Biomedical Engineering Reference
In-Depth Information
34
35
(Klein-Nishina equation). In PET, Ollinger
used the recon-
structed emission and attenuation data to directly calculate the scatter that
would be expected given the object density and radiotracer distribution.
These methods have been shown to be highly accurate. All of these meth-
ods, of course, rely on spatially registered data sets. The dual-modality
PET-CT and SPECT-CT systems also can use the x-ray CT data in scatter
correction algorithms.
The attenuation data from simultaneous emission-transmission scans have
uses beyond providing attenuation and scatter correction factors. They have
been used in respiratory research in SPECT to define differences between
lung volume and distribution patterns of radioactively labeled aerosols.
and Watson
36-38
In the heart, Iida et al. used the reconstructed attenuation images to estimate
tissue bulk in an effort to correct for resolution limitations in PET myocardial
perfusion studies to produce a ''perfusable tissue index,'' a measure of the via-
ble tissue.
39
More sophisticated methods for partial volume correction are dis-
cussed in the next section.
11.4.2
Partial Volume Correction
Nuclear medicine images do not exhibit the same high spatial resolution that is
seen in x-ray CT or MRI. Generally, the main reasons for this are limited photon
flux (to minimize the radiation dose to the patient) and restrictions imposed by
the physical limitations in detecting high energy gamma radiation with scintil-
lation crystals. The typical spatial resolution in emission tomography varies
from 2 to 4 mm for high resolution PET scanners to approximately 12 to 18 mm
for SPECT systems. This leads to a characteristic known as the partial volume
effect. This “blurring,” due to limited spatial resolution, causes an object to
appear larger than it is if its true size is less than approximately three times
the system resolution.
40
While the
total
reconstructed counts within the object
are conserved, the count
is decreased from the true value because the
data are ''smeared'' over a larger area. A simple method to estimate the true
count density is to take the total counts in the “blurred” representation of
the object (organ, tumor, etc.) and normalize the count density to the actual
area of the object as measured by x-ray CT or MRI. More sophisticated
approaches use the anatomical data to perform a pixel-by-pixel correction of
the emission data, with
density
a priori
knowledge of the expected biodistribution
41
patterns.
An example is shown in Figure 11.9.
These procedures work, in general, in the following way. The structural
and functional data are first coregistered to the same space with a suitable
algorithm. Next, the structural scan is segmented into a number of dis-
crete, homogeneous compartments between which radiotracer uptake is
known to differ (e.g., differences between gray and white matter uptake in
the brain reflecting differences in glucose metabolism, blood flow, or receptor
density). A probability is then assigned to a photon arising from each of the dif-
ferent compartments (say 4:1:0 for gray matter:white matter:CSF, skull or
