Minireviews
Copyright ©The Author(s) 2021.
World J Gastroenterol. Apr 14, 2021; 27(14): 1392-1405
Published online Apr 14, 2021. doi: 10.3748/wjg.v27.i14.1392
Table 2 Artificial intelligence application for esophageal adenocarcinoma
AI ApplicationStudy designData categoryType of ImagesAI architectureTraining datasetValidation Method or datasetAUCSENSPEACCPPVNPVCompared with expertsRef.
DetectionRetrospectiveStill imageWLICAD-SVM64 imagesLOOCVNA95%NA75%NANANAvan der Sommen et al[36], 2014
DetectionRetrospectiveStill imageWLICAD-SVM100 ImagesLOOCVNA83% (per-image) 86% (per-patient)83% (per-image) 87% (per-patient)NANANAInferiorvan der Sommen et al[37], 2016
DetectionRetrospectiveStill imageVLECAD60 imagesLOOCV0.9590%93%NANANASuperiorSwager et al[38], 2017
DetectionRetrospectiveStill imageVLECAD60 imagesLOOCV0.90-0.93NANANANANASuperiorvan der Sommen et al[40], 2018
DiagnosisRetrospectiveStill imageWLI/NBICNN8 patientsCaffe DL frameworkNA88% (WLI)/88% (NBI) (per-patient) 69% (WLI)/71% (NBI) (per-image)NA90%NANANAHorie et al[21], 2019
DetectionRetrospectiveStill imageWLECNN-SSD100 images/39patients20% patients/5-fold-CV/LOOCVNA96%92%NANANANAGhatwary et al[41], 2019
DetectionRetrospectiveStill imageWLI/NBICNN- Inception-ResNet-v21832 images458 imagesNA96.4%94.2%95.4%NANANAHashimoto et al[42], 2019
DetectionRetrospectiveStill imageWLICAD60 imagesLOOCV0.9295%85%91.7%NANANAde Groof et al[43], 2019
DetectionRetrospectiveStill imageWLICAD-ResNet-UNet1544 images4-fold-CV (internal validation)/160 images (external validation)NA87.6% (internal validation) 92.5% (external validation)88.6% (internal validation) 82.5% (external validation)88.2% (internal validation) 87.5% (external validation)NANANAde Groof et al[44], 2019
DiagnosisRetrospectiveStill imageWLI/NBICAD-ResNet248 imagesLOOCVNA97% (WLI)/94% (NBI) (Augsburg data)92% (MICCAI)88% (WLI)/80% (NBI) (Augsburg data)100% (MICCAI)NANANANAEbigbo et al[45], 2019
DiagnosisRetrospectiveRandom images from real-time videoWLICAD-ResNet-/DeepLab V.3+129 images36 images (real time)NA83.7%100%89.9%NANANAEbigbo et al[49], 2020
SurveillanceProspectiveReal-time imageWLI/NBI/VLEIRISNAReal-time imageNANANANANANANATrindade et al[50], 2019
DetectionProspectiveLive endoscopic procedureLive endoscopic procedureCAD-ResNet/U-Net1544 images48 levels/144 images/20 live endoscopic procedureNA90.9% (per level) 75.8% (per image) 90% (per patient)89.2% (per level) 86.5% (per image) 90% (per patient)89.6% (per level) 84.0% (per image) 90% (per patient)NANANAde Groof et al[51], 2020