Segmenting melanoma Lesion using Single Shot Detector (SSD) and Level Set Segmentation Technique

摘要: Melanoma is a lethal type of skin cancer that orginates fron melanocytes cells and it responsible several deaths annually due to exposure ultraviolet radiations. Early diagnosis proper treatment melanoma significantly improves the patient's survival rate. In computer aided diagnosis, automatic segmentation first step in early accurate lesion area. However, presence natural or clinical artifacts hinders precise segmentation. The goal our work establish novel pipeline automatically pre-process, localize then segment precisely improve its accuracy. proposed method, dermoscopic images are segmented three steps: 1. Preprocessing using morphological operations remove hair. 2. Localization by utilizing deep convolutional neural network named as Single-Shot Detection (SSD) network, 3. Segmentation level set algorithm. approach was evaluated on ISBI 2016 challenge dataset (Skin Lesion Analysis Towards Challenge Dataset). On ISIC 2016, method achieved an average Jc, Di Ac 0.82, 0.901 0.90 respectively. results also compared with state-of-the-art methods justify effectiveness approach.