Dermoscopic Image-Based Skin Cancer Detection Using a CNN-Driven U-Net Model

Skin cancer is one of the most serious diseases that people can have. Skin cancer detection is frequently complicated by the present healthcare system, which can result in potentially fatal delays if not detected promptly. Early identification is crucial to a good recovery from skin cancer. Deep learning (DL) algorithms have been a popular approach for detecting skin cancer in recent years. The goal of this research is to use a DL model to create a multi-classification approach for the diagnosis of skin cancer. In this work, we introduce a novel deep-learning skin cancer classification network model, U Net, which is based on convolutional neural networks (CNNs). We evaluate it using three open-source standard datasets: ISIC 2020, HAM10000, and PH2. The Proposed U Net achieved an accuracy of 89.29%, a recall of 84.52%, a precision of 84.56%, and an F1-score of 88% in the classification of the various types of skin cancer. Our suggested UNet model fared better than baseline models, offering dermatologists and other healthcare providers significant support in the study of skin cancer.