computation process, and, in general, the computation time decreases with
the decreasing number of subsets (i.e., more projections in each subset).
Corrections for detection efficiency variations, noise component, random
coincidences, scatter coincidences, and photon attenuation are made prior
to reconstruction in the FBP method. In the MLEM or OSEM method,
these factors are inherently incorporated a priori in the estimated image
and need not be applied separately. In general, iterative reconstruction
methods do not produce artifacts observed with the FBP method and
provide a better signal-to-noise ratio in regions of low tracer uptake (Fig.
12.14). Overall, iterative methods provide high-quality images and are cur-
rently included in image reconstruction in PET and SPECT.
Another algorithm, the row-action maximum likelihood algorithm
(RAMLA), has been proposed as a special case of OSEM requiring
sequences of orthogonal projections, which lead to faster convergence than
the OSEM itself.
SPECT/CT
Accurate medical diagnosis of human disease can be made if both anatom-
ical and functional status of the patient’s disease is known. In the inter-
pretation of nuclear medicine studies, physicians always like to have a
comparison between high-resolution CT or MR images and low-resolution
PET or SPECT images for accurate localization of lesions. In PET and
SPECT imaging, in vivo measurement of organ physiology, cellular metab-
olism, and perfusion and other functional status of the organ is made.
However, these studies have poor resolution due to poor photon flux and
lack anatomical detail. On the other hand, computed tomography (CT) or
magnetic resonance (MR) imaging provide excellent spatial resolution with
high anatomical detail, but little functional information.
Single Photon Emission Computed Tomography 169Fig. 12.14. Comparison of filtered backprojection and iterative OSEM method with
attenuation correction.