Generation of Intracranial Hemorrhage CT Scan Images using GAN and VAE Frameworks for Improved Medical Imaging

Intracranial hemorrhages are one of many neurological diseases that can be diagnosed and followed up on with the help of medical imaging. However, due to privacy concerns and need for more data, the availability of annotated medical imaging datasets, particularly those pertaining to severe illnesses like intracranial hemorrhages are frequently constrained. In order to solve the lack of medical data, synthetic data generation techniques like Generative Adversarial Networks (GAN) and Variational Autoencoders (VAE) have shown significant improvements. This research approaches the synthesis of intracranial hemorrhage computer tomography images through GAN and VAE constructs to advance deep-learning models for ICH identification. Performance evaluation of the proposed methodology relies on the dice coefficient measurements during synthetic dataset comparison against the small realworld dataset.