
MSc Dissertation: Synthetic Histopathology Image Generation with Diffusion Models
University of Surrey, Guildford, UK
My MSc thesis focused on addressing data scarcity and privacy constraints in computational pathology by generating high-fidelity synthetic histopathology images. The project involved developing a multi-faceted evaluation framework and fine-tuning a Latent Diffusion Model (Stable Diffusion v1.5) on the TCGA-BRCA dataset. The generated data was proven to be a safe and effective tool for data augmentation, significantly improving a downstream cancer classifier's performance (up to a +.247 AUROC increase) in simulated data-scarce scenarios.