MRI of small bowel Crohn's disease: determining the reproducibility of bowel wall gadolinium enhancement measurements

This study aims to determine inter- and intra-observer variation in MRI measurements of relative bowel wall signal intensity (SI) in Crohn's disease. Twenty-one small bowel MRI examinations (11 male, mean age 40), including T1-weighted acquisitions acquired 30 to 120s post-gadolinium, were analysed. Maximal bowel wall SI (most avid, conspicuous contrast enhancement) in designated diseased segments was measured by two radiologists and two trainees using self-positioned "free" regions of interest (ROIs) followed by "fixed" ROIs chosen by one radiologist, and this procedure was repeated 1 month later. Relative enhancement (post-contrast SI minus pre-contrast SI/pre-contrast SI) was calculated. Data were analysed using Bland-Altman limits of agreement and intra-class correlation. Inter-observer agreement for relative enhancement was poor (spanning over 120%) using a free ROI-95% limits of agreement -0.69, 0.70 and -0.47, 0.74 for radiologists and trainees, respectively, only marginally improved by use of a fixed ROI -0.60, 0.67 and -0.59, 0.49. Intra-class correlation ranged from 0.46 to 0.72. Intra-observer agreement was slightly better and optimised using a fixed ROI-95% limits of agreement -0.52, 0.50 and -0.34, 0.28 for radiologists and trainees, respectively. Intra-class correlation ranged from 0.49 to 0.86. Relative bowel wall signal intensity measurements demonstrate wide limits of observer agreement, unrelated to reader experience but improved using fixed ROIs.