Deep Learning Based Pox Disease Detection and Generation of Synthesis Data with GAN Model

This study focuses on the classification of smallpox histopathological images using deep learning-based approaches. Given the challenges posed by limited dataset availability in medical image analysis, multiple techniques were employed to improve classification performance. Initially, a baseline classification was performed using a small dataset consisting of 600 images, which yielded a moderate accuracy of 75$\%$. To enhance the model's generalization capability, data augmentation techniques were applied, expanding the dataset by four times and increasing the classification accuracy to 96.23$\%$. Additionally, Generative Adversarial Networks (GAN) were used to generate synthetic data, further augmenting the dataset. The classification model trained on GAN-generated synthetic data achieved a high accuracy of 98.33$\%$ after proper hyperparameter tuning, demonstrating the efficacy of using synthetic data to improve performance. This study highlights the importance of data augmentation and synthetic data generation in medical image classification tasks, especially when working with limited datasets. The results suggest that a well-tuned deep learning model, combined with advanced data generation techniques, can provide accurate and reliable classification outcomes, potentially contributing to more effective diagnostic processes in clinical applications.