Melanoma Detection Among Various Skin Lesions

Melanoma is a type of skin cancer that starts and evolves from the “pigment-producing” cells known as “melanocytes.” There has been quite some research done in the area of melanoma classification through image detection and classification using machine learning—specifically deep learning and neural networks. Researchers have used convolutional neural networks (CNN), deep neural networks (DNN); some have even used recurrent neural network (RNN) and transfer learning. The research work has not been up to the mark as of yet, we know this because there has been no news of models being put to clinical testing. Another setback to the process of creating a perfect algorithm and mode is the lack of data regarding melanoma; the largest dataset publicly available is the one provided by ISIC for its 2020 competitions; it has 25,333 images in the dataset, but the issue with these is that they do not contain just the images and data for melanoma, they are a dataset for eight different skin lesions to be detected and classified. In this paper, we have proposed a model for melanoma detection using CNN architecture. We have also discussed the issues with our model and the accuracy achieved by it.