DeepMelaNet: Advancing Melanoma Stage Classification in Skin Cancer Diagnosis

Abstract

Melanoma skin cancer is a global public health threat due to its increasing rates and the possibility of severe outcomes if not adequately addressed. Melanoma is caused by ultraviolet radiation and, among its two stages, malignant is more dangerous than benign. The diagnosis of melanoma is typically based on visual inspection and manual methods carried out by experienced physicians. However, this method is usually slow and has a high probability of error. Deep-learning-based models can offer better and low-cost treatments for people with melanoma. This study aimed to develop a deep-learning model to classify melanoma skin cancer in its early stages. This study presents a modified deep-learning model, named DeepMelaNet, to correctly classify skin cancer images as benign or malignant. The proposed classifier achieved an accuracy of 93.40%, a precision of 98%, a recall of 94%, and an F1 score of 93% on a dataset of 10,000 melanoma skin cancer images, offering a practical solution that can help healthcare professionals in early skin cancer prediction.