Melanoma Detection from Skin Lesions using Convolution Neural Network

Melanoma is a deadly cancer of the skin as well as being the second most common cancer in elderly people. Females aged 15-29 are also prone to this disease. It is regarded to be a fast-growing cancer, although it can be cured in early measures, it spreads faster if it is not treated in the preliminary stage. Even a person may develop melanoma if he or she has never incurred any sun damage. Cancer longevity is 5 years of survival with a 92 percent likelihood, which means 92 out of 100 people diagnosed with melanoma in 5 years, will be alive. Generally, melanoma occurs when the Pigment-producing cells which provide the skin with pigment become malignant. This paper focuses on detecting whether the skin lesion image is benign or malignant. Here, a convolutional neural network (CNN) has been used to predict melanoma from images of skin lesions. In this study, Support Vector Machine has also been used for a comparison purpose. However, CNN outperformed SVM and produced 96% accuracy. Thus, it can be used for real-life applications.