Aortic Aneurysm, Bicuspid Aortic Valve (BAV), TAVR
Conditions
Keywords
Bicuspid Aortic Valve, Aortic Dilation, Vascular Deformation Mapping, Risk Prediction
Brief summary
Bicuspid aortic valve (BAV) is the most common congenital valvular malformation, characterized by heterogeneous phenotypic subtypes that predispose patients to secondary aortic pathologies, including valvular dysfunction and ascending aortic dilation. With approximately 50% of BAV patients developing aortic dilation, a prevalence that continues to rise, accurate assessment of postoperative aortic remodeling remains a critical unmet clinical need for early risk stratification and optimized therapeutic decision-making. Currently, clinical surveillance relies heavily on periodic manual measurement of the maximum aortic diameter on follow-up computed tomography angiography (CTA), yet this approach suffers from several inherent limitations. It is a lagging indicator that detects irreversible wall damage only after significant enlargement has occurred. It oversimplifies complex three-dimensional morphological changes into a single linear dimension. It exhibits substantial intra- and inter-observer variability. It is also inefficient for large-scale longitudinal data management. Although alternative metrics such as computational fluid dynamics (CFD) derived hemodynamic parameters and morphological geometric features have been explored, existing methods remain constrained by static single-time-point analyses that fail to capture the dynamic biomechanical evolution driving aneurysm progression, high technical barriers that preclude routine clinical integration, and a lack of comprehensive models that systematically integrate dynamic deformation, static anatomy, and hemodynamic information. To address these gaps, this study aims to develop a fully automated, quantitative, and dynamic risk prediction system that leverages vascular deformation mapping (VDM) for noninvasive early detection of regional aortic deformation, integrates multiparameter features including dynamic deformational, static anatomical, and hemodynamic characteristics through an artificial intelligence model, and delivers intuitive structured reports to directly support clinical decision-making, thereby enabling earlier intervention and improved patient outcomes.
Detailed description
This retrospective multi-center study aims to develop an automated AI system integrating Vascular Deformation Mapping (VDM) and Computational Fluid Dynamics (CFD) to predict aortic dilation risk in BAV patients after TAVR. Approximately 1,000 patients with pre-operative, post-operative, and follow-up CTA will be enrolled from two Chinese hospitals. The system automatically segments the aorta using 3D U-Net++, quantifies local deformation via deformable registration, and extracts dynamic, anatomical, and hemodynamic features. An XGBoost model trained on historical data (n=200) with 1-year outcomes outputs risk probability and category. The primary outcome is a validated prediction model; secondary outcomes include deformation pattern quantification and automated report generation. All data are anonymized. The target sample size is 1,000 patients.
Interventions
PROCEDURETranscatheter Aortic Valve Replacement
Transcatheter Aortic Valve Replacement (TAVR) is a minimally invasive procedure in which a collapsible replacement valve is inserted via catheter through the femoral artery or other access routes and deployed within the native diseased aortic valve. In this study, TAVR was performed as standard clinical care in BAV patients with severe aortic stenosis or regurgitation. Post-procedural CTA imaging was obtained as part of routine follow-up to monitor aortic remodeling and detect potential dilation. The present study retrospectively analyzes the serial CTA images acquired before and after this procedure; no additional intervention is administered for research purposes.
Sponsors
Second Affiliated Hospital, Zhejiang University, School of Medicine
First Affiliated Hospital of Wenzhou Medical University
Study design
Observational model
COHORT
Time perspective
RETROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
18 Years to 85 Years
Healthy volunteers
No
Inclusion criteria
Patients are eligible for inclusion if they are male or non-pregnant female aged 18 to 85 years, diagnosed with Bicuspid Aortic Valve (BAV) by CTA, and have successfully undergone Transcatheter Aortic Valve Replacement (TAVR) with available pre-operative, post-operative (1-3 months), and follow-up (6-12 months) CTA imaging. The CTA scan range must cover cranially at least the origin of the brachiocephalic trunk, left common carotid artery, and left subclavian artery, and caudally at least the origin of the internal and external iliac arteries.
Exclusion criteria
Patients are excluded if they have traumatic or iatrogenic aortic dissection, isolated aortic aneurysm without BAV, connective tissue disorders such as Marfan syndrome or Ehlers-Danlos syndrome, or prior cardiac or aortic surgery including surgical aortic valve replacement or aortic repair. Additional exclusions include insufficient number of CTA scans or inadequate scan range, poor CTA image quality due to motion artifacts or inadequate contrast enhancement precluding accurate segmentation, or image registration failure precluding completion of VDM analysis.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Development and Validation of a Multi-dimensional Risk Prediction Model for Aortic Dilation | Post-TAVR 1-year follow-up | The model is developed using machine learning (XGBoost) integrating dynamic deformation features (e.g., radial displacement percentiles from VDM), static anatomical features (e.g., aneurysm volume), and optional hemodynamic features (e.g., wall shear stress from CFD). Model performance (discrimination and calibration) will be assessed using the Area Under the Receiver Operating Characteristic Curve (AUC) and calibration plots. The outcome is the model's predictive accuracy for aortic dilation status. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Quantification of Aortic Deformation via Vascular Deformation Mapping (VDM) | Post-TAVR 3-month follow-up | Aortic deformation is quantified by calculating the 3D displacement field between pre- and post-TAVR CTA images using a highly regularized deformable registration algorithm. The primary metric is the 90th percentile of radial displacement (mm) of the aortic surface. Higher values indicate greater local expansion. |
| Change in Aortic Dimensions Measured by Automated 3D Analysis | Post-TAVR 1-year follow-up (relative to pre-TAVR baseline) | Change in maximum aortic diameter, cross-sectional area, and volume, automatically measured from 3D aortic models segmented by a deep learning algorithm. This provides a comprehensive assessment of morphological change beyond the traditional single-diameter measurement. |
| Clinical Utility Assessment of the Automated Reporting System | Upon study completion, up to 36 months | The clinical utility of the automated system is evaluated by the proportion of successfully generated, structured clinical reports that are interpretable without manual correction. This outcome measures the feasibility of integrating this advanced AI-driven analysis into routine clinical workflow. |
Countries
China
Outcome results
None listed