Improved Image Quality for Asymmetric Double-focal Cone-beam SPECT

Z. J. Cao, B. M.W. Tsui

Research output: Contribution to journalArticlepeer-review

1 Scopus citations

Abstract

To optimize both spatial resolution and detection efficiency in brain SPECT imaging using a rectangular camera, an asymmetric double-focal cone-beam collimator is proposed with the focal points located near the base plane of the patient's head. To fit the entire head into the field-of-view of the collimator with dimensions of 50cmx40cm and at a radius-of-rotation of 15 cm, the focal lengths of the collimator are 55 and 70 cm, respectively, in the transverse and axial directions. With this geometry, the artifacts in the reconstructed image produced by the Feldkamp algorithm are more severe compared to those in a symmetric cone-beam geometry, due to the larger vertex angle between the top of the head and the base plane. To improve the reconstructed image quality, a fully three-dimensional (3D) reconstruction algorithm developed previously for single-focal cone-beam SPECT was extended to the asymmetric double-focal conebeam geometry. The algorithm involves nonstationary 2D filtering and a reprojection technique for estimation of the missing data caused by a single-orbit cone-beam geometry. The results from simulation studies with the 3D Defrise slab phantom demonstrated that the fully 3D algorithm provided a much improved image quality in terms of reduced slice-to-slice cross talks and shape elongation compared to that produced by the conventional Feldkamp algorithm.

Original languageEnglish (US)
Pages (from-to)1145-1148
Number of pages4
JournalIEEE Transactions on Nuclear Science
Volume40
Issue number4
DOIs
StatePublished - Aug 1993
Externally publishedYes

ASJC Scopus subject areas

  • Nuclear and High Energy Physics
  • Nuclear Energy and Engineering
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Improved Image Quality for Asymmetric Double-focal Cone-beam SPECT'. Together they form a unique fingerprint.

Cite this