AI & ML interests
Multimodal AI for biomedicine
Recent Activity
Papers
CONFLUX: A Latent Diffusion Model for 3D Chest-CT Synthesis with RL Post-Training
Discrete Diffusion Language Models for Interactive Radiology Report Drafting
Organization Card
The Gevaert Lab develops machine-learning methods that integrate genomics, pathology, imaging, and clinical data — advancing precision medicine in oncology and cardiovascular disease, and building toward a medical digital twin that predicts disease trajectories and treatment response.
Stanford Medicine · Department of Medicine & Biomedical Data Science · Directed by Olivier Gevaert
Lab website · Publications · GitHub · Stanford profile
Research
We pursue the twin across the molecular-to-clinical stack, each area anchored by open methods:
- Molecular & epigenomic modeling — driver genes and methylation subtypes from multi-omics (MethylMix, AMARETTO).
- Computational pathology — gene expression from whole-slide images (SEQUOIA) and synthetic tissue generation.
- Quantitative imaging & radiogenomics — imaging phenotype linked to molecular state and outcome (LungNet).
- Integration & generative modeling — multimodal fusion and meta-learning (GeNNius); conditional 3D image synthesis (CONFLUX, BrainG3N).
Open resources — models, datasets, and demos are listed below (non-commercial research; see each repository for license and citation).
models 162
gevaertlab/conflux
Unconditional Image Generation • Updated • 25 • 2
gevaertlab/diffusiongemma-radiology-vqa
Image-Text-to-Text • Updated • 2
gevaertlab/he2rna-ucec-4
Updated • 2
gevaertlab/he2rna-ucec-3
Updated • 1
gevaertlab/he2rna-ucec-2
Updated • 1
gevaertlab/he2rna-ucec-1
Updated • 1
gevaertlab/he2rna-ucec-0
Updated • 1
gevaertlab/he2rna-thca-4
Updated • 1
gevaertlab/he2rna-thca-3
Updated • 1
gevaertlab/he2rna-thca-2
Updated • 1