An overview of computed tomography (1984)
Type of ContentTheses / Dissertations
Thesis DisciplineElectrical Engineering
Degree NameDoctor of Philosophy
PublisherUniversity of Canterbury. Electrical Engineering
AuthorsGarden, Kathryn L.show all
An overview of computed (or computerised. or computer-assisted) tomography (CT) is presented. A unified descriptive account of the underlying principles is given. The theory of CT image reconstruction is outlined. The applications of CT are reviewed, and several illustrative examples of practical CT systems are described in detail. Promising approaches to CT software and hardware problems (which often arise due to the departure of a system's real-world behaviour from the ideal concepts of theory) are identified. The implications of CT image reconstruction from incomplete projection data are explored. It is found that when images which are of "piecewise constant density" are reconstructed from few projections, the boundaries of constituent regions are often clearly recognisable. Based on the excellent ability of the human brain to extract structural information from degraded images an interactive reconstruction technique is developed. The interactive reconstruction scheme is found to be an effective and efficient way of incorporating a priori information, as well as constraining the image to be of an "acceptable form", based on the past perceptual experience of the human operator. The effects on reconstructed images of changes in the cross-section during scanning are investigated. It is found that significant variations in the densities of constituent regions (inside a cross-section) do not prevent the boundaries of those regions being clearly recognisable in the reconstruction. The implications of this finding for CT in general, and for X-ray CT imaging of the heart in particular, are discussed. The problem of spatially varying resolution in single photon emission CT (SPECT) is addressed. The effect of measuring rays within a "solid angle" (a diverging cone beam, of Gaussian cross-section) is analysed, and a computational procedural framework for compensating for the above effect is presented and illustrated.