All publications

Preprint / 2025

Wall Shear Stress Estimation in Abdominal Aortic Aneurysms: Towards Generalisable Neural Surrogate Models

Patryk Rygiel, Julian Suk, Christoph Brune, Kak Khee Yeung, Jelmer M. Wolterink

arXiv · Preprint

CFD and a LaB-GATr surrogate estimate transient wall shear stress; four panels show the evaluated generalization settings.
CFD and the learned surrogate take vessel geometry and an inflow waveform as inputs. The study tests generalization across physiological conditions, remodeling, topology and mesh resolution. Figure 1 · Rygiel et al. (2025) · Rygiel et al. · CC BY 4.0

Interactive examples

Generalisation challenges

This project dates from early 2025. Our newer AB-GATr work provides more accurate wall shear stress and velocity estimates.

Generalization beyond a fixed vessel model

A useful surrogate must cope with the variation encountered in vascular modeling. This study uses geometric descriptors and projective geometric algebra to predict transient wall shear stress on abdominal aortic aneurysm surfaces.

The evaluation considers external cases, remodeled geometries, changing boundary conditions, new artery branches, and different mesh resolutions. These experiments test several distinct ways a model can encounter inputs that differ from its training data.

Cite this work

@misc{wallshearstressaaa2025,
  title = {Wall Shear Stress Estimation in Abdominal Aortic Aneurysms: Towards Generalisable Neural Surrogate Models},
  author = { Patryk Rygiel and Julian Suk and Christoph Brune and Kak Khee Yeung and Jelmer M. Wolterink },
  year = {2025},
  doi = {10.48550/arXiv.2507.22817},
  eprint = {2507.22817},
  archivePrefix = {arXiv},
  url = { https://doi.org/10.48550/arXiv.2507.22817 }
}