Table 4.
DenseNet121 - Not Pretrained (High Accuracy) | Resnet34 - Pretrained (Poorer Accuracy) | |
---|---|---|
Positive Predictive Value | 93.3 % | 85.1 % |
Negative Predictive Value | 75.5 % | 46.6 % |
Accuracy | 88.6 % | 71.4 % |
F1 Score | 88.2 % | 72.6 % |
ROC AUC | 0.95 | 0.84 |
Sensitivity | 90.1 % | 74.1 % |
Specificity | 82.4 % | 63.6 % |