QUANTITATIVE RECONSTRUCTION FOR CONVERGING BEAM SPECT

The objective of the proposed research is to achieve quantitative
reconstruction with high computational efficiency for single photon
emission computed tomography (SPECT) using converging-hole collimators.
Converging-hole collimators, such as fan beam and cone beam collimators
with single or multiple focal points, provide better trade-off between
spatial resolution and detection efficiency, hence improved imaging
quality, as compared to parallel-hole collimators. They have found
useful applications in three-dimensional (3D) SPECT imaging of the brain
and heart in the clinics. However, to achieve quantitative
reconstruction for converging beam SPECT in routine clinical use, several
important problems need to be solved. First, the projection data
collected from a single planar orbit cone beam geometry are not
sufficient for exact reconstruction. Second, few truly 3D algorithms are
currently available for reconstruction, and they require very intensive
computation. The commonly used efficient algorithms make approximations
which introduce artifacts and inaccuracy in the reconstructed images.
Third, the design and use of multifocal converging beam collimators are
in the early stage, and a truly 3D algorithm for image reconstruction
remains to be developed. In addition to the special problems, physical
factors, such as photon attenuation, scatter, and spatially variant
detector response, degrade SPECT imaging quality. Efficient
compensations for these factors are difficult for converging beam SPECT.
To solve the above problems, we propose the following research aims.
First, for single-orbit cone beam geometry, the dimension and location of
the regions with missing data as well as the amount of missing data in a
particular part of the regions will be determined to improve estimation
of missing data. The results will also be used to optimize the design of
multi-orbit cone beam geometry to ensure complete sampling and to fully
utilize the measured data as well. Second, the fully 3D algorithm
previous developed by us will be extended to multi-orbit cone beam
reconstruction. Third, a truly 3D algorithm will be derived for
multifocal cone beam reconstruction. Finally, compensations for the
degrading factors will be performed. The Chang algorithm will be
extended to converging beam SPECT. To compensate for detector response,
a preprocessing filtering will be derived based on the frequency-distance
principle. Scatter compensation will be made using the newly developed
slab technique and dual-photopeak window technique. To evaluate the
proposed algorithms and techniques, a cone beam and a multifocal cone
beam collimators will be designed and constructed. Experiments using the
converging-hole collimators and Monte Carlo simulations will be carried
out, and then the reconstructed images will be compared to the phantoms
by means of various quantitation methods.