Corrections for Physical Factors - Correction Techniques in Emission Tomography - page 79

measurement of the proper attenuation in every scan session. However, this
usually comes at the cost of long transmission scan times to ensure a satisfying
statistical quality in the transmission data as poor transmission data will lead
to an excessive amount of noise in the attenuation-corrected emission data
and thus to inferior image quality.
A first step in solving this problem is a spatial smoothing of the blank
and transmission data [74]. More advanced methods involve the utilization
of a priori knowledge of attenuation values for dierent tissue values [65]
[87] [88]. For this approach, the noisy attenuation map is first calculated by
reconstructing

Z

log b i

t i =

(r)dr; (5.16)

i

obtaining an image of the attenuation map (r). Then, (r) is segmented using
histogram techniques into a limited number of body tissues (soft tissue, lung
tissue, bones) whose average values are known (see Table 5.1), resulting in a
noise-free attenuation coecient map from which the attenuation correction
values can be determined by calculating the line integrals along each line
of response. Furthermore, to account for individual variations in -values, a
weighted sum of both the segmented and the original attenuation map can
easily be generated (see Figure 5.9).
Other strategies to avoid long transmission scan times involve operating
the blank and transmission scans in singles rather than in coincidence mode
as the singles counting rate is much higher than the coincidence counting rate
[26]. Obviously, this singles mode can also be applied when using gamma de-
caying sources like 137 Cs (E = 661:7 keV) instead of positron emitters. This
reduces the cost of replacing the sources after a short time, as the half-life of
137 Cs is 30 years, ensuring a long system life time [42]. However, if the gamma
energy is significantly different from 511 keV, then the measured attenuation
data has to be scaled to PET energies, which again can be done in a seg-
mentation approach similar to the one described above [11]. Additionally, the
singles-based transmission data is heavily scatter-contaminated as scattered
singles events cannot be distinguished from non-scattered events as opposed

FIGURE 5.9: Attenuation maps derived without segmentation (left) and
with segmentation (right).