vesselFM

TL;DR: VesselFM is a foundation model for universal 3D blood vessel segmentation in arbitrary imaging domains. For details, please refer to our manuscript and our GitHub repo.

Checkpoints

We provide the following checkpoints:

• vesselFM_all.pt: VesselFM pre-trained on a version of our three proposed data sources (D_real, D_drand, and D_flow), including our evaluation set for potential increased performance. This checkpoint will be automatically downloaded in vesselfm/seg/inference.py.

• vesselFM_all_ema.pt: EMA version of vesselFM_all.pt.

• vesselFM_base.pt: VesselFM pre-trained on a version of our three proposed data sources (D_real, D_drand, and D_flow). Performance is similar to the model reported in the manuscript.

Citing vesselFM

If you find our work useful, please cite:

@InProceedings{Wittmann_2025_CVPR,
    author    = {Wittmann, Bastian and Wattenberg, Yannick and Amiranashvili, Tamaz and Shit, Suprosanna and Menze, Bjoern},
    title     = {vesselFM: A Foundation Model for Universal 3D Blood Vessel Segmentation},
    booktitle = {Proceedings of the Computer Vision and Pattern Recognition Conference (CVPR)},
    month     = {June},
    year      = {2025},
    pages     = {20874-20884}
}