Related Experiment Video
Updated: Jan 9, 2026
09:32
Manufacturing Abdominal Aorta Hydrogel Tissue-Mimicking Phantoms for Ultrasound Elastography Validation
Published on: September 19, 2018
15.8K
Geometric deep learning for local growth prediction on abdominal aortic aneurysm surfaces
Dieuwertje Alblas1, Patryk Rygiel1, Julian Suk1
1Department of Applied Mathematics, Technical Medical Centre, University of Twente, Drienerlolaan 5, Enschede, 7522 NB, The Netherlands.
Artificial Intelligence in Medicine
|December 2, 2025
Summary
This study introduces a new AI model to predict abdominal aortic aneurysm (AAA) growth using 3D vascular models. The model offers personalized monitoring, improving patient surveillance and predicting repair eligibility.
Area of Science:
- Biomedical engineering
- Artificial intelligence in medicine
- Cardiovascular imaging analysis
Background:
- Abdominal aortic aneurysms (AAAs) are life-threatening dilatations of the aorta.
- Current AAA monitoring relies on maximum diameter, which may not accurately predict growth.
- Personalized growth prediction is needed to optimize surveillance strategies.
Purpose of the Study:
- To develop and validate an SE(3)-symmetric transformer model for predicting AAA growth.
- To assess the model's ability to personalize AAA surveillance and predict surgical repair eligibility.
Main Methods:
- Utilized a longitudinal dataset of 113 computed tomography angiography (CTA) scans from 24 AAA patients.
- Employed an SE(3)-symmetric transformer model operating directly on the vascular surface with local, multi-physical features.
- Validated the model on an external dataset of 25 CTAs from 7 AAA patients.
Main Results:
- The model achieved a median diameter error of 1.18 mm in predicting AAA growth to the next scan.
- Demonstrated high accuracy (0.93) in identifying patients eligible for elective repair within two years.
- Confirmed the feasibility of local directional AAA growth prediction from vascular surfaces.
Conclusions:
- AI-driven, surface-based AAA growth prediction is feasible and accurate.
- This approach can enhance personalized surveillance strategies for AAA patients.
- The model shows potential for improving clinical decision-making in AAA management.
