Table 1.
Dice Similarity Coefficient (DSC), sensitivity, positive predictive value (PPV), training time and number of parameters for our five experimental models.
| Model | DSC | Sensitivity | PPV | Training time | # Parameters |
|---|---|---|---|---|---|
| U-Net(FLAIR) | 0.5440 | 0.4594 | 0.6275 | 1 h 52 m 55 s | 7,859,715 |
| U-Net(IAM) | 0.5274 | 0.4179 | 0.6769 | 1 h 53 m 52 s | 7,859,715 |
| U-Net(F+I) | 0.5281 | 0.4902 | 0.6268 | 1 h 24 m 22 s | 7,861,315 |
| Saliency U-Net(F+I) | 0.5535 | 0.4730 | 0.6034 | 1 h 30 m 1 s | 2,756,803 |
| Dilated Saliency U-Net(F+I) | 0.5588 | 0.4747 | 0.6374 | 1 h 4 m 18 s | 2,623,683 |
Values in bold are the highest scores and in italic the second highest. In the brackets after the model names, the input data type is specified. “FLAIR” is equivalent to T2-FLAIR and “F+I” refers to taking both T2-FLAIR and IAM as input.