Table 4.
Performance metrics for classification with dermatoscopic datasets.
| Method | AUC | Specificity | Sensitivity | F1-score |
|---|---|---|---|---|
| ISIC2016 | ||||
| Without segmentation | 0.765 | 0.726 | 0.860 | 0.864 |
| Contextual segmentation | 0.719 | 0.641 | 0.826 | 0.833 |
| Refined contextual segmentation | 0.727 | 0.698 | 0.844 | 0.845 |
| ISIC2017 | ||||
| Without segmentation | 0.790 | 0.741 | 0.761 | 0.740 |
| Contextual segmentation | 0.750 | 0.744 | 0.726 | 0.723 |
| Refined contextual segmentation | 0.774 | 0.785 | 0.766 | 0.762 |
| HAM 10,000 | ||||
| Without segmentation | 0.891 | 0.933 | 0.866 | 0.871 |
| Contextual segmentation | 0.831 | 0.884 | 0.825 | 0.810 |
| Refined contextual segmentation | 0.871 | 0.919 | 0.873 | 0.866 |