Artificial intelligence and early esophageal cancer

Table 1 Application of artificial intelligence in endoscopic detection of Barrett’s esophagus

Ref.	Target disease	Endoscopic modality	AI technology	Database	Outcomes
van der Sommen et al[23], 2016	Early neoplasia in BE	WLI	SVM	100 images	Per-image sensitivity 83%/specificity 83%; Per-patient sensitivity 86%/specificity: 87%
Struyvenberg et al[24], 2021	BE	WLI/NBI	CNN	Train 494364 images/1247 images; test 183 images/157 videos	Images: Accuracy 84%/sensitivity 88%/specificity 78%; Videos: Accuracy 83%/sensitivity 85%/specificity 83%
de Groof et al[25], 2020	Early neoplasia in BE	WLI	ResNet-UNet	Train 1544 images; test 160 images	Dataset 4: Accuracy 89%/sensitivity 90%/specificity 88%; Dataset 5: Accuracy 88%/sensitivity 93%/specificity 83%
de Groof et al[26], 2020	Barrett’s neoplasia	WLI	ResNet-UNet	Train 1544 images; test 20 patients	Accuracy 90%/sensitivity 91%/specificity 89%
Hong et al[27], 2017	BE	Endomicroscopy	CNN	Train 236 images; test 26 images	Accuracy 80.77%
Hashimoto et al[28], 2020	Early neoplasia in BE	WLI/NBI	CNN	Train 1832 images; test 458 images	Accuracy 95.4%/sensitivity 96.4%/ specificity 94.2%
de Groof et al[29], 2019	Barrett’s neoplasia	WLI	SVM	60 images	Accuracy 92%/sensitivity 95%/specificity 85%

BE: Barrett’s esophagus; WLI: White light imaging; SVM: Support vector machine; NBI: Narrow band imaging; CNN: Convolutional neural network.

Table 2 Application of artificial intelligence in endoscopic detection of early esophageal cancer

Ref.	Target disease	Endoscopic modality	AI technology	Database	Outcomes
Ebigbo et al[37], 2019	EAC	WLI/NBI	CNN	248 images	Augsburg database: Sensitivity 97%/specificity 88% (WLI); Sensitivity 94%/specificity 80% (NBI); MICCAI database: Sensitivity 92%/specificity 100%
Ebigbo et al[38], 2020	EAC	WLI	CNN	Train 129 images; test 62 images	Accuracy 89.9%/sensitivity 83.7%/specificity 100%
Horie et al[39], 2019	EC	WLI/NBI	CNN	Train 8428 images; test 1118 images	Accuracy 98%/sensitivity 98%
Cai et al[40], 2019	ESCC	WLI	DNN	Train 2428 images; test 187 images	Accuracy 91.4%/sensitivity 97.8%/specificity 85.4%
Ohmori et al[41], 2020	ESCC	WLI/NBI/BLI	CNN	Train 22562 images; test 727 images	Non-ME: Accuracy 81.0%/sensitivity 90%/specificity 76% (WLI); Accuracy 77%/sensitivity 100%/specificity 63% (NBI/BLI); ME: Accuracy 77%/sensitivity 98%/specificity 56%
Liu et al[42], 2020	EC	WLI	CNN	Train 1017 images; test 255 images	Accuracy 85.83%/sensitivity 94.23%/specificity 94.67%
Kumagai et al[43], 2019	ESCC	ECS	CNN	Train 4715 images; test 1520 images	Accuracy 90.9%/sensitivity 92.6%/specificity 89.3%
Guo et al[44], 2020	ESCC	NBI	CNN	Train 6473 images; test 6671 images and 80 videos	Images: Sensitivity 98.04%/specificity 95.03%; videos: Non-ME sensitivity 60.8% (per frame)/100% (per lesion); ME sensitivity 96.1% (per frame)/100% (per lesion)
Tokai et al[46], 2020	ESCC	WLI/NBI	CNN	Train 1751 images; test 291 images	Accuracy 80.9%/sensitivity 84.1%/specificity 73.3%
Nakagawa et al[47], 2019	ESCC	WLI/NBI	CNN	Train 14338 images; test 914 images	Accuracy 91%/sensitivity 90.1%/specificity 95.8%
Zhao et al[48], 2019	ESCC	NBI	Double-labeling FCN	1350 images	Lesion level: Accuracy 89.2%; pixel level: Accuracy 93%
Everson et al[49]	ESCC	NBI	CNN	7046 images	Accuracy 93.7%/sensitivity 89.3%/specificity 98%
Uema et al[50], 2021	ESCC	NBI	CNN	Train 1777 images; test 747 images	Accuracy 84.2%
Fukuda et al[51], 2020	ESCC	NBI/BLI	CNN	Train 28333 images; test 144 patients	Accuracy 63%/sensitivity 91%/specificities 51% (detection); accuracy 88%/sensitivity 86%/specificities 89% (characterization)
Shimamoto et al[52], 2020	ESCC	WLI/NBI/BLI	CNN	Train 23977 images; test 102 videos	Non-ME: Accuracy 87%/sensitivity 50%/specificity 99%; ME: Accuracy 89%/sensitivity 71%/specificity 95%
Waki et al[53], 2021	ESCC	WLI/NBI/BLI	CNN	Train 18797 images; test 100 videos	Sensitivity 85.7%/specificity 40%

EAC: Esophageal adenocarcinoma; WLI: White light imaging; NBI: Narrow band imaging; CNN: Convolutional neural network; EC: Esophageal cancer; ESCC: Esophageal squamous cell carcinoma; DNN: Deep neural network; BLI: Blue laser imaging; ME: Magnification endoscopy; ECS: Endocytoscopic system; FCN: Fully convolutional network.