TABLE 4.

Cross‐sectional imaging based artificial intelligence in bowel fibrosis of IBD

Type of subjects	Type of study design	Sample size	Study results	Reference	AUC in the training cohort	AUC in the test cohort
Adult patients with CD	Retrospective multicenter study	212 bowel lesions of 167 CD patients	This radiomic model enabled the accurate differentiation of moderate‐severe from non‐mild intestinal fibrosis in CD, showing remarkable robustness in different inflammatory severities, CD locations, and CT scanners	Li et al 53	0.888	0.832
Adult patients with CD	Retrospective multicenter study	312 bowel segments of 235 CD patients	Based on the same patient cohorts mentioned above (reference 52 ), the diagnostic performance of the deep learning model developed by using a 3D deep convolutional neural network was not inferior to that of the radiomic model, but it was more time‐saving than the radiomic model (48.4 vs. 599.8 s)	Meng et al 54	0.828	0.811
Paediatric patients with CD	Retrospective single institution study	64 bowel segments of 25 patients	Texture analysis of enteric phase T1‐weighted fat suppressed postcontrast MRI images can distinguish fibrotic from nonfibrotic strictures, providing a noninvasive biomarker of stricture composition that can guide therapy arrangement	Tabari et al 55	NA	0.995
Rat model of colitis	Retrospective animal study	45 rat model of inflammation (10 control and 35 irradiated with visible lesions)	This approach offers practitioners a valuable tool to evaluate antifibrotic treatments under development and to extrapolate such noninvasive MRI scores model for patients with the aim of identifying early stages of fibrosis improving disease management	Morilla et al 56	NA	0.875

Abbreviations: AUC, area under the curve; CD, Crohn's disease; IBD, inflammatory bowel disease; MRI, magnetic resonance imaging.