Table 1 Summary of existing techniques in skin lesion segmentation
From: Skin lesion segmentation using deep learning algorithm with ant colony optimization
Reference | Study focus | Methods used | Dataset(s) | Key findings |
|---|---|---|---|---|
Lallas, et al. [14] | Automated skin lesion segmentation | U-Net architecture with data augmentation | ISIC 2018 | Achieved a Dice coefficient of 85% and demonstrated the effectiveness of U-Net for segmentation tasks. |
Ronneberger, et al. [15] | Multi-class skin lesion classification | Deep learning (ResNet, DenseNet) | ISIC 2018 | Enhanced classification accuracy by incorporating a multi-class approach, achieving an accuracy of 89%. |
Cordes, et al. [16] | Skin lesion segmentation with ensemble models | Ensemble of CNNs | ISIC 2018 | Improved segmentation accuracy by combining multiple CNN models, leading to a Dice coefficient of 87%. |
Winkler, et al. [17] | Transfer learning for skin lesion detection | Transfer learning (VGG19, ResNet-50) | ISIC 2018 | Demonstrated the effectiveness of transfer learning, achieving an accuracy of 90% in lesion detection. |
Kuo, et al. [18] | Comparative analysis of segmentation methods | Comparison of CNN, U-Net, and SegNet | ISIC 2018 | Compared different segmentation methods, with U-Net performing the best with a Dice coefficient of 86%. |