Advancements in Convolutional Neural Networks for Accurate and Efficient Skin Cancer Classification: A Comprehensive Survey

The cancer of the skin, particularly melanoma, is considered to be one of the most life-threatening types of cancer. The major cause of skin cancer is an unrepaired deoxyribonucleic acid (DNA) disruption in skin cells which results in heritable skin imperfections or mutations that result from the disruptions. It is very important to note that skin cancers are most curable at their earliest stages, but they tend to spread to multiple parts of the body and appear as lesions. Among the lesion parameters that doctors use to identify and distinguish between non-cancer skin cancer and melanoma skin cancer is similarity, structure, spread, color, etc. These parameters provide the doctors with a valuable tool to diagnose skin cancer early on. Due to the parameter study and significant increase in skin cancer incidence rates, many researchers decided to quickly identify and diagnose skin cancer. Early detection of skin cancer is essential to lowering the mortality rate. Using deep learning and transfer learning models, researchers have been able to detect skin cancer using various algorithms. A recent study demonstrates the capability of deep learning and deep transfer learning methods to classify and detect skin cancer early. The results of this study suggest that a comprehensive investigation into the methods of Deep Learning (DL) algorithms for locating skin cancer at several stages and types is required based on the analysis of the results of this study. An investigation into articles published in reputed journals was carried out with regard to the diagnosis of skin cancer based on the findings.