Table 2 Segmentation performance of our proposed networks on the test sets.
From: Connected-UNets: a deep learning architecture for breast mass segmentation
Proposed architectures | Dice score (%) | IoU score (%) | Dice score (%) | IoU score (%) | Dice score (%) | IoU score (%) |
|---|---|---|---|---|---|---|
| Â | CBIS-DDSM | INbreast | Private | |||
Standard UNet | 78.62 | 64.87 | 89.21 | 79.5 | 89.87 | 86.43 |
Connected-UNets | 82.22 | 69.82 | 93.36 | 85.75 | 95.72 | 91.95 |
Standard AUNet | 80.39 | 67.29 | 91.35 | 82.59 | 90.25 | 88.02 |
Connected-AUNets | 83.84 | 72.19 | 93.52 | 86.01 | 95.82 | 92.17 |
Standard ResUNet | 80.94 | 68.05 | 92.71 | 84.58 | 93.58 | 89.79 |
Connected-ResUNets | 85.01 | 73.95 | 94.13 | 87.63 | 95.88 | 92.27 |
