Table 2 Background/benign/malignant segmentation results.
From: BUS-UCLM: Breast ultrasound lesion segmentation dataset
Network | Class | IoU | Acc | Dice | Precision | Recall |
|---|---|---|---|---|---|---|
UNet | Background | 95.63% | 98.41% | 97.77% | 97.13% | 98.41% |
Benign | 33.65% | 41.09% | 50.20% | 64.76% | 41.09% | |
Malignant | 41.55% | 47.51% | 58.66% | 75.62% | 45.96% | |
Average | 56.94% | 62.34% | 68.87% | 79.17% | 61.82% | |
AttUNet | Background | 95.66% | 98.21% | 97.78% | 97.35% | 98.19% |
Benign | 33.41% | 40.41% | 50.01% | 66.42% | 40.41% | |
Malignant | 43.53% | 52.51% | 60.56% | 72.61% | 52.51% | |
Average | 57.53% | 63.71% | 69.45% | 78.79% | 63.70% | |
Sk-UNet | Background | 95.54% | 98.13% | 97.72% | 97.17% | 98.13% |
Benign | 34.12% | 41.50% | 50.77% | 61.67% | 42.06% | |
Malignant | 42.83% | 51.97% | 59.78% | 62.42% | 47.09% | |
Average | 57.50% | 63.87% | 69.43% | 73.75% | 62.43% | |
DeepLabv3 | Background | 95.74% | 98.34% | 97.82% | 97.31% | 98.36% |
Benign | 33.16% | 40.62% | 49.76% | 58.98% | 40.54% | |
Malignant | 40.78% | 49.82% | 57.80% | 69.83% | 49.57% | |
Average | 56.56% | 62.93% | 68.46% | 75.37% | 62.82% | |
Mask R-CNN | Background | 96.42% | 98.71% | 98.18% | 97.65% | 98.71% |
Benign | 45.85% | 60.08% | 62.87% | 65.93% | 60.08% | |
Malignant | 54.12% | 64.36% | 70.23% | 77.28% | 64.36% | |
Average | 65.46% | 74.38% | 77.09% | 80.29% | 74.38% |
