English

A Kernel Method for CT Reconstruction: a Fast Implementation using Circulant Matrices

Numerical Analysis 2014-12-19 v3

Abstract

Modern tomography involves gathering projection data from multiple directions and feeding them into a software algorithm for tomographic reconstruction. We focus our study on image reconstruction from Radon data in the setting of Computerized Tomography (CT), a non-invasive medical procedure that uses an X-rays equipment to produce cross-sectional images of the body. The detectors of the machine measure the X-ray projections through the sample, producing a so-called sinogram, which is actually the Radon transform of the attenuation coefficient. We provide a kernel-based reconstruction algorithm adapted to these kind of data, i.e. the Algebraic Reconstruction Technique (ART). The novel idea of this work is to reduce the complexity of the ART, through a faster implementation of the matrix-vector product, based on the storage of the involved matrix as a circulant matrix. We provide numerical results for both artificially generated (phantom data) and real data. The aim is to study the behaviour of the method with respect to its image reconstruction accuracy and its computational efficiency. In addition, we try to solve the trade-off between accuracy and efficiency and the one between accuracy and numerical stability.

Keywords

Cite

@article{arxiv.1403.0219,
  title  = {A Kernel Method for CT Reconstruction: a Fast Implementation using Circulant Matrices},
  author = {Maria Angela Narduzzo},
  journal= {arXiv preprint arXiv:1403.0219},
  year   = {2014}
}

Comments

Master's thesis, 108 pages This paper has been withdraw by the author due to some crucial errors

R2 v1 2026-06-22T03:18:36.097Z