Abstract
Accurate analysis of catheterization procedures is critical for improving safety and efficiency in X-ray guided endovascular interventions. Medical image analysis in this setting is severely constrained by limited annotated data and substantial domain variation. While diffusion-based generative models have shown promise for medical image synthesis, extending them to temporally coherent video remains challenging, particularly for thin, elongated structures such as catheters and guidewires, where small spatial errors accumulate across frames, disrupting structural continuity and degrading downstream segmentation performance. To address this, we propose a structure-aware diffusion framework for synthetic fluoroscopic video generation that explicitly enforces geometric consistency across frames. Building on a latent video diffusion backbone, we introduce three domain-specific objectives: shape-sensitive supervision via signed distance maps (SDMs) to preserve structural continuity, a domain-adaptive grayscale variational autoencoder to align latent representations with X-ray intensity statistics, and a perceptual similarity objective to restore fine-grained detail while maintaining temporal coherence. We evaluate the framework on real-animal and phantom X-ray datasets, demonstrating improved temporal coherence and structural fidelity over a strong diffusion baseline using Fréchet Video Distance (FVD), Structural Similarity Index Measure (SSIM), and Skeleton Dice (SD). Downstream segmentation results further confirm consistent performance gains in both centralized and federated learning settings under data scarcity and domain shift.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/DEMI_019.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=4SCBFzxwWR
BibTex
@InProceedings{KonCha_StructureAware_MICCAISAT2026,
author = { Kongtongvattana, Chayun AND Huang, Baoru AND Olajide, Olufemi AND Nguyen, Anh},
title = { { Structure-Aware Diffusion for Synthetic Fluoroscopic Video Generation } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17276},
month = {pending},
page = {pending}
}