Proposed Convolution Neural Network for Skin Cancer Diagnosis and Classification

Abstract

Skin cancer is a lethal condition that, if not detected in its early stages, becomes more difficult to cure and can have fatal outcomes. Thus, skin cancer must be diagnosed accurately, precisely, and as early as possible so that it doesn't progress into further stages. Traditional methods for diagnosing skin cancer involve numerous tests and consultations with dermatologist experts. Because many kinds of skin cancer might seem similar, especially in their early stages, correct skin cancer detection can be challenging, even for dermatologist experts. This paper proposed a Convolutional Neural Network (CNN) for diagnosing and stratifying skin cancer into seven classes. The proposed CNN model consists of 26 layers. The images utilized for training and testing the model were obtained from the HAM10000 dataset, which was then augmented using various techniques and then classified by the proposed CNN model into seven labeled classes, including AKIEC, BCC, BKL, DF, MEL, NV, and VASC. The presented CNN model was shown to have a high accuracy of 99.94%, outperforming state-of-the-art algorithms for accurately diagnosing and categorizing skin cancer. This paper aims to prevent premature mortality, provide health in resource-constrained settings, and seek patients' healthy lives, which can be done through an accurate and early-stage skin cancer diagnosis.