Comparative prediction results on the ISIC2018
| Dataset | Models | Year | DSC↑ | SE↑ | SP↑ | ACC↑ |
|---|---|---|---|---|---|---|
| ISIC2018 | U-Net* | 2015 | 0.8851 | 0.8735 | 0.9744 | 0.9539 |
| SCR-Net* | 2021 | 0.8886 | 0.8892 | 0.9714 | 0.9547 | |
| C2SDG* | 2021 | 0.8806 | 0.8970 | 0.9643 | 0.9506 | |
| ATTENTION SWIN U-NET* | 2022 | 0.8540 | 0.8057 | 0.9826 | 0.9480 | |
| MALUNet* | 2022 | 0.8931 | 0.8890 | 0.9725 | 0.9548 | |
| UNeXt-S# | 2022 | 0.8910 | 0.8577 | 0.9818 | 0.9555 | |
| EGE-UNet* | 2023 | 0.8819 | 0.9009 | 0.9638 | 0.9510 | |
| VM-UNet* | 2024 | 0.8891 | 0.8809 | 0.9743 | 0.9554 | |
| VM-UNet v2* | 2024 | 0.8902 | 0.8959 | 0.9702 | 0.9551 | |
| LightM-UNet* | 2024 | 0.8898 | 0.8829 | 0.9765 | 0.9555 | |
| UltraLight VM-UNet* | 2024 | 0.8940 | 0.8680 | 0.9781 | 0.9558 | |
| UltraLight VM-UNet# | 2024 | 0.8905 | 0.8724 | 0.9766 | 0.9545 | |
| UCM-Net (Baseline) | 2024 | 0.9060 | 0.9041 | 0.9753 | 0.9602 | |
| MUCM-Net (1-patch) | 2024 | 0.9111 | 0.8953 | 0.9811 | 0.9629 | |
| MUCM-Net (2-patch) | 2024 | 0.9040 | 0.9151 | 0.9706 | 0.9588 | |
| MUCM-Net (4-patch) | 2024 | 0.9058 | 0.8752 | 0.9846 | 0.9614 | |
| MUCM-Net (8-patch) | 2024 | 0.9095 | 0.9046 | 0.9772 | 0.9618 |
* the results cited from UltraLight VM-Net; # the results tested by us. ISIC: International Skin Imaging Collaboration; DSC: Dice Similarity Coefficient; SE: sensitivity; SP: specificity; ACC: accuracy. ↑: the higher the value, the better the performance