Fast and efficient tetrahedral volume mesh reconstruction with CAD-ASTRA

Reconstructing x-ray CT data on cubic voxels leads to discretization errors due to the finite size of these voxels. These errors are known as Partial Volume Effects (PVE). Reconstructing x-ray data on a tetrahedral mesh adapted to the sample generates a memory efficient multi-resolution representation that has the potential to diminish PVE. Such an adaptive method requires reconstructing attenuation values of the scanned sample on tetrahedra of the mesh. While a wide variety of open source software is available for iterative reconstruction methods on cubic voxels, research on fast iterative reconstructions on tetrahedral meshes is limited. In this work, we present an extension of the GPU accelerated CAD-ASTRA mesh projector that solves the inverse problem on tetrahedra. With an appropriate starting condition, such a representation has the potential to greatly increase the accuracy of a CT reconstruction.