IEEE Trans. Nucl. Sci. NS-25, pp. 160-163 (1978)

The characteristics of computed-tomographic reconstruction noise and their effect on detectability

K. M. Hanson
Los Alamos National Laboratory
D. P. Boyd
University of California, San Francisco


The process of reconstruction of a 2-D density distribution from 1-D projections has the effect of suppressing the low frequency portions of the noise power spectral density distribution. This low frequency suppression is demonstrated for computed tomographic (CT) reconstructions from an EMI 5005 scanner. An alternative method of displaying the implied long-range negative correlation in CT noise is through the noise granularity function. It is expected that large objects are easier to detect in the presence of computed tomographic noise than in the presence of white noise of the same rms magnitude. A simple detectability phantom is described which allows this hypothesis to be tested on x-ray CT scanners.

Keywords: CT noise, noise granularity, detectability, detectability phantom, noise power spectral density, EMI 5005 scanner

Get full paper (pdf, 2.8 MB)
Return to publication list
Send e-mail to author at