Biomedical Engineering Reference
In-Depth Information
FIGURE 11.9
Partial volume effect correction in an iterative reconstruction algorithm. The images show
A) segmented MRI scan, B) extracted gray matter compartment, C) FDG PET scan, D) partial
volume corrected FDG PET scan, and E) the partial volume corrected PET scan superim-
posed on the structural MRI scan. (Images courtsey of Dr Claire Labbé, University Hospital
of Geneva and CERMEP.)
outside the brain in this example), and the relative “spill-in” and “spill-out” for
each compartment calculated. This process may be performed iteratively using
updated estimates of the radiotracer distribution at each step. There are many
variations on this general approach (see, for example, Rousset et al.
42
).
11.4.3
Anatomically Guided Reconstruction
New developments in reconstruction algorithms, which apply partial vol-
ume correction in an earlier step in the processing sequence, incorporate
knowledge of the object's boundaries into the reconstruction in an attempt
to improve upon the limited spatial resolution by restoration of the emis-
sion data during reconstruction.
43-45
Much prior information can be built into
iterative reconstruction schema, such as photon attenuation, scattering, and
an intrinsic correction for the blurring or limited spatial resolution of the
measuring systems.
46
The underlying idea behind so called ''anatomically
guided reconstruction'' is that the functional image is strongly correlated
with the anatomical distribution of known sites of radiopharmaceutical
uptake, e.g., receptors, tumors, boundaries of organs, the cerebral gray
white
matter interface, etc. These algorithms produce a probabilistic estimate of the
likely distribution which has given rise to the two-dimensional projection
data measured, test this against the actual measured projection data, modify
the estimated reconstruction in some deterministic way, and then again test
