,

Abdominal aortic aneurysm (AAA) is a life-threatening dilation of the main abdominal artery that can rupture if left untreated. Currently, approximately 75% of AAAs in the Netherlands are treated by Endovascular Aneurysm Repair (EVAR), a minimally invasive procedure where a stent graft is placed via the groin. While EVAR offers better short-term outcomes than open surgery, it carries a significant long-term burden: 1 in 5 patients requires re-intervention within 5 years, and all patients undergo lifelong annual follow-up with imaging and hospital visits.

This project, funded by ZonMw, aims to develop and validate artificial intelligence models that personalize both EVAR treatment planning and follow-up protocols. We will create two complementary AI prediction models: (1) a pre-operative model that uses patient imaging and clinical data to predict postoperative complication risk, enabling surgeons to optimize treatment plans (e.g., select appropriate stent grafts, add preventive measures); and (2) a post-operative model that incorporates actual procedural data to stratify follow-up intensity — identifying high-risk patients needing frequent monitoring while allowing low-risk patients extended follow-up intervals up to 5 years.

A unique aspect of this project is the integration of these AI models into shared decision-making tools. We will evaluate three innovative methods for communicating personalized risk predictions to patients: conventional 3D visualization software, patient-specific 3D-printed physical models, and interactive holograms using smart glasses. Through active patient participation via our partnership with Harteraad (the Dutch cardiovascular patient association), we will determine the most effective approach for supporting informed, shared decisions about treatment and follow-up.

The project includes an early health technology assessment to evaluate cost-effectiveness and implementation requirements, ensuring sustainable integration into clinical practice. Ultimately, this research aims to improve quality of care for EVAR patients while reducing unnecessary healthcare burden and costs.

Highlights:

• Development of geometric deep learning models operating directly on 3D vascular mesh geometries
• First prospective clinical validation of AI-driven personalized EVAR follow-up protocols
• Novel patient communication tools combining 3D printing, holographic visualization, and clinical software
• Active patient participation through Harteraad partnership (~200 patient questionnaire + focus groups)
• Collaboration with Pie Medical Imaging for direct integration into widely used 3mensio Vascular software

 Related publications from the research group:

1. van Veldhuizen, W.A., de Vries, J.P.P., Tuinstra, A., Zuidema, R., IJpma, F.F., Wolterink, J.M., ... & van Sambeek, M.R. (2024). Machine learning based prediction of post-operative infrarenal endograft apposition for abdominal aortic aneurysms. European Journal of Vascular and Endovascular Surgery. 10.1016/j.ejvs.2024.07.003
2. Alblas, D., Hofman, M., Brune, C., Yeung, K.K., & Wolterink, J.M. (2023). Implicit neural representations for modeling of abdominal aortic aneurysm progression. International Conference on Functional Imaging and Modeling of the Heart (FIMH), 356-365. Springer Nature. 10.1007/978-3-031-35302-4_37
3. Rygiel, P., Alblas, D., Brune, C., Yeung, K.K., & Wolterink, J.M. (2024). Global Control for Local SO(3)-Equivariant Scale-Invariant Vessel Segmentation. arXiv preprint arXiv:2403.15314. 10.48550/arXiv.2403.15314
4. Suk, J., de Haan, P., Lippe, P., Brune, C., & Wolterink, J.M. (2024). Mesh neural networks for SE(3)-equivariant hemodynamics estimation on the artery wall. Computers in Biology and Medicine, 173, 108328. 10.1016/j.compbiomed.2024.108328
5. van Rijswijk, R.E., Groot Jebbink, E., Zeebregts, C.J., & Reijnen, M.M. (2022). A systematic review of anatomic predictors of abdominal aortic aneurysm remodeling after endovascular repair. Journal of Vascular Surgery, 75(5). 10.1016/j.jvs.2021.11.071