Biomedical Engineering Reference
In-Depth Information
6.1.5 Partial volume
Partial volume corrections are necessary due to the discrete nature with
which the true distribution of radioactivity is mapped into an image. Dis-
cretization of the imaged volume into voxels that may be large with respect
to the spatial variation in the radioactivity distribution leads to an averaging
of this variation on the scale of the voxel dimension. Another length scale is
provided by the spatial resolution of the PET scanner. This macro parameter
depends on a number of characteristics:
•
geometrical arrangement of the detectors (e.g., the axial dimension of
a crystal element sets a lower limit to finest axial spacing in sinogram
space). This arrangement also introduces the typical spatial variation of
spatial resolution over the FOV of the scanner.
•
the physics of positron annihilation; due to annihilation photon non-
collinearity, a larger scanner bore will lead to a decreased spatial reso-
lution of the scanner.
•
the possibility to determine depth of interaction within a crystal; by
using a binary depth of interaction scheme a recorded event can be
attributed to either the surface or the back part of a crystal element
(typically 1 cm long). Depth of interaction can be measured by defining
a physically separated front and back layer for each original crystal (by
combining two different scintillators, or having a separate readout for
two stacked crystals that both are half the length of an original crystal),
double readout of a single crystal (front and back side) and so on. This
effectively reduces the distance between lines of responses, enabling a
finer sinogram sampling.
•choices for span, ring difference and radial and angular mashing factors;
increasing those factors generally leads to increased spatial resolution of
the scanner, but they may be necessitated by count statistical consider-
ations.
•
the reconstruction algorithm.
The NEMA performance evaluation protocol [2] describes a method to
determine spatial resolution of a PET camera. The source should be smaller
than 4 times the expected FWHM spatial resolution of the scanner. The ac-
tivity should be low enough to make contributions of randoms and scatters
to the measured data negligible. Reconstruction should be done via filtered
back projection (a non-iterative reconstruction method) on a voxel grid that
provides at least 5 pixels on the FWHM. Profiles through the maximum of
the measured radioactivity distribution should be summed over two FWHM
in the two directions perpendicular to the direction in which the FWHM is
measured. The maximum activity should be determined from a parabolic fit to
Search WWH ::
Custom Search