The mitral valve computational anatomy and geometry analysis

Date

2015-12-10

Authors

Khalighi, Amir Hossein

Journal Title

Journal ISSN

Volume Title

Publisher

Abstract

We present a novel methodology to characterize and quantify the Mitral Valve (MV) geometry and physical attributes in a multi-resolution framework. A multi-scale decomposition was implemented to model the MV geometry by using superquadric shape primitives and spectral reconstruction of the finer-scale geometric details. Superquadrics provide a basis to normalize the size and approximate a basic model of the MV geometry. The point-wise difference between the original geometry and the superquadric model denotes the finer-scale geometric details, which can be modeled as a scalar attribute for the MV model development. The additive decomposition of the basic MV geometry from geometric details (attributes) allows recovering the actual geometry by superposition of the superquadric approximation and the finer-details model. We implemented a lasso optimization algorithm to perform spectral analysis and develop the Fourier reconstruction of the geometric details. The spectral modeling enabled us to resample the geometric details or use spectral filters in order to adjust the spatial resolution in the model reconstruction. It also provides the basis to control the level of detail in the final model reconstruction by applying low-pass filters in the frequency domain. The higher-order attributes such as internal fiber architecture can be integrated with the geometric models using the same framework. We applied our pipeline to create models of three ovine MVs based on computed-tomography 3D images with micrometer resolution. We were able to quantify the MV leaflet geometry, reconstruct models with custom level of geometric details, and develop medial representation of the MV leaflet structure. The results show that our methodology for geometry analysis provides a basis for assessing patient-specific geometries and facilitates developing population-averaged models. Ultimately, this approach allows building personalized image-based computational models for medical device design and surgical treatment simulations.

Description

LCSH Subject Headings

Citation