Table 5.
Comparison of the proposed architectures and state-of-the-art methods.
| Reference | Method | Dataset | Dice score (%) | IoU score (%) |
|---|---|---|---|---|
| Tsochatzidis et al.42 | UNet+ | CBIS-DDSM | 72.2 | 56.5 |
| Sun et al.38 | Attention UNet | INbreast | 79.1 | – |
| CBIS-DDSM | 81.8 | – | ||
| Ravitha Rajalakshmi et al.31 | Deeply supervised U-Net (DS U-Net) | INbreast | 79 | 83.2 |
| CBIS-DDSM | 82.9 | – | ||
| Dhungel et al.7 | Deep structured output learning + refinement | INbreast | 85 | – |
| Dhungel et al.12 | CNN + CRF | INbreast | 90.06 | – |
| Abdelhafiz et al.33 | R-UNet | INbreast | 90.5 | 89.1 |
| Zhu et al.26 | Multi-scale FCN-CRF | INbreast | 90.97 | – |
| Wang et al.40 | ResNet34 + ASPP | INbreast | 91.1 | – |
| Singh et al.27 | conditional GAN (cGAN) | INbreast | 91.47 | 83.58 |
| Al-Antari et al.25 | Full resolution convolutional network (FrCN) | INbreast | 92.63 | 86.37 |
| Li et al.32 | Conditional Residual UNet | INbreast | 92.72 | – |
| Proposed architectures | Connected-UNets | INbreast | 95.16 | 90.77 |
| CBIS-DDSM | 87.02 | 77.07 | ||
| Connected-AUNets | INbreast | 94.89 | 90.28 | |
| CBIS-DDSM | 87.95 | 78.89 | ||
| Connected-ResUNets | INbreast | 95.28 | 91.03 | |
| CBIS-DDSM | 89.52 | 80.02 |