Automated Skin Lesion and Cancer Classification Using AI-Driven-Deep Convolutional Networks

Abstract

Skin cancer is one of the most common cancers worldwide, and the longer it remains untreated, the greater the likelihood of serious consequences or even death. Traditional diagnostic techniques depend on the expert's visual evaluation and are thus subjective, laborious, and very variable. As a result, getting a proper diagnostic done quickly is crucial. To enhance the accuracy and efficacy of skin lesion diagnosis, this effort seeks to develop an automated classification system using dermoscopy images. The authors of this study propose building and testing a skin lesion classification AI system that makes use of various ML and DL techniques, including Xception, ResNet, CNN, ANN, and LSTM models. The proposed convolutional neural network (CNN) model achieved the highest reliability and accuracy in skin cancer diagnosis (98.49 F1-score, 98.58% ACC, 98.57% PRE, and 98.59% REC) compared to all other models. This method can help dermatologists identify skin disorders earlier and more accurately, which in turn improves clinical decision-making and decreases the incidence of misdiagnosis.