect subclinical disease activity and be a predictor of radiographic progression. A semi-quantitative method established by the spondyloarthritis research consortium of Canada (SPARCC) is commonly used to assess the disease activity in MR images of the spine. This study aims to evaluate thresholding for quantification of subtle bone marrow inflammation in the spine and the sacroiliac (SI) joints of patients with PsA and compare it with the SPARCC scoring system. Methods Short tau inversion recovery (STIR) MR images of the spine (N = 85) and the SI joints (N = 95) of patients with PsA (N = 41) were analyzed. A threshold was applied to visible bone marrow in order to mask areas with higher signal intensity, which are consistent with inflammation. These areas were considered as inflammatory lesions. The volume and relative signal intensity of the lesions were calculated. Results from thresholding were compared to SPARCC scores using linear mixed-effects models. The specificity and sensitivity of thresholding were also calculated. Results A significant positive correlation between the volumes and mean relative signal intensities, which were calculated by thresholding analysis, and the SPARCC scores was detected for both spine (p < 0.001) and SI joints (p < 0.001). For the spine, thresholding had sensitivity and specificity of 83% and 76% respectively, while for the SI joints the values were 51% and 88% respectively. Conclusions Thresholding allows quantification of subtle bone marrow inflammatory edema in patients with psoriatic arthritis, and could support SPARCC scoring of the spine. Improved image processing and inclusion of automatic segmentation are required for thresholding of STIR images to become a rapid and reliable method for quantitative measures of inflammation.