Minireviews
Copyright ©The Author(s) 2021.
World J Gastroenterol. May 28, 2021; 27(20): 2531-2544
Published online May 28, 2021. doi: 10.3748/wjg.v27.i20.2531
Table 1 Summary of studies using deep learning for detection of esophageal precancerous lesions
Ref.
Year
Imaging
Study design
Study aim
DL model
Dataset
Outcomes
Cai et al[30]2019WLERetrospectiveDetection of precancerous lesions and early ESCC--2615 imagesSensitivity: 97.8%. Specificity: 85.4%. Accuracy: 91.4%
Guo et al[31]2020NBI, M-NBIRetrospectiveDetection of precancerous lesions and early ESCCSegNet13144 images and 168865 video framesSensitivity: 96.10% for M-NBI videos, 60.80% for non-M-NBI videos, 98.04% for images. Specificity: 99.90% for non-M-NBI/M-NBI videos, 95.30% for images
de Groof et al[32]2020WLERetrospectiveDetection of Barrett’s neoplasiaResNet/U-Ne1544 imagesSensitivity: 91%. Specificity: 89%. Accuracy: 90%
de Groof et al[33]2020WLERetrospectiveDetection of Barrett’s neoplasiaResNet/U-Ne494364 unlabeled images and 1704 labeled imagesSensitivity: 90%. Specificity: 88%. Accuracy: 89%
Struyvenberg et al[34]2021NBIRetrospectiveDetection of Barrett’s neoplasiaResNet/U-Ne2677 imagesSensitivity: 88%. Specificity: 78%. Accuracy: 84%
Hashimoto et al[35]2020WLE, NBIRetrospectiveRecognition of early neoplasia in BEInception-ResNet-v2, YOLO-v22290 imagesSensitivity: 96.4%. Specificity: 94.2%. Accuracy: 95.4%
Hussein et al[36]2020WLERetrospectiveDiagnosis of early neoplasia in BEResnet101266930 video framesSensitivity: 88.26%. Specificity: 80.13%
Ebigbo et al[37]2020WLERetrospectiveDiagnosis of early EAC in BE DeepLab V.3+, Resnet101191 imagesSensitivity: 83.7%. Specificity: 100%. Accuracy: 89.9%
Liu et al[38]2020WLERetrospectiveDetection of esophageal cancer from precancerous lesionsInception-ResNet1272 imagesSensitivity: 94.23%. Specificity: 94.67%. Accuracy: 85.83%
Wu et al[39]2021WLERetrospectiveAutomatic classification and segmentation for esophageal lesionsELNet1051 imagesClassification sensitivity: 90.34%. Classification specificity: 97.18%. Classification accuracy: 96.28%. Segmentation sensitivity: 80.18%. Segmentation Specificity: 96.55%, Segmentation accuracy: 94.62%
Ghatwary et al[40]2021WLERetrospectiveDetection of esophageal abnormalities from endoscopic videosDenseConvLstm, Faster R-CNN42425 video framesSensitivity: 93.7%. F-measure: 93.2%
Table 2 Summary of studies using deep learning for detection of gastric precancerous lesions
Ref.
Year
Imaging
Study design
Study aim
DL model
Dataset
Outcomes
Shichijo et al[47]2017WLERetrospectiveDiagnosis of H. pylori infectionGoogLeNet43689 imagesSensitivity: 88.9%; Specificity: 87.4%; Accuracy: 87.7%
Itoh et al[48]2018WLERetrospectiveAnalysis of H. pylori infectionGoogLeNet179 imagesSensitivity: 86.7%; Specificity: 86.7%
Zheng et al[49]2019WLERetrospectiveEvaluation of H. pylori infection statusResNet-5015484 imagesSensitivity: 91.6%; Specificity: 98.6%; Accuracy: 93.8%
Nakashima et al[50]2018BLI-bright, LCIProspectivePrediction of H. pylori infection statusGoogLeNet666 imagesSensitivity: 96.7%; Specificity: 86.7%
Nakashima et al[51]2020WLE, LCIProspectiveDiagnosis of H. pylori infection--13127 imagesFor currently infected patients, the sensitivity and specificity are 62.5% and 92.5%, respectively
Guimarães et al[53]2020WLERetrospectiveDiagnosis of atrophic gastritisVGG16270 imagesAccuracy: 93%
Zhang et al[54]2020 WLERetrospectiveDiagnosis of atrophic gastritisDenseNet1215470 imagesSensitivity: 94.5%; Specificity: 94.0%; Accuracy: 94.2%
Horiuchi et al[55] 2020M-NBIRetrospectiveDifferentiation between early gastric cancer and gastritisGoogLeNet2826 imagesSensitivity: 95.4%; Specificity: 71.0%; Accuracy: 85.3%
Wang et al[57]2019 WLERetrospectiveLocalization and identification of GIMDeepLab V.3+200 imagesAccuracy: 89.51%
Zheng et al[58]2020WLERetrospectiveDetection of atrophic gastritis and GIM ResNet-503759 imagesSensitivity for atrophic gastritis: 87.2%; Specificity for atrophic gastritis: 91.1%; Sensitivity for GIM: 90.3%; Specificity for GIM: 93.7%
Yan et al[18]2020NBI, M-NBIRetrospectiveDiagnosis of GIMEfficientNetB42357 imagesSensitivity: 91.9%; Specificity: 86.0%; Accuracy: 88.8%
Cho et al[60]2019 WLEProspectiveClassification of multiclass gastric neoplasmsInception-Resnet-v25217 images Accuracy: 84.6%
Inoue et al[61]2020WLE, NBIRetrospectiveDetection of duodenal adenomas and high-grade dysplasiasSingle-Shot Multibox Detector1511 imagesFor high-grade dysplasia, the sensitivity and specificity are all 100%
Lui et al[62]2020NBIRetrospectiveClassification of gastric lesionsResNet3000 imagesSensitivity: 97.1%; Specificity: 85.9%; Accuracy: 91.0%