Table 1. Comparison of performances of deep learning and manual in internal and external test.
Internal test | External test | |||||||
---|---|---|---|---|---|---|---|---|
Resnet50 | Xception | InceptionV3 | Manual | Resnet50 | Xception | InceptionV3 | Manual | |
Sensitivity | 0.90 | 0.85 | 0.93 | 0.62 | 0.86 | 0.86 | 0.90 | 0.62 |
specificity | 0.90 | 0.97 | 0.95 | 0.97 | 1.0 | 0.90 | 0.90 | 0.95 |
PPV | 0.90 | 0.97 | 0.95 | 0.96 | 1.0 | 0.90 | 0.90 | 0.93 |
NPV | 0.90 | 0.87 | 0.93 | 0.72 | 0.87 | 0.86 | 0.90 | 0.70 |
F1-score | 0.90 | 0.91 | 0.94 | 0.79 | 0.92 | 0.88 | 0.90 | 0.77 |
AUC | 0.950 | 0.928 | 0.970 | 0.80 | 0.955 | 0.936 | 0.967 | 0.785 |
MCC | 0.80 | 0.832 | 0.876 | 0.64 | 0.858 | 0.757 | 0.805 | 0.60 |
Abbreviations: PPV, positive predictive value; NPV, negative predictive value, AUC, area under the ROC curve, MCC, Matthews correlation coefficient.