Table 7 Performance comparison across datasets and recent state-of-the-art methods.
From: Advancing skin cancer diagnosis with deep learning and attention mechanisms
Model | Dataset | Dice score | Accuracy (%) | Sensitivity (%) | Precision (%) | F1-Score (%) |
|---|---|---|---|---|---|---|
Proposed Model | HAM10000 (Original) | 0.988 | 97.8 | 98.3 | 98.1 | 98.2 |
ISIC (Preliminary) | 0.91 | 91.2 | 90.7 | 92.5 | 91.0 | |
PH2 (Preliminary) | 0.933 | 93.5 | 92.0 | 94.0 | 93.3 | |
UNet | HAM10000 (Original) | 0.850 | 88.4 | 85.5 | 85.2 | 85.3 |
ISIC (Preliminary) | 0.78 | 84.5 | 83.2 | 81.3 | 82.1 | |
PH2 (Preliminary) | 0.80 | 85.2 | 84.0 | 82.5 | 83.2 | |
DenseNet | HAM10000 (Original) | 0.870 | 90.1 | 88.2 | 89.5 | 88.9 |
ISIC (Preliminary) | 0.83 | 88.2 | 86.5 | 87.8 | 87.1 | |
PH2 (Preliminary) | 0.85 | 89.0 | 87.4 | 88.2 | 87.8 | |
Attention UNet | HAM10000 (Original) | 0.920 | 94.0 | 93.0 | 93.4 | 93.2 |
ISIC (Preliminary) | 0.85 | 89.5 | 87.8 | 89.0 | 88.4 | |
PH2 (Preliminary) | 0.88 | 90.3 | 89.5 | 90.2 | 89.9 | |
UNet++ | HAM10000 (Original) | 0.930 | 94.5 | 94.0 | 94.8 | 94.3 |
ISIC (Preliminary) | 0.87 | 92.0 | 90.0 | 91.2 | 90.6 | |
PH2 (Preliminary) | 0.90 | 92.8 | 91.6 | 92.2 | 91.9 | |
TransUNet | HAM10000 (Original) | 0.950 | 95.2 | 94.7 | 94.9 | 94.8 |
ISIC (Preliminary) | 0.88 | 90.0 | 88.2 | 89.5 | 88.8 | |
PH2 (Preliminary) | 0.91 | 93.1 | 92.4 | 93.0 | 92.7 |
