The loss of detection sensitivity incurred by any stage of image processing may normally be characterized by the frequency dependence of the detective quantum efficiency (DQE) of that stage of processing, provided the image is represented in continuous coordinates. However, limitations to the DQE concept arise when discretely sampled projection data are used to obtain discretely sampled computed tomographic (CT) reconstructions. The source of these limitations is the aliasing produced by the discrete sampling which mixes contributions from various frequencies. An associated problem is that the SNR for the detection of an object can depend upon the position of the object relative to the discrete reconstruction pixels. The effective SNR for discrete images must take into account this variation, While there may be no loss in the detection SNR for reconstructions in continuous coordinates (DQE = 100%!), a reduction in the SNR will result from aliasing for discrete reconstructions. A simple one-dimensional model elucidates the characteristics of discrete CT reconstruction.
Keywords: detective quantum efficiency (DQE), noise equivalent quanta (NEQ), detectability, detection signal-to-noise ration (SNR), CT reconstruction, discrete sampling, aliasing
Get full paper (pdf, 328 KB)
Return to publication list
Send e-mail to author at kmh@hansonhub.com