Skin Lesion Segmentation Based on Edge Attention Vnet with Balanced Focal Tversky Loss

Segmentation of skin lesions from dermoscopic images plays an essential role in the early detection cancer. However, lesion segmentation is still challenging due to artifacts such as indistinguishability between and normal skin, hair on reflections obtained dermoscopy images. In this study, edge attention network (ET-Net) combining guidance module (EGM) weighted aggregation added 2D volumetric convolutional neural (Vnet 2D) maximize performance segmentation. addition, proposed fusion model presents a new loss function by balanced binary cross-entropy (BBCE) focal Tversky (FTL). The has been tested ISIC 2018 Task 1 Lesion Boundary Challenge dataset. outperformed state-of-the-art studies result tests.