SegSkin: An Effective Application for Skin Lesion Segmentation Using Attention-Based VGG-UNet

Skin cancer is a worldwide health problem with high morbidity and mortality rates. To successfully diagnose skin lesions, computer-aided automatic diagnostic systems are required. It offers dermatologists a reliable approach to assessing skin lesions. The segmentation is difficult because of the artifacts in dermoscopic images like hair and the low contrast. In this paper, we present a new model that integrates two architectures, the U-Net and the VGG-19, as well as the attention mechanism. Furthermore, to improve the segmentation results, we also employ image preprocessing, including the Dull-Razor algorithm for hair removal and Contrast Limited Adaptive Histogram Equalization (CLAHE) to improve the image quality. We implement our approach as an online application to help the dermatologist segment the skin lesion for further diagnosis; this application is named SegSkin. Moreover, we evaluate our model on four datasets: ISIC 2016, ISIC 2017, ISIC 2018, and PH2. The proposed model has obtained satisfactory results compared to the state-of-the-art.