The sinogram polygonizer for reconstructing 3D shapes.

Area of Science:

• Medical Imaging
• Computer Graphics
• Computational Geometry

Background:

• Conventional 3D shape reconstruction from X-ray computed tomography (CT) involves volume reconstruction followed by isosurface polygonization (e.g., Marching Cubes).
• This two-step process can introduce artifacts and may not efficiently handle the continuous nature of CT data.

Purpose of the Study:

• To propose a novel method, the sinogram polygonizer, for direct 3D shape reconstruction from sinograms.
• To generate accurate, high-quality triangle meshes directly from CT projection data, bypassing intermediate volume reconstruction.

Main Methods:

• The sinogram polygonizer treats CT values as a continuous function.
• It directly extracts a triangle mesh using tetrahedral mesh deformation, incorporating quadratic error metric minimization and optimal Delaunay triangulation.
• Analytical gradient estimation of CT values is employed.

Main Results:

• The method directly reconstructs 3D shapes from sinograms, producing triangle meshes.
• Sharp features are well approximated due to analytical gradient estimation, even with coarse meshes.
• Aliasing artifacts commonly seen on triangle meshes are eliminated.

Conclusions:

• The sinogram polygonizer offers a direct and efficient approach for 3D shape reconstruction from CT data.
• This method yields high-quality, artifact-free triangle meshes with accurate feature representation.
• It presents a significant advancement over conventional two-step reconstruction techniques.