Table 10 Comparison of proposed methodology with state-of-the-art techniques and competition participants.The results presented in this table are resulted from models trained and tested on individual ISIC-2016 and ISIC-2017 datasets.
From: Melanoma segmentation using deep learning with test-time augmentations and conditional random fields
Method | Dice | Jaccard | Precision | Recall |
|---|---|---|---|---|
Trained and tested on ISIC-2016 | ||||
Team-EXB13 | 91.00 | 84.30 | 96.50 | 91.00 |
Team-CUMED23 | 89.70 | 82.90 | 91.10 | 95.70 |
Team-Rahman13 | 89.50 | 82.22 | 88.00 | 96.90 |
Bi et al.5 | 91.18 | 84.64 | 92.17 | 96.54 |
Yuan et al.6 | 91.20 | 84.70 | 96.60 | 91.80 |
Lesion Net11 | 92.39 | 86.47 | 96.45 | 93.62 |
Proposed UNet | 90.39 | 85.96 | 87.46 | 92.76 |
Proposed ResUNet | 88.33 | 84.15 | 84.75 | 92.22 |
Proposed ResUNet++ | 90.27 | 85.88 | 93.60 | 86.27 |
Trained and tested on ISIC-2017 | ||||
Yuan et el. (CDNN)6 | 84.90 | 76.50 | 97.50 | 82.50 |
Li et al.24 | 84.70 | 76.20 | 97.80 | 82.00 |
Bi et al. (ResNet)25 | 84.40 | 76.00 | 98.50 | 80.20 |
Lin et al. (UNet)7 | 77 | 62.00 | – | – |
Al-masni et al. (FrCN)12 | 87.08 | 77.11 | 96.69 | 85.40 |
Kashan et al. (ResUNet)10 | 85.80 | 77.20 | – | – |
Lesion Net11 | 87.87 | 78.28 | 96.08 | 86.23 |
Proposed UNet | 90.39 | 80.05 | 86.77 | 82.17 |
Proposed ResUNet | 79.28 | 77.66 | 82.46 | 80.47 |
Proposed ResUNet++ | 82.96 | 80.03 | 84.01 | 85.26 |
