Imagining Infinity: Endless CT Datasets through Conditional Diffusion Models
Name
Ekaterina Sedykh
Abstract
Medical imaging, the technique of creating visual representations of the interior of a body for clinical analysis and medical intervention, critically depends on the availability of extensive and high-quality datasets. However, the acquisition of such datasets is often limited by logistical, ethical, and privacy concerns. Diffusion models, known for their effectiveness in generating high-quality synthetic data, can mitigate these challenges by producing realistic medical images for model training and research purposes. This thesis addresses the challenge of data scarcity in medical imaging by exploring the efficacy of diffusion models in generating synthetic CT images that can be used to enhance dataset diversity and volume. Here, we demonstrate that diffusion models can effectively generate synthetic CT images that closely mimic real diagnostic images, thereby potentially expanding the breadth of available training data for medical AI applications. The results reveal that the synthetic images produced are not only almost indistinguishable from real images but also retain the necessary clinical details, which is an advancement over previous generative models that often sacrificed clinically relevant details. This work further exemplifies the utility of synthetic data generated by diffusion models in improving the training and performance of AI systems in diagnosing and analyzing medical images. The integration of diffusion models into medical imaging practices promises to significantly strengthen the AI-driven diagnostic tools. By providing a novel method for generating synthetic medical images, this research highlights the potential of advanced generative models in overcoming practical and ethical barriers in medical research.
Graduation Thesis language
English
Graduation Thesis type
Master - Computer Science
Supervisor(s)
Dmytro Fishman
Defence year
2024