Table 2.
Applying AI based on EUS in the differential diagnosis of pancreatic cancer and other pancreatic tumors
| Reference | Sample size | Data source | Algorithms | Aim | Best result |
|---|---|---|---|---|---|
| Zhu et al. 101 | 388 cases | EUS | SVM | PC vs CP | Accuracy (94.2%), Sensitivity (96.25%), Specificity (93.38%), PPV (92.21%), NPV (96.68%) |
| Udriştoiu et al. 102 | 65 cases | multi-sequences EUS | CNN, LSTM neural network | PDAC vs CPP vs PNET | AUC (0.98), Accuracy (98.26%) |
| Tong et al. 103 | 558 cases | CE-EUS | CNN (ResNet-50) | PDAC vs CP | AUC (0.986) |
| Tonozuka et al. 104 | 1390 images | EUS | CNN | PDAC vs CP | AUC (0.940), Sensitivity (92.4%), Specificity (84.1%), PPV (86.8%), NPV (90.7%) |
| Sǎftoiu et al. 105 | 167 cases | CEH-EUS | ANN | PC vs CP | Specificity (94.44%), Sensitivity (94.64%), PPV (97.24%), NPV (89.47%) |
| Marya et al. 106 | 1174461 images from 583 cases | EUS | CNN | AIP vs PDAC vs CP vs NP | For PDAC: AUC (0.976), Sensitivity (95%), Specificity (91%), PPV (87%), NPV (97%) |
| Sǎftoiu et al. 107 | 68 cases | EUS elastography | NN_MLP | PC vs CP vs NP vs PNET | AUC (0.932), Accuracy (89.7%), Sensitivity (91.4%), Specificity (87.9%), PPV (88.9%), NPV (90.6%) |
| Norton et al. 108 | 35 cases | EUS | NN | PC vs CP | Accuracy (80%), Sensitivity (100%), Specificity (50%) |
| Sǎftoiu et al. 109 | 258 cases | EUS elastography | ANN_MLP | PC vs CP | AUC (0.94), Accuracy (91.14%), Sensitivity (87.59%), Specificity (82.94%), PPV (96.25%), NPV (57.22%) |
| Kuwahara et al. 110 | 3,970 images | EUS | CNN (ResNet-50) | Diagnosis of malignancy in IPMN | AUC (0.98), Accuracy (94.0%), Sensitivity (95.7%), Specificity (92.6%), PPV (91.7%), NPV (96.2%) |
| Iwasa et al. 111 | 100 cases | CE-EUS | U-Net | PC segmentation | Median IoU (0.77) |
| Zhang et al. 112 | 2207+19486 images | EUS | ResNet | Pancreas segmentation; station recognition | Pancreas segmentation: DSC (71.5%); Station recognition: accuracy (82.4%) |
Abbreviations: AIP: autoimmune pancreatitis; ANN: artificial neural network; AUC: area under the curve; CE-EUS: contrast-enhanced EUS; CEH-EUS: contrast-enhanced harmonic EUS; CNN: convolutional neural network; CP: chronic pancreatitis; CPP: chronic pseudotumoral pancreatitis; DSC: dice similarity coefficient; IPMN: intraductal papillary mucinous neoplasm; IoU: intersection over union; LSTM: long short-term memory; MLP: multilayer perceptron; NN: neural network; NP: normal pancreas; NPV: negative predictive value; PDAC: pancreatic ductal adenocarcinoma; PNET: pancreatic neuroendocrine tumor; PPV: positive predictive value; SVM: support vector machine.