| Single photon emission computed tomography (SPECT) has gained a lot of attention in medical imaging since it enables us to non-invasively visualize either anatomy information of human bodies or biological functions of human organs with the use of radiopharmaceuticals. The biological information it provides has become an important reference in clinical diagnosis of heart disease, brain diseases and cancers. However, due to a constrained radiation dose in clinical imaging, the amount of emitted photons that can be measured by the detector is quite limited, which results in noisy projection measurements.; The objective of this research is to investigate different types of measurements for iterative reconstructions with respect to noise, truncation and image representation effects.; Firstly, we propose a planar-integral system (e.g., a rotating-slat/slit system) which measures planar-integral projections, and compare it with a conventional line-integral system (e.g., a parallel-hole system) in terms of noise propagation and effects of insufficient number of measurements. The rotating-slat system is shown to advantage for small objects while the parallel-hole system outperforms for large objects. The criteria of measurement sufficiency are also given for the planar-integral system.; Secondly, we extend the planar-integral idea to cone-beam geometry, and propose a rotating single-slit (RSS) system and a rotating double-slits (RDS) system for small animal imaging. They are investigated and compared based on sensitivity uniformity, noise properties and spatial resolution.; Thirdly, we evaluate the image representation effects for iterative methods. We propose a new voxel-based iterative method with immediate after-backprojection filtering. It is shown that this new method significantly outperforms the general voxel-based methods in terms of image noise reduction, and achieves comparable resolution-to-noise performance to the blob-based iterative algorithms, but with much less computational cost.; Finally, we explore the capability of iterative reconstructions for ROI reconstruction with different types of truncated measurements. Computer simulation shows that iterative methods can apply to more truncation cases than the two existing analytical methods. A mathematical tool is presented to determine the recoverable image region using iterative methods, for a given truncation situation. |
