Image Processing Techniques in Emission Tomography - Correction Techniques in Emission Tomography - page 138

TABLE 7.2: Software

Software Source Language Platform GUI Examples

ITK yes C++ Linux, PC, MAC no yes

FAIR yes MATLAB Linux, PC, MAC no yes

FSL yes C++ Linux, PC, MAC yes yes

AIR yes C Linux, PC, MAC yes no

ART no C++ Linux, PC, MAC no no

the PVE lie in the limited resolution of the respective imaging devices. Struc-
tures smaller than the system's resolution cannot be resolved which results in
blurred boundaries for example. If the size of the structures to be examined
is close to the imaging system's resolution, e.g., in imaging of small tumors,
the impact of the PVE cannot be neglected.
This section deals with the mechanisms of the PVE and its impact on
PET imaging (Section 7.5.1) as well as correction methods reducing the image
degradation caused by PVE (Section 7.5.2).

7.5.1 The partial volume effect in PET imaging
Compared to morphological scanning techniques, one of the most decisive
properties of PET is the quantification of tracer uptake in vivo. Tracer up-
take can be quantified by means of the standardized uptake value (SUV), a
normalized index taking different aspects into account, e.g., tracer dose, body
weight and radioactive decay. The SUV allows a more precise diagnosis in tu-
mor imaging and gives reliable information, e.g., about the progress of tumor
therapies. The partial volume effect, however, makes it dicult to take full
advantage of the potential of PET and has a strong impact on the accuracy of
tracer concentration measurements like the SUV, particularly in small struc-
tures. The extent of this effect depends strongly on the resolution of the PET
scanner.
In the following we show that partial volume correction is necessary to
improve the quality of PET images and hence the quantification of tracer
uptake. The influence of the PVE can be described by the image degradation
model (see Equation (7.1))

I = P ? I u + N : (7.24)

Here, I is the PET image after reconstruction, I u is the clean and uncorrupted
image of the scanned object and N is additive noise. P denotes the point spread
function which is, in general, the imaging system's response to a point source,
i.e., how the system depicts an object smaller than the system's resolution.
The signal acquired by scanning a point source resembles a Gaussian function
(a Gaussian function is thus often used in practice to represent the PSF). The
width of that signal at half its highest value is the full width at half maximum
(FWHM), describing the system's resolution. Current human PET imaging