Spatial variation of SNR in two- and three-dimensional neuro-PET

Author

Li, Henry H. ; Votaw, John R.

Author_Institution

Dept. of Radiol., Emory Univ., Atlanta, GA, USA

Volume

44

Issue

2

fYear

1997

fDate

4/1/1997 12:00:00 AM

Firstpage

225

Lastpage

229

Abstract

A method for region of interest (ROI) evaluation for three-dimensional (3-D) positron emission tomography (PET) in the sinogram space was implemented, according to the fully 3-D filtered back-projection algorithm. With this method, the statistical error in the image that propagates from the Poisson noise in the raw data was computed. The signal-to-noise ratio (SNR) for ROIs at various locations inside a cylindrical phantom was computed from both scanner data and simulation data and was verified via the standard deviation method through multiple measurements. As a comparison, two-dimensional (2-D) scans were also collected and similar computations carried out, Results show that the SNR increases with radius due to decreased attenuation at the edge of the phantom. For 3-D scans, the SNR drops gradually for ROIs outside the central 8 cm of the field of view (FOV). Also, it was found that the random events must be recorded and considered in the error computation

Keywords

brain; error statistics; filtering theory; image reconstruction; medical image processing; neurophysiology; positron emission tomography; random noise; Poisson noise; SNR; cylindrical phantom; error computation; field of view; fully 3-D filtered back-projection algorithm; multiple measurements; positron emission tomography; random events; scanner data; signal-to-noise ratio; simulation data; sinogram space; spatial variation; standard deviation method; statistical error; three-dimensional neuro-PET; two-dimensional neuro-PET; Attenuation; Computational modeling; Data mining; Filtering; Image reconstruction; Imaging phantoms; Measurement standards; Positron emission tomography; Signal to noise ratio; Two dimensional displays;

fLanguage

English

Journal_Title

Nuclear Science, IEEE Transactions on

Publisher

ieee

ISSN

0018-9499

Type

jour

DOI

10.1109/23.568811

Filename

568811