An efficient approach to detect melanoma skin cancer using a custom CNN model

Abstract

Despite only making up 1% of all occurrences of skin cancer, melanoma is one of the most prevalent forms to cause fatalities in recent years. Melanoma has a survival rate of more than 50% from the early stages to the end. To survive this type of cancer, it is essential to identify lesions on the skin early and to keep an eye out for any complications. If skin cancer is not detected and treated early, it is among the most fatal cancers. Of the skin cancers, which are among the deadliest, melanoma is the most unexpected. Like most other diseases, melanoma may be treatable if caught early enough. Due to the high cost of having a dermatologist screen every patient and the difficulty of human judgment, an automated system for melanoma diagnosis is required. Due to its promising pattern recognition skills, Convolutional Neural Network (CNN) models have recently gained a lot of interest in medical imaging. Melanoma diagnosis from dermoscopic skin samples automatically is a difficult task. In contrast to other types, melanoma ranks as the most serious type of skin cancer. However, those who are diagnosed early on have a better prognosis; several methods of spontaneous melanoma recognition and diagnosis have been researched by different researchers for the objective of providing a supplementary

opinion to professionals. Building models using existing data has proven problem- atic due to the imbalance between classes. However, these issues may be solved by

implementing a deep learning approach as a machine vision tool. The purpose of the current study was to determine how well dermoscopy and deep learning classified melanoma. In this paper, we introduce a brand-new deep learning model that was created to categorize melanoma skin cancer. And we have compared the result of our suggested model with pre-trained VGG16, VGG19, and AlexNet. According to experimental data, we discovered that our model worked well and could accurately categorize melanoma skin cancer. Also, the proposed system is competitive in the area of melanoma detection and superior in terms of accuracy and can be employed in the clinical decision-making procedure for melanoma skin cancer early detection.