Table 2.

Characteristics of included studies whose primary outcome was polyp characterization.

Authors	Year	Recruitment	Machine learning approach	Image modality	Patients, n	Polypsor lesions, n	Total images, n	Images for training, n	Images for validation, n
Tischendorf et al [36]	2010	Prospective pilot	SVMa classifier	Magnification NBIb	223	209	—c	208	—
Gross et al [37]	2011	Prospective	SVM classifier	Magnification NBI	214	434	—	433	—
Ganz et al [13]	2012	Retrospective	Shape-UCMd	NBI	—	—	—	58	87
Takemura et al [38]	2012	Retrospective	SVM classifier	Magnification NBI	—	371	—	1519	371
Mori et al [39]	2015	Retrospective	ECe-CADf	EC	152	176	—	—	—
Kominami et al [11]	2016	Retrospective	SVM classifier	Magnification NBI	41	118	—	2247	—
Misawa et al [40]	2016	Retrospective	EndoBRAINg	NBI and EC	—	85	1079	979	100
Mesejo et al [41]	2016	Retrospective	SfMh	White light and NBI	—	76	—	—	—
Mori et al [42]	2016	Retrospective	SVM classifier	EC-CAD	123	205	—	—	6051
Takeda et al [43]	2017	Retrospective	SVM classifier	EC-CAD	242	375	5843	5643	200
Byrne et al [44]	2017	Retrospective	DCNNi	NBI	—	125	—	60,089	—
Komeda et al [45]	2017	Retrospective	CNN	Endoscopic images	—	—	1200	—	—
Misawa et al [46]	2017	Retrospective	EndoBRAIN and ECVj-CAD	NBI	100	124	1834	173	1661
Mori et al [47]	2018	Retrospective	—	EC	—	144	—	—	—
Chen et al [48]	2018	Prospective	DNNk	NBI	193	284	2441	2157	284
Renner et al [49]	2018	Retrospective	DNN	NBI and HDWLl	250	231	788	602	186
Mori et al [50]	2018	Prospective	SVM classifier	NBI and EC	325	466	—	61,925	450
Mori et al [50]	2018	Prospective	SVM classifier	NBI and EC	325	466	—	61,925	450
Kudo et al [51]	2019	Retrospective	EndoBRAIN system	White light, NBI, and EC	89	100	—	69,142	5065
Kudo et al [51]	2019	Retrospective	EndoBRAIN system	White light, NBI, and EC	89	100	—	69,142	5065
Figueiredo et al [52]	2019	Retrospective	Segmentation algorithm	NBI	10	11	86	43	43
Rodriguez-Diaz et al [53]	2020	Retrospective	DeepLab framework	High magnification NBI	286	607	740	—	—
Yang et al [54]	2020	Retrospective	CNN-Inception-ResNet	White light	1339	—	3828	—	240
Zachariah et al [55]	2020	Retrospective	CNN-Inception-ResNet	NBI and white light	—	—	6223	—	634

^aSVM: support vector machine.

^bNBI: narrow band imaging.

^cThis value was not reported.

^dShape-UCM is an algorithm for automatic polyp segmentation.

^eEC: endocytoscopy.

^fCAD: computer-aided diagnosis.

^gEndoBRAIN is a novel artificial intelligence system.

^hSfM: structure from motion.

ⁱDCNN: deep convolutional neural network.

^jECV: endocytoscopic vascular pattern.

^kDNN: deep neural network.

^lHDWL: high-definition white light.