Related Experiment Video
Updated: Jul 28, 2025
09:32
Manufacturing Abdominal Aorta Hydrogel Tissue-Mimicking Phantoms for Ultrasound Elastography Validation
Published on: September 19, 2018
15.4K
Computer-aided shape features extraction and regression models for predicting the ascending aortic aneurysm growth
Leonardo Geronzi1, Antonio Martinez1, Michel Rochette2
1University of Rome Tor Vergata, Department of Enterprise Engineering "Mario Lucertini", Rome, Italy; Ansys France, Villeurbanne, France.
Computers in Biology and Medicine
|June 1, 2023
Summary
Predicting ascending aortic aneurysm growth is challenging. Global shape features, particularly those derived from Partial Least Squares (PLS), show significant promise in improving growth prediction accuracy compared to local features.
Area of Science:
- Cardiovascular Imaging
- Medical Image Analysis
- Computational Anatomy
Background:
- Ascending aortic aneurysm (AAA) growth prediction remains a clinical challenge.
- Accurate prediction is crucial for timely intervention and patient management.
- Current methods may not fully capture the complex biomechanical factors influencing aneurysm expansion.
Purpose of the Study:
- To evaluate and compare the efficacy of local and global shape features in predicting ascending aortic aneurysm growth.
- To identify specific shape characteristics associated with faster aneurysm expansion.
- To develop and assess regression models for AAA growth prediction.
Main Methods:
- Utilized 3D imaging data from 70 patients with ascending aortic aneurysms.
- Computed local shape features (diameter-to-length ratio, wall distensibility, tortuosity) and global shape features (via Principal Component Analysis and Partial Least Squares).
- Developed Gaussian Support Vector Machine and PLS linear regression models for growth prediction using leave-one-out cross-validation.
Main Results:
- Partial Least Squares (PLS)-derived global shape features achieved the lowest root mean square error (0.066 mm/month) in growth prediction.
- Principal Component Analysis (PCA)-based global features yielded a prediction error of 0.083 mm/month.
- Local shape features resulted in a prediction error of 0.112 mm/month.
- Aneurysms near the aortic root with larger initial diameters exhibited faster growth rates.
Conclusions:
- Global shape features offer a significant advantage in predicting ascending aortic aneurysm growth.
- PLS-based shape modes demonstrate superior predictive performance.
- Further research into advanced shape analysis techniques is warranted for improved clinical risk stratification.
