Performance and comparison of artificial intelligence and human experts in the detection and classification of colonic polyps

Table 1 Characteristics and results of the eligible studies

Author	Years	Main purpose	Model type	Image modality; magnified (if any)	Train set	Test set	Different seniority; endoscopists	External validation
Halligan [24]	2006	Detection	CAD	CT colonography	239 patients	110 patients	No; 10 experts	No
Petrick [38]	2008	Detection	CAD system	CT colonography	UC	UC	No; 4 experts	No
Tischendorf [39]	2010	Classification	Linear classifier、K-NN、SVM	NBI; × 100	UC	UC	No; 2 experts	No
Ignjatovic [7]	2011	Classification	UC	NBI	UC	30 polyps	Yes; 2 experts, 1 novice	No
Gross [17]	2011	Classification	SVM	NBI; × 150	NA	434 polyps	Yes; 2 experts, 2 novices	No
Mang [40]	2012	Detection	CTC CAD system	CT colonography	UC	UC	No; 0	No
Mesejo [23]	2016	Classification	RF、RS、SVM	WL, NBI	UC	UC	Yes; 1 expert, 1 novice	No
Mori [41]	2018	Classification	SVM	NBI, methylene blue staining; × 520	UC	UC	No; 0	No
Renner [42]	2018	Classification	CNN*	WL, NBI;@@@Without magnification	602 images	186 images	No; 8 experts	No
Chen [21]	2018	Classification	DNN-CAD*	NBI; @@@Optical maximum magnification	2157 images	284 images	Yes; 2 experts, 4 novices	No
Shin [43]	2018	Detection	SVM	UC	UC	UC	No; 0	No
Byrne [11]	2019	Classification	CNN	NBI	223 videos, @@@60,089 images	125 videos	No; 0	No
Cristina [44]	2019	Classification	SVM	WL, NBI; @@@Without magnification	UC	UC	No; 2 experts	No
Zachariah [35]	2020	Classification	CNN	WL, NBI	5278 images	634 images	No; 0	Yes
Shahidi [45]	2020	Classification	CNN	WL, NBI, near-focus	UC	UC	No; 0	No
Qadir [14]	2020	Detection	Faster R-CNN	UC	UC	UC	No; 0	No

CT colonography, computed tomographic colonography; NBI, narrow-band imaging; WL, white light; CAD, computer-aided diagnosis; K-NN, k-nearest neighbor; RF, random forests; RS, random subspaces; SVM, support vector machine; CNN, convolutional neural network; DNN-CAD, computer-aided diagnosis with a deep neural network; UC, unclear

ISSN: 1471-230X