Precision CT-based Aortic Valve Reconstruction: Minimal Variation Geometry Invariant Parametric Reconstruction Approach for Aortic Stenosis and Bicuspid Valves.

BACKGROUND AND

OBJECTIVES

Accurate reconstruction of the aortic valve complex is challenging due to its significant anatomical variability, presence of calcifications, and dynamic deformations across cardiac cycles. This study aimed to introduce and evaluate a novel, fully automated pipeline-Minimal Variation Geometry Invariant Parametric Reconstruction (MVGIPR)-for precisely reconstructing the aortic valve complex from computed tomography scans. By integrating computer-aided design, image processing, meshing, and geometry optimization into one cohesive framework, MVGIPR seeks to advance diagnostic and interventional planning in both aortic stenosis and bicuspid aortic valve cases.

METHODS

We retrospectively analyzed 80 patient datasets (covering aortic stenosis and bicuspid aortic valve pathologies), supplemented by ex-vivo micro-computed tomography and synthetic phantom experiments. The fully automated pipeline employed advanced computer-aided design for constructing parametric models, coupled with a semi-automated image processing workflow for segmenting and isolating the aortic valve complex. Further, a specialized meshing algorithm generated high-fidelity surface representations, while differential geometry metrics-specifically length, curvature, and torsion-were optimized to minimize noise and preserve essential anatomical details.

RESULTS

MVGIPR demonstrated strong concordance with high-resolution imaging and ground truth datasets. In clinical scans, minimal mean signed Euclidean distance errors of -0.2 ± 0.1 mm and -0.1 ± 0.08 mm were observed, whereas ex-vivo micro-computed tomography assessments yielded 233 ± 447 μm errors. Phantom data reconstructions showed negligible deviations (0.003 ± 0.04 mm) and a 2.9% surface area discrepancy, underscoring the pipeline's robustness in both planar and three-dimensional applications. Additionally, the automated approach reduced inter-observer variability by providing consistent, reproducible reconstructions across diverse imaging modalities.

CONCLUSIONS

MVGIPR offers a comprehensive, fully automated pipeline that integrates computer-aided design, image processing, meshing, and geometry optimization. By preserving critical anatomical features and minimizing noise, the method addresses key challenges in aortic valve complex reconstruction, thereby enhancing diagnostic and prognostic assessments. Its proven accuracy across clinical, ex-vivo, and synthetic settings underscores its potential for broader clinical integration, including enabling personalized treatment strategies and improving outcomes in valvular heart diseases by providing a robust foundation for advanced interventional planning.