Performance of Non-Contrast MR Enterography to Localize and Predict Disease Activity in Crohn's Disease.

Srigouri Yalamanchili¹, Michael Macari¹, Rafael Rivera¹, Danny Kim¹, Alec Megibow¹, James Babb¹, Joseph Levy¹, Kerry Zabriskie¹, Sooah Kim²

Secific findings at MR imaging- mural T2 high signal intensity and contrast enhancement patterns- have been proposed as accurate markers of disease activity. The purpose of our study is to assess the performance using conventional non-contrast MR imaging sequences as a tool to localize and predict disease activity in Crohns disease in comparison to post-contrast images. Thirty patients referred for MR enterography were evaluated at 1.5 T using steady state free precession, single shot fast spin echo, fat suppressed T2, and pre- and post contrast enhanced T1-weighted sequences. Images were reviewed by two radiologists in two separate sessions; the readers interpreted non-contrast images of MRI during the first session and whole images including post-contrast sequences during the second session. The readers evaluated the presence or absence of disease in regard to the presence of active inflammation using a six-point grading system (0, No evidence of disease; 1, definite absence of active inflammation; 2, probable absence of active inflammation; 3, equivocal; 4, probable presence of active inflammation; and 5, definite presence of active inflammation). Reference standard was constructed by combining surgical, endoscopy, physical, and all available imaging findings. There were 34 bowel segments with active inflammation on reference standard in 18 subjects (proximal ileum, n=1, distal ileum, n=4; terminal ileum, n=16; cecum, n=6; ascending colon, n=1; rectosigmoid, n=2). Inter-reader agreement was significantly higher for whole imaging including post-contrast images relative to non-contrast images (Kappa 0.31 for non-contrast and 0.41 for whole images). Although there is no significant difference localizing abnormal bowel segment involved with Crohns disease using non-contrast images, inter-reader agreement, sensitivity, and accuracy are significantly higher in the assessment of active disease adding post-contrast images to non-contrast images for MRI interpretation.