An Ensemble of Transfer Learning Models for the Prediction of Skin Lesions with Conditional Generative Adversarial Networks

Abstract

Skin cancer is one of the most serious forms of the disease, and it can spread to other parts of the body if not detected early. Therefore, it is crucial to diagnose and treat skin cancer patients at an early stage. Due to the fact that a manual diagnosis of skin cancer is both time-consuming and expensive, an incorrect diagnosis is made due to the high degree of similarity between the various skin lesions. Improved categorization of multi-class skin lesions requires the development of automated diagnostic systems. We offer a fully automated method for classifying several skin lesions by fine-tuning the deep learning models, namely VGG16, ResNet50, and ResNet101. Prior to model creation, the training dataset should undergo data augmentation using traditional image transformation techniques and generative adversarial networks (GANs) to prevent class imbalance issues that may lead to model overfitting. In this study, we investigate the feasibility of creating dermoscopic images that have a realistic appearance using conditional generative adversarial network (CGAN) techniques. Afterward, the traditional augmentation methods are used to augment our existing training set to improve the performance of pretrained deep models on the skin lesion classification task. This improved performance is then compared to the models developed using the unbalanced dataset. In addition, we formed an ensemble of finely tuned transfer learning models, which we trained on balanced and unbalanced datasets. These models were used to make predictions about the data. With appropriate data augmentation, the proposed models attained an accuracy of 92% for VGG16, 92% for ResNet50, and 92.25% for ResNet101. The ensemble of these models increased the accuracy to 93.5%. There was a comprehensive discussion on the performance of the models. It is possible to conclude that using such a method leads to enhanced performance in skin lesion categorization compared to the efforts made in the past.