Corrections for Scanner-Related Factors - Correction Techniques in Emission Tomography - page 111

the measured data, and the FWHM and FWTM values of the spatial resolu-
tion should be determined from a linear interpolation between the neighboring
measured data points. The alternative method of line of point transfer func-
tions is hardly used anymore. Data should be obtained for various locations
in the FOV of the PET scanner.
As a practical issue many preclinical scanners do not provide an FBP re-
construction algorithm. This has led to a highly varying body of literature on
the actually obtainable spatial resolution for a given scanner by applying all
kinds of iterative reconstruction algorithms. It is important to note that itera-
tively reconstructed point source measurements against a cold background do
depend strongly on the number of iterations or other tweaking factors. At least
a non-zero background concentration should be present in the measurement
of a spatial resolution based on an iterative reconstruction technique.
The nite spatial resolution of PET scanners (clinically 6 mm, preclini-
cally 2 mm) leads to another eect that needs to be considered if quantitative
data analysis is to follow the reconstruction of PET images. Radioactivity in
structures with dimensions smaller than 2 FWHM spatial resolution of the
scanner will be affected by the partial volume effect (PVE). In these structures
the measured radioactivity will be spread out over a volume roughly equal to
a cylinder with a diameter of 2 FWHM in the in-plane direction and a
length of 2 FWHM in the axial direction. This leads to an underestimation
of the local radioactivity within the cylinder. Furthermore, the background
activity surrounding the structure will influence the measured radioactivity
concentration (via spill in for a structure in a background with a higher ra-
dioactivity concentration, or via spill out for a structure in a background with
a lower radioactivity concentration). As an example, quantification of glucose
metabolism in the myocardium is hindered by the high concentration of tracer
in the blood pool at the start of the measurement (this activity spills into the
myocardium). At the end of the measurement, the activity in the myocardium
spills into the blood pool, which at that point in time will have a low remain-
ing activity concentration. Phantom experiments with spheres of various sizes
allow for the calculation of recovery coecients that allow for compensation
of these effects in spherical objects. But in the case when the tumor shape is
not known, they are of limited usefulness. In a recent paper, various strategies
to correct for the PVE were evaluated [9]. Corrections can be applied at var-
ious stages in the data analysis process: during reconstruction (see Chapter
3), during kinetic modeling or on the reconstructed images themselves. PVC
improves accuracy of SUV without decreasing (clinical) test-retest variability
significantly and it has a small, but significant effect on observed tumor re-
sponses. Reconstruction-based PVC outperforms image-based methods, but
requires dedicated reconstruction software. Image-based methods are good
alternatives because of their ease of implementation and their similar perfor-
mance in clinical studies. For more detailed information on the methods that
were evaluated one could have a look at the references of this chapter as well
as in [16]. Furthermore PVE correction approaches have developed in differ-