Automated quantification of wrist bone marrow oedema, pre- and post-treatment, in early rheumatoid arthritis

• DOI: 10.1093/rap/rkae073
• Publication Date: 2024-06-21T06:16:00Z
• Authors: Chungwun Yiu; James Francis Gri...; Fan Xiao; Lin Shi; Bingjing Zhou; Su Wu; Lai-Shan Tam
• Keywords: 03 medical and health sciences; 0302 clinical medicine; Original Article

Abstract Objective Bone inflammation (osteitis) in early RA (ERA) manifests as bone marrow oedema (BME) and precedes the development of erosion. In this prospective, single-centre study, we developed an automated post-processing pipeline for quantifying severity wrist BME on T2-weighted fat-suppressed MRI. Methods A total 80 ERA patients [mean age 54 years (s.d. 12), 62 females] were enrolled at baseline 49 (40 females) after 1 year treatment. For segmentation, a framework based convolutional neural network (nnU-Net) was trained validated (5-fold cross-validation) 15 areas 60 patients. quantification, identified by Gaussian mixture model clustering thresholding. proportion (%) relative intensity within each area compared with visual semi-quantitative assessment MRI score (RAMRIS). Results overall Sørensen–Dice similarity coefficient 0.91 0.02) ground truth manual segmentation. High correlation (Pearson r = 0.928, P < 0.001) between RAMRIS found. The decreased treatment, correlating highly (r 0.852, reduction score. Conclusion had excellent segmentation performance reliable quantification both patients, providing more objective efficient alternative to scoring.