view article

Figure 6
Center of rotation (COR) demonstration on dose fractionated, 360°, simulated data. A simulated low-contrast 3D Shepp–Logan phantom was generated to which Poisson noise was applied with a cumulative fluence of 300000 photons per pixel. As this fluence is divided into Nθ = 360, 720, 1440, and 2880 projections (with [\bar{n}] = 833, 417, 208, and 104 photons per pixel per projection, respectively), one can see that individual projections become increasingly noisy in a subset of projections shown in a sinogram (a) and also in the single projection SNR (c). Since the same cumulative fluence was used, the θ-sums shown in (b) and the θ-sum SNR shown in (c) are the same no matter the number of projections Nθ used. [A sinogram with [\bar{n}] = 300000 photons per pixel per projection is also shown on the right in (a) and (b), showing no visible Poisson noise.] In spite of the very low SNR of individual projections (which scales as [{\bar{n}} ^{\,1/2}] as expected) shown in (c), the phase symmetry method accurately finds the center of rotation with low variance as shown in (d). The COR error bars in (d) represent the standard deviation in COR results from 100 different instances of Poisson noise at the indicated fluence per projection.

Journal logoJOURNAL OF
ISSN: 1600-5775
Follow J. Synchrotron Rad.
Sign up for e-alerts
Follow J. Synchrotron Rad. on Twitter
Follow us on facebook
Sign up for RSS feeds