M.Sc. Tezi Görüntüleme
3D reconstruction of vessels is an important research area for human health and care. Thus there are lots of intensive studies in this open area.
In this thesis a 3D representation of coronary vessels is reconstructed from its three angiographic images taken at arbitrary views.
A vessel is modeled by generalized cylinders having elliptical cross-sections. To build this generalized cylinder one must have a medial axis and ellipses that can be placed on the medial axis orthogonally. To obtain the medial axis and ellipses the vessels edges and centerline information is extracted by fitting a parametric model of vessel projection. Marquardt-Levenberg nonlinear minimization algorithm is used to determine model parameters.
By using vessel centerline information from three images, 3D vessel medial axis is reconstructed with the assistance of projection parameters obtained in DICOM files that the images are stored. Once the medial axis is reconstructed, ellipse parameters is obtained from three boundary points in three images of the vessel and placed on the medial axis according to generalized cylinders theory.
Cubic smoothing spline is used to smooth and also interpolate medial axis and ellipse parameters to obtain continuity on the vessel.
Key Words:3D Reconstruction, Edge Estimation, Angiography, Generalized Cylinders, Non-linear Model Fitting, Cubic Smoothing Splines