Quantifying synovitis in rheumatoid arthritis using computer-assisted manual segmentation with 3 tesla MRI scanning

Abstract

To investigate the reliability, validity and feasibility of a computer-assisted manual segmentation method for determining the synovial membrane volume as a surrogate measure for synovitis in patients with rheumatoid arthritis (RA). The 3 Tesla (T) MRI scans were acquired in 22 early RA and 16 established RA patients. Synovial membrane volumes in postcontrast T1w axial images at three wrist joint regions were determined by two nonradiologist observers using a computer-assisted manual segmentation method. Intraobserver reliability, measured by intraclass correlation coefficients (ICCs), was excellent in the early (ICC = 0.99) and established (ICC = 0.99) RA cohorts. Interobserver reliability (mean ICC [95% Confidence Interval]) was moderate to excellent in the early and established RA groups (ICCs = 0.87 [0.68,0.94] and 0.88 [0.66, 0.96], respectively). There was a strong correlation between the synovial membrane volumes derived by segmentation and the RA MRI scoring system (RAMRIS) scores for synovitis at all joints in the early (Spearman rho = 0.86-0.96) and established (Spearman rho = 0.85-0.93) RA cohorts. The entire segmentation technique took 19 to 21 min per patient. Measurement of MRI synovitis using a computer-assisted manual segmentation method demonstrated excellent intraobserver and very good interobserver reliability, content validity (represented by its strong correlation with RAMRIS synovitis), and moderate feasibility.