Table 3. Evaluation of DRUNet comparing with different layers and different networks.
| Networks | PSNR | SSIM | Tenengrad | |||||
|---|---|---|---|---|---|---|---|---|
| T1WI | T2WI | T1WI | T2WI | T1WI | T2WI | |||
| DRUNet-34 | 34.19±2.07 | 33.43±1.11 | 0.96±0.04 | 0.96±0.06 | 17.85±2.87 | 19.82±3.67 | ||
| DRUNet-50 | 34.23±2.02 | 33.44±1.12 | 0.97±0.04 | 0.97±0.04 | 18.10±2.94 | 19.57±3.40 | ||
| DRUNet-101 | 34.25±2.06 | 33.50±1.08 | 0.97±0.03 | 0.98±0.05 | 17.03±2.75 | 19.76±3.54 | ||
| U-Net (17) | 34.08±2.16 | 27.61±0.06 | 0.90±0.06 | 0.73±0.12 | 13.62±2.36 | 21.79±0.70 | ||
DRUNet-X, double ResNet-U-Net with double X convolutional layers; PSNR, peak signal-to-noise ratio; SSIM, structural similarity; Tenengrad, Tenenbaum gradient; T1WI, T1-weighted images; T2WI, T2-weighted images.