Abstract

Malignant melanoma represents a highly aggressive form of skin cancer associated with significant mortality rates in the absence of early detection. Consequently, the establishment of precise diagnostic protocols is imperative. Conventional diagnostic methods, which rely heavily on microscopic examination and the clinical expertise of dermatologists, are often labor-intensive and subject to interpretation variability. To address these limitations and enhance diagnostic precision, Artificial Intelligence (AI) has emerged as a valuable adjunctive tool for clinicians. This study investigates the efficacy of Deep Learning (DL) methodologies for melanoma detection utilizing cutaneous image processing. A comparative analysis was conducted across a spectrum of Convolutional Neural Network (CNN) architectures, specifically DenseNet201, MobileNetV2, ResNet50V2, ResNet152V2, Xception, VGG16, VGG19, and GoogleNet. All models were trained and evaluated on a graphical processing unit (GPU) environment using a dataset comprising 7,146 images. The empirical results demonstrate that the GoogleNet architecture yielded the superior performance, achieving accuracies of 74.91% on the training set and 76.08% on the testing set.