ORIGINAL RESEARCH article
Front. Physiol.
Sec. Computational Physiology and Medicine
Predicting the growth of asymptomatic small abdominal aortic aneurysms (AAA) based on deep learning
Provisionally accepted
- 1Northeastern University College of Medicine and Biological Information Engineering, Shenyang, China
- 2National Key Laboratory of Frigid Zone Cardiovascular Disease, Cardiovascular Research Institute and Department of Cardiology, General Hospital of Northern Theatre Command, Shenyang, China
- 3Department of Radiology, General Hospital of Northern Theatre Command, Shenyang, China
- 4Department of Radiology, Northern Theater Command General Hospital, Shenyang, China
- 5State Key Laboratory of Robotics, Chinese Academy of Sciences Shenyang Institute of Automation, Shenyang, China
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
Accurate prediction of asymptomatic small abdominal aortic aneurysm (AAA) growth is crucial for risk stratification and personalized surveillance. This study developed an end-to-end deep learning framework to predict rapid expansion (≥0.5 cm/6 months) using computed tomography angiography (CTA) images from 81 asymptomatic patients with small AAA (31 rapid growth and 50 stable patients). The pipeline integrated three core components: a ResNet50 classifier for identifying aortic images (99.86% accuracy, 99.91% F1-score), a YOLOv11 detector for localizing aneurysms (precision-recall: 0.902), and a MedMamba-based feature fusion model that combined imaging features with clinical metadata via multi-head self-attention. Model robustness was ensured through stratified 5-fold cross-validation and comprehensive data augmentation. The fusion model achieved a predictive accuracy of 98.75% and an F1-score of 97.78, outperforming seven classical deep learning backbones. Furthermore, explainability analyses confirmed the model's reliance on established clinical risk factors and highlighted biologically plausible imaging regions for prediction. The proposed ResNet50-YOLOv11-MedMamba framework demonstrates the feasibility of automating AAA growth prediction directly from CTA, showing promising potential to enhance clinical decision-making.
Keywords: multi-head self-attention, computed tomography angiography, Growth prediction, deep learning, Abdominal Aortic Aneurysm
Received: 13 Sep 2025; Accepted: 03 Dec 2025.
Copyright: © 2025 Cheng, Zhang, Wang, Sun, Wang, Wang and Wang. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Xiaozeng Wang
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.