This project focuses on comparing the generation of brain MRI images using three types of Generative Adversarial Networks (GANs): Deep Convolutional GAN (DCGAN), Super-Resolution GAN (SRGAN), and Conditional GAN (cGAN). Brain tumors, whether benign or malignant, pose significant health risks due to their impact on brain function and the potential for life-threatening complications. Magnetic Resonance Imaging (MRI) is a crucial diagnostic tool for detecting brain tumors, but obtaining a large and diverse dataset of real MRI images can be challenging. In this project, we leverage the power of GANs to address data scarcity in terms of brain MRI images.
The project explores the application of GANs to produce synthetic brain MRI images. GANs have shown significant potential in generating synthetic MRI data that can capture the distribution of real MRI images. GANs are also popular for tasks such as segmentation, noise removal, and super-resolution of brain MRI images. Specifically, we employ DCGANs, a class of deep-learning models that learn the underlying data distribution from available samples. By training GANs on existing brain MRI images, we aim to generate new, realistic MRI scans that can augment the dataset for brain tumor detection.