Biomedical Engineering Reference
In-Depth Information
modalities are due to the vastly different nature of the data: anatomy and high
(
1 mm) resolution for MRI and CT compared with physiological function
and intermediate resolution (4 to 15 mm) for emission tomography.
Registration of SPECT or PET data with anatomical data alone (i.e., without
fiducial markers), often has the fundamental problem of trying to match dis-
similar data. Functional data from emission tomography may contain insuf-
ficient anatomical detail for many algorithms. It is therefore imperative that
systematic sources of error be minimized.
5.4.1.1
Geometric and Alignment Effects
5.4.1.1.1
Center of Rotation Errors
SPECT systems produce distortions in the reconstructed data if alignment of
the detector heads and electronic alignment of the center of the data acquisi-
tion matrix are not carefully measured and corrected prior to reconstruction.
The effect of this is to produce halo-like artifacts, which spatially distort the
true distribution. Similarly, nonuniformities in detector response lead to “ring
artifacts”: circular intensity distortions in the reconstructed image. The rea-
son for this is the use of a single (or very few) detector which is moved or
rotated to various positions about the subject to acquire all of the data. Thus
a nonuniform area will trace out a consistent, circular path at a set radius
from the center of the acquisition matrix.
5.4.1.1.2
Nonuniform Sampling Effects
The sampling in emission tomography systems is usually greater towards the
center of the field of view, and therefore any nonuniformity is greatly enhanced
the closer it is to the center of the field of view. It has been calculated that a 2%
nonuniformity at the center of the field of view can give an apparent 50% non-
uniformity in the reconstructed image. This falls off rapidly with radius.
62
An
example of nonuniformity “ring” artifacts is shown in Figure 5.3.
Full ring PET systems have a varying distance between adjacent parallel
projections, decreasing towards the edge of the field of view, due to its circu-
lar geometry. This effect would cause spatial distortions if ignored; however,
these are usually corrected prior to reconstruction. One effect that is not usu-
ally corrected by the reconstruction algorithm, though, is variation in the res-
olution of the system towards the edge of the field of view. The lines of
response towards the edge of the field are measured at an increasing angle
relative to the face of the crystals, thus decreasing the accuracy of localiza-
tion. This leads to an ellipsoidal point response function as this degradation
occurs preferentially for those lines of response at higher incidence angles to
the detectors.
5.4.1.1.3
Voxel Dimensions
An important issue for accurate coregistration is the accurate measurement
of the reconstructed voxel dimensions. In PET, for example, it is often assumed
that the detector radius, which determines the dimensions of the coordinate
