Enhancing myocardium and fibrosis segmentation in LGE-SA CMR images using diffusion models
Abstract
Myocardial scar segmentation from Late Gadolinium Enhancement (LGE)-Cardiac Magnetic Resonance (CMR) is essential for infarct quantification and clinical decision-making. Accurate measurement of scar tissue traditionally relies on manual delineation, which is time-consuming and observer-dependent. However, the development of automated and reproducible segmentation models remains technically challenging due to pronounced class imbalance, ill-defined tissue interfaces, and heterogeneous image quality. These challenges are further exacerbated by the limited availability of expert-annotated datasets, thereby necessitating data-efficient training paradigms. This dissertation examines the extent to which synthetic image generation using Diffusion Models can mitigate annotation scarcity for myocardium and scar segmentation. We propose a mask-conditioned diffusion model that synthesizes LGE image slices directly from input segmentation masks, thereby enabling anatomically constrained sampling at inference time without the need for accompanying reference images. The generative model is trained in a two-dimensional setting using a DDPM-style noise-prediction objective and is assessed as a source of synthetic training data for nnU-Net. In addition, a baseline learning curve derived exclusively from real data is established using nested, incrementally expanded subsets that preserve the prevalence of fibrosis. We further compare mask-conditioned diffusion models against exemplar-guided…
Read the paper
Related studies
- Learning from First-Generation Qualitative Approaches in the IS Discipline: An Evolutionary View and Some Implications for Authors and Evaluators (PART 1/2)
- Commentary on the clinical and preclinical dosage limits of interstitially administered magnetic fluids for therapeutic hyperthermia based on current practice and efficacy models
- Prosthesis: New Technological Opportunities and Innovative Biomedical Devices
- The history and future of neural modeling for cochlear implants
- Ryanodine receptors: waking up from refractoriness
- Novel Transgenic Rabbit Model Sheds Light on the Puzzling Role of Matrix Metalloproteinase-12 in Atherosclerosis
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.