TABLE 1.

AI in the Diagnosis of Esophageal Adenocarcinoma

					Performance
References	AI Model	Study Type	Aim	Endoscopic Technique	Accuracy	Sensitivity	Specificity
de Groof et al14	ML—SVM	Retrospective	Detection of early Barrett’s neoplasia	WLE	92%	95%	85%
de Groof et al15	DL	Prospective	Detection of early Barrett’s neoplasia	WLE	90%	91%	89%
Ebigbo et al16	DL	Prospective	Detection of early Barrett’s neoplasia	WLE	89.9%	83.7%	100%
de Groof et al17	DL	Retrospective	AI vs. endoscopists in detection of early Barrett’s neoplasia	WLE AI EE	88%  73%	93%  72%	83%  74%
Swager et al18	ML—SVM	Retrospective	AI vs. endoscopists in detection of early Barrett’s neoplasia	VLE AI EE	AUC=0.95  AUC=0.81	90%  85%	93%  68%
van der Sommen et al19	ML—SVM	Retrospective	AI vs. endoscopists in detection of early Barrett’s neoplasia	WLE AI Best endoscopist	—  —	86%  90%	87%  91%
Ebigbo et al20	DL	Retrospective	AI vs. endoscopists in predicting invasion in Barrett’s cancer	WLE AI EE	77%  63%	64%  78%	71%  70%

AI indicates artificial intelligence; AUC, area under the curve; DL, deep learning; EE, expert endoscopists; ML, machine learning; SVM, support vector machine; VLE, volumetric laser endomicroscopy; WLE: white-light endoscopy.