Table 3.
Performance of the CNN model and mean ADC.
| Internal Test Set (140) | External Test Set (102) | Tra and Val Sets (560) | |||
|---|---|---|---|---|---|
| CNN Model | Mean ADC | CNN Model | Mean ADC | Mean ADC | |
| Accuracy | 0.907 | 0.836 | 0.902 | 0.824 | 0.845 |
| Sensitivity | 0.893 | 0.845 | 0.873 | 0.818 | 0.864 |
| Specificity | 0.929 | 0.821 | 0.894 | 0.829 | 0.820 |
| PPV | 0.949 | 0.877 | 0.906 | 0.849 | 0.861 |
| NPV | 0.852 | 0.780 | 0.857 | 0.796 | 0.823 |
| F1 score | 0.908 | 0.836 | 0.882 | 0.824 | 0.845 |
| kappa value | 0.809 | 0.661 | 0.764 | 0.646 | 0.684 |
| AUC (95% CI) | 0.977 (0.957-0.998) | 0.866 (0.805-0.927) | 0.926 (0.876-0.976) | 0.845 (0.766-0.925) | 0.868 (0.838-0.899) |
AUC, area under the receiver operating characteristic curve; CI, confidence interval; CNN, convolutional neural network; NPV, negative predictive value; PPV, positive predictive value; Tra and val sets, training and validation sets; ADC, apparent diffusion coefficient.