Enhancing skin cancer detection using transfer learning and AdaBoost: a deep learning approach

Abstract

Skin cancer is one of the most prevalent types of cancer worldwide, with early detection playing a critical role in improving patient outcomes. In this study, we propose a deep learning model based on LeNet-7 combined with adaptive boosting (AdaBoost) to classify skin lesions as either benign or malignant using the International Skin Imaging Collaboration (ISIC) dataset. We evaluate the proposed model alongside other well-established deep learning architectures, such as residual network (ResNet), VGGNet, and the traditional LeNet model, through various performance metrics including precision, recall, F1-score, specificity, Matthew’s correlation coefficient (MCC), area under the receiver operating characteristic curve (AUC-ROC), and testing accuracy. Our results demonstrate that the proposed model (LeNet-7+AdaBoost) significantly outperforms the other models, achieving a testing accuracy of 91.3%, precision of 0.92, recall of 0.91, and AUC-ROC of 0.93. The model successfully addresses issues of overfitting and generalization, providing a robust solution for skin cancer classification. However, some misclassifications of visually similar benign and malignant lesions highlight areas for future improvement. The proposed model shows promise in real-world medical applications and paves the way for further research into optimizing deep learning models for skin cancer detection.