Harnessing artificial intelligence in gastrointestinal endoscopy for early detection of dysplastic lesions in inflammatory bowel disease

Table 1 Summary of studies on the application of artificial intelligence for dysplasia detection in inflammatory bowel disease

Ref.	Country	AI model/technique	Study design	Dataset	AI performance
					Mode	Sensitivity (%)	Specificity (%)	Accuracy (%)
Guerrero Vinsard et al[12], 2023	United States	IBD-CADe system with HD-WLE and chromoendoscopy (Scaled-YOLOv4)	Retrospective	3437 colonoscopies, 1692 endoscopic images (HD-WLE and dye chromoendoscopy)	HD-WLE	95.1	98.8	96.8
					Chromoendoscopy	67.4	88	77.8
Yamamoto et al[13], 2022	Japan	Deep CNN (EfficientNet-B3)	Retrospective (Pilot Study)	862 endoscopic images (WLE and chromoendoscopy)	-	72.5	82.9	79
Abdelrahim et al[32], 2024	United Kingdom	Deep CNN (RetinaNet; ResNet-101 backbone)	Prospective	478 endoscopic images and real-time colonoscopy data	Lesion detection	93.5	80.6	N/A
					Lesion characterization	87.5	80.6	N/A
Maeda et al[33], 2021	Japan	EndoBRAIN-EYE	Case report	Single case (patient with ulcerative colitis)	-	N/A	N/A	N/A
Fukunaga et al[34], 2021	Japan	EndoBRAIN	Case report	Single case (patient with pancolitis)	-	N/A	N/A	N/A

AI: Artificial intelligence; IBD: Inflammatory bowel disease; CADe: Computer-aided detection; HD-WLE: High-definition white light endoscopy; CNN: Convolutional neural network; WLE: White-light endoscopy; UC: Ulcerative colitis; N/A: Not available.