Abstract
OBJECTIVES: To achieve automated quantification of visceral adipose tissue (VAT) distribution in CT images and screen out parameters with discriminative value for inflammatory bowel disease (IBD) subtypes. METHODS: This retrospective multicenter study included Crohn's disease (CD) and ulcerative colitis (UC) patients from three institutions between 2012 and 2021, with patients with acute appendicitis as controls. An automatic VAT segmentation algorithm was developed using abdominal CT scans. The VAT volume, as well as the coefficient of variation (CV) of areas within the lumbar region, was calculated. Binary logistic regression and receiver operating characteristic analysis was performed to evaluate the potential of indicators to distinguish between IBD subtypes. RESULTS: The study included 772 patients (365 CDs, median age [inter-quartile range] = 31.0. (25.0, 42.0) years, 255 males; 241 UCs, 46.0 (34.0, 55.5) years, 138 males; 166 controls, 40.0 (29.0, 53.0) years, 80 males). CD patients had lower VAT volume (CD = 1584.95 ± 1128.31 cm(3), UC = 1855.30 ± 1326.12 cm(3), controls = 2470.91 ± 1646.42 cm(3)) but a higher CV (CD = 29.42 ± 15.54 %, p = 0.006 and p ˂ 0.001) compared to UC and controls (25.69 ± 12.61 % vs. 23.42 ± 15.62 %, p = 0.11). Multivariate analysis showed CV was a significant predictor for CD (odds ratio = 6.05 (1.17, 31.12), p = 0.03). The inclusion of CV improved diagnostic efficiency (AUC = 0.811 (0.774, 0.844) vs. 0.803 (0.766, 0.836), p = 0.08). CONCLUSION: CT-based VAT distribution can serve as a potential biomarker for distinguishing IBD subtypes. CRITICAL RELEVANCE STATEMENT: Visceral fat distribution features extracted from CT images using an automated segmentation algorithm (1.14 min) show differences between Crohn's disease and ulcerative colitis and are promising for practical radiological screening. KEY POINTS: • Radiological parameters reflecting visceral fat distribution were extracted for the discrimination of Crohn's disease (CD) and ulcerative colitis (UC). • In CD, visceral fat was concentrated in the lower lumbar vertebrae, and the coefficient of variation was a significant predictor (OR = 6.05 (1.17, 31.12), p = 0.03). • The differences between CD, UC, and controls are promising for practical radiological screening.