Generative models improve radiomics: reproducibility and performance in low dose CTs

This thesis focused on using generative models to improve radiomics reproducibility and performance in low-dose CTs. Different generative models were included as testing models, and models were trained based on paired simulation data and unpaired real data. Before delving into the story of low-dose CT radiomics enhancement, this thesis investigated some details about generative models for low-dose denoising and applied low-dose CT radiomics to a new application. The results showed that low-dose CT radiomics achieved good performance in the new applications, and generative models can improve radiomics reproducibility and performance in low-dose CTs.