A Comprehensive Approach for Skin Cancer Detection and Classification Using Machine Learning and Deep Neural Network Algorithm

Millions of people worldwide are affected by skin cancer each year, making it a serious public health issue, caused by environmental and genetic factors. In the United States alone, 5.4 million [1, 2] people are affected by skin cancer and diagnosed every year. The worldwide figures are equally alarming [3, 4]. The deadliest type of skin cancer is melanoma. Early detection of skin cancer is a difficult and expensive process. Early diagnosis and detection are crucial because treatment at early stages can increase the chance of cure, recovery, and survival rates. Traditional approaches have long been plagued by issues including biased outcomes and imprecision due to the similarity of skin cancer lesions. Therefore, this study developed an approach that uses conventional camera images from a dataset to identify and classify skin abnormalities as benign or malignant. We have used techniques to develop a model that detects and classifies seven kinds of skin cancer specifically Vascular Lesions, Dermatofibroma, Basal Cell Carcinoma, Melanocytic Nevi, Melanoma, Benign Keratosis-Like Lesions, and Actinic Keratoses with an accuracy of 98.56% by a combination of deep neural network and machine learning algorithms with image processing technologies using a dataset from Kaggle. There are 10,015 dermatoscopic images in the dataset. Additionally, we have created a web application with an intuitive user interface that enable users to rapidly upload dermatoscopic images to identify and classify skin cancer.