Enhancing Skin Cancer Classification Through GAN-Generated Synthetic Images for Improved CNN Trainings

Abstract: Skin cancer is among the most prevalent forms of cancer globally, and early detection is crucial for improving patient outcomes. While Convolutional Neural Networks (CNNs) have demonstrated strong performance in image-based skin cancer classification, their effectiveness is highly dependent on large and diverse datasets. However, medical image datasets often suffer from class imbalance and limited sample sizes. This study presents a novel approach to augmenting training data by leveraging Generative Adversarial Networks (GANs) to synthesize high-quality skin lesion images. These synthetic images are used to enhance the training set for a CNN-based classification model. Experimental results show a significant improvement in classification accuracy and robustness when synthetic images are included, particularly for underrepresented classes. The findings highlight the potential of GAN-augmented datasets in addressing data scarcity and improving diagnostic performance in medical imaging applications. Keywords: Skin Cancer Classification, Convolutional Neural Networks (CNNs), Generative Adversarial Networks (GANs), Data Augmentation, Synthetic Medical Images, Deep Learning, Class Imbalance, Medical Image Analysis.