Research portfolio
Selected research projects in medical image computing, structural representation learning, and generative modeling.
ECCV 2026 Spotlight
Tokenizing Vessel-like 3D Biomedical Graph Representations for Reconstruction and Generation
Compact learned representations for faithful reconstruction and generation of complex three-dimensional vascular and airway graphs.
MICCAI 2025
Semantically Consistent Discrete Diffusion for 3D Biological Graph Modeling
Generating anatomically plausible 3D biological graphs by correcting semantic and structural inconsistencies during discrete diffusion.
Research overview
MICCAI 2024
3D Vessel Graph Generation Using Denoising Diffusion
A two-stage diffusion framework for synthesizing realistic vascular graphs, including networks with cycles.
Research overview
MIDL 2023 · PMLR Vol. 227
Towards Complete 3D Vessel Graph Generation from Images
Directly predicting radius-aware 3D vessel graphs from images and merging patch-wise predictions into full-volume networks.
Published paper Figure 2.
MIDL 2023 · PMLR Vol. 227
Improving Vision Transformer Autoencoder for Self-supervised Medical Image Representations
Self-supervised medical image learning with masked reconstruction, structure-aware losses and contrastive masked views.
Original paper Figure 1
MICCAI 2023
Self-Pruning Graph Neural Network for Predicting Inflammatory Disease Activity in Multiple Sclerosis from Brain MR Images
Learning a patient-specific lesion graph from brain MRI and identifying the lesions most relevant to predicting future inflammatory activity.
Original MICCAI 2023 Figure 1.
