Skin Cancer Detection Using Convolution Neural Network

Abstract

Skin cancer is a prevalent medical condition, with approximately 2 million people diagnosed annually. Shockingly, one person dies of skin cancer every hour, making it a crucial public health concern. Among the three types of skin cancer, Melanoma is the most lethal and notorious for its rapid spread. The significance of early detection of Melanoma cannot be overstated. According to studies, the five-year survival rate for melanoma patients whose cancer is detected early is 99%. However, if the disease has metastasized to other organs, the five-year survival rate drops to only 27%. This statistic emphasizes the importance of timely detection and treatment of this deadly disease. Given the high stakes involved, it is imperative to investigate various machine learning algorithms' effectiveness in detecting skin cancer accurately. Using machine learning algorithms, we can analyze features extracted from digital images of skin lesions to classify them as either Melanoma or Benign and even predict the likelihood of malignancy. This paper investigates different CNNs and their efficiencies in detecting skin cancer. In this paper, we have investigated the effectiveness of various CNN models, including Alexnet, Resnet50, and a customized CNN model, in detecting skin cancer.