Balancing Fidelity, Utility, and Privacy in Synthetic Cardiac MRI Generation: A Comparative Study
2026-03-04 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionMachine Learning
AI summaryⓘ
The authors studied how to create fake but useful heart MRI images using three different types of AI models to help with limited real data and privacy rules. They tested Denoising Diffusion Probabilistic Models (DDPM), Latent Diffusion Models (LDM), and Flow Matching (FM) to see which made the best images that are both helpful for medical tasks and protect patient privacy. Their results showed DDPM works best overall, balancing good image quality, usefulness, and privacy. FM also showed strong privacy but wasn’t as good for medical tasks. This work helps understand how to safely make synthetic medical data for research.
Cardiac MRIDeep learningSynthetic dataDenoising Diffusion Probabilistic ModelsLatent Diffusion ModelsFlow MatchingImage synthesisData privacyMedical imagingSegmentation
Authors
Madhura Edirisooriya, Dasuni Kawya, Ishan Kumarasinghe, Isuri Devindi, Mary M. Maleckar, Roshan Ragel, Isuru Nawinne, Vajira Thambawita
Abstract
Deep learning in cardiac MRI (CMR) is fundamentally constrained by both data scarcity and privacy regulations. This study systematically benchmarks three generative architectures: Denoising Diffusion Probabilistic Models (DDPM), Latent Diffusion Models (LDM), and Flow Matching (FM) for synthetic CMR generation. Utilizing a two-stage pipeline where anatomical masks condition image synthesis, we evaluate generated data across three critical axes: fidelity, utility, and privacy. Our results show that diffusion-based models, particularly DDPM, provide the most effective balance between downstream segmentation utility, image fidelity, and privacy preservation under limited-data conditions, while FM demonstrates promising privacy characteristics with slightly lower task-level performance. These findings quantify the trade-offs between cross-domain generalization and patient confidentiality, establishing a framework for safe and effective synthetic data augmentation in medical imaging.