Quantification in 3D positron emission tomography
Acquisition and reconstruction of data in three-dimensional positron emission
tomography (3D PET) was introduced in 1990 almost 20 years after the first PET scanners
were developed. 3D PET offers a significant sensitivity improvement over conventional, sliceoriented
2D PET, but at the cost of a three-fold increase in acceptance of scattered events. In
addition, processing time is increased and new methods for applying corrections such as for
photon attenuation, calibration, and detector/geometry normalisation are required. 3D PET
raised concerns that the high quantitative accuracy that was possible with 2D PET (with its
moderate sensitivity) would not be matched in 3D, primarily because of the greatly increased
scattered photon component in the measured data.
The aim of this thesis was to develop methods that enable quantitatively accurate
measurements with 3D PET. A technique to correct for scattered photons prior to
reconstruction has been developed, implemented and assessed. A device for normalising the
data for detector efficiency and the geometry of the cylindrical detector system has been
developed, and the factors affecting reconstruction investigated. A new approach to calibration
of the reconstructed data to produce images of activity concentration which is independent of
scatter has been implemented. Finally, the techniques have been applied to data from brain
scans of human subjects.
Evaluation of images reconstructed from 3D PET demonstrates that the methodology
developed in this work produces data accurate to within 10% of the true activity concentration
in an object with reasonably homogeneous density. 3D PET is shown to be as accurate as 2D
PET, but with a sensitivity advantage that improves signal-to-noise by approximately a factor
of three in the human brain and slightly less in other regions of the body.