A Comparative Analysis of Pre-Trained Convolutional Neural Networks for Melanoma Early Detection

One of the most prevalent and lethal types of cancer in the world is skin cancer, especially melanoma. Improving patient outcomes requires early detection, and automated systems powered by deep learning (DL) have demonstrated significant potential to support medical professionals. Several pre-trained convolutional neural network (CNN) models used for melanoma early detection are compared in this study. They are VGG16, MobileNetV2, EfficientNetB7, and DenseNet121. The HAM10000 dataset, which includes a wide range of skin lesions, was used to assess these models. The aim was to find the best pre-trained model for the classification of melanoma. We compared the accuracy, precision, recall, and F-1 scores of various models to assess their capacity to extract reliable features from the dataset. The CNN (2D) model performed the best, according to the data, in terms of accuracy (87%), precision (86%), recall (85%), and F1-score (88%). By thoroughly contrasting the most widely used pre-trained models and revealing their relative advantages and disadvantages for melanoma diagnosis, this study advances the field. According to the results, these pre-trained models can greatly improve automated melanoma detection, facilitating early diagnosis and offering dermatologists and other healthcare professionals invaluable assistance.