All publications

Conference paper / 2026

GReAT: Leveraging Geometric Artery Data to Improve Wall Shear Stress Assessment

Julian Suk, Jolanda J. Wentzel, Patryk Rygiel, Joost Daemen, Daniel Rueckert, Jelmer M. Wolterink

Shape in Medical Imaging (ShapeMI) · MICCAI 2025 workshop · Published · 2025 conference, 2026 proceedings

Three coronary artery surfaces display regions of low, intermediate and high wall shear stress in pink, yellow and green.
Coronary examples from the IMPACT dataset illustrate spatial wall-shear-stress classes used in the study: right coronary, left anterior descending and left circumflex arteries. Figure 1 · Suk et al. (2025 preprint) · Suk et al. · CC BY 4.0. Original figure from arXiv:2508.19030v1.

Learning from unlabeled artery geometry

Large collections of vascular shapes can provide useful training information even without matching flow simulations. GReAT uses heat-kernel signatures as a self-supervised learning target, then transfers the resulting geometric representation to coronary wall shear stress assessment.

The downstream task identifies regions with low, medium, or high time-averaged wall shear stress. The study examines whether geometric pretraining helps when only a small patient cohort is available.

Cite this work

@inproceedings{great2026,
  title = {GReAT: Leveraging Geometric Artery Data to Improve Wall Shear Stress Assessment},
  author = { Julian Suk and Jolanda J. Wentzel and Patryk Rygiel and Joost Daemen and Daniel Rueckert and Jelmer M. Wolterink },
  year = {2026},
  booktitle = {Shape in Medical Imaging (ShapeMI)},
  pages = {277--291},
  doi = {10.1007/978-3-032-06774-6_21},
  eprint = {2508.19030},
  archivePrefix = {arXiv},
  url = { https://doi.org/10.1007/978-3-032-06774-6_21 }
}