MELANOMA SKIN CANCER DETECTION MODEL USING CONVOLUTIONAL NEURAL NETWORKS

ABSTRACT

Skin cancer is a common type of cancer. Despite being the least common type of skin cancer, melanoma, also known as malignant melanoma, is the deadliest, accounting for 75% of all skin cancer deaths. Trying to detect it as early as possible and treating it with minimal surgery is the best strategy to combat it. In this study, we comprehensively examine melanoma and find that convolutional neural networks perform better when given deeper, wider, and higher resolution images. We suggest an automated melanoma detection model based on these findings that analyses skin lesion photos using Matlab's deep network designer. This model used the International Skin Imaging Collaboration (ISIC) 2020 Challenge Dataset, which contains images from a variety of primary medical sources for both training and testing. When compared to more established melanoma classifiers on the same dataset from earlier evaluations, experimental results on this dataset demonstrated state-of-the-art classification performance. A 91% accuracy rate was achieved by the CNN, which could reduce human mistakes in the diagnosis process.