Exponentially Carpet Weaving Optimization Enabled GoogleNet for Melanoma Classification using Skin Images

Abstract

Skin cancer is an invasive condition characterized by abnormal proliferation of melanocyte cells in body, where the cells can multiply and spread through lymph nodes for damaging nearby tissues. The early detection is essential for better treatment. Currently, it is widely recognized that the melanoma is the common skin cancer since it is significantly more likely to spread to other parts of the body when it is not treated or diagnosed promptly. To mitigate this challenge, an approach is proposed for melanoma classification using skin images named Exponentially Carpet Weaving Optimization enabled GoogleNet (ECWO_GoogleNet). The input skin image is pre-processed by Adaptive Kalman filter. Then, the skin lesion segmentation is performed by utilizing ENet, and then feature extraction is carried out. At last, the melanoma classification is conducted by GoogleNet, which is classified into Melanoma, Melanocytic nevus, Basal cell carcinoma, Actinic keratosis, Benign keratosis (solar lentigo / seborrheic keratosis / lichen planus-like keratosis), Dermatofibroma, Vascular lesion, and Squamous cell carcinoma. Here, GoogleNet is trained by ECWO. The analytic measures of ECWO_GoogleNet namely, Accuracy, True Positive Rate (TPR) and True Negative Rate (TNR) achieved 91.89%, 91.99%, and 91.27%.