Region-of-interest image reconstruction in computed tomography
| dc.contributor.advisor | Hussein, Esam | |
| dc.contributor.author | Enjil Ela, Esmaeil | |
| dc.date.accessioned | 2023-03-01T16:41:13Z | |
| dc.date.available | 2023-03-01T16:41:13Z | |
| dc.date.issued | 2012 | |
| dc.date.updated | 2020-06-19T00:00:00Z | |
| dc.description.abstract | Computed tomography (CT) produces slice (section) images using multiple exposures to x-ray beams. It is often desired to enhance the image of a smaller interior region-of-interest (RoI). Magnifying the image may visually improve the appearance of the RoI, but it does not fundamentally alter the image attributes. The current practice is to acquire another CT image with finer resolution, resulting in additional cost and radiation exposure. The objective of this work is to develop numerical methods to produce refined interior images, while acquiring no or minimal additional radiation exposure. A section image with a spatial resolution lower than that desired in the RoI, along with associated measurements, is required. This image can be an existing image, in which an RoI is to be refined, or a coarser image acquired from radiation exposure confined to the RoI. The section image enables the estimation of the contribution, (to acquired measurements), of the material outside the RoI. Measurements that depend only on the RoI are then calculated by virtually isolating the RoI from its surroundings, and subtracting the contribution of the outside region from recorded measurements. When a section image is available from an existing tomograph, it has a high degree of overdetermination (number of measurements over the number of unknown pixel attributes). It is shown that this degree of overdetermination is further increased significantly after applying virtual isolation, due to the overlapping nature of CT projections. This enables RoI image reconstruction over finer pixels, hence improving the spatial resolution of the image. Moreover, the reduced size of the RoI enables the use of the powerful, but more computationally demanding, iterative image reconstruction methods; namely the modified convex maximum likelihood (MCML) algorithm and the conjugate gradient (CG) method. These methods are known for their enhancement of image contrast, due to their lower susceptibility to error propagation; a feature desired to control the effect of the increased measurement uncertainty associated with virtual isolation. These improvements in RoI image spatial resolution and material contrast were numerically demonstrated in four simulated human phantoms in which noisy measurements were synthesized. When an RoI image is obtained with a confined (narrow) field-of-view, to reduce radiation exposure, it is shown that the section coarse image required in virtual isolation can be adequately reconstructed from available measurements using the MCML method. Applying either the MCML or CG method to reconstruct the RoI image was shown to produce images comparable in quality to entire-section images generated by a wide field-of-view conventional CT system. | |
| dc.description.copyright | ©Esmaeil Enjil Ela, 2012 | |
| dc.description.note | Scanned from archival print submission. | |
| dc.format | text/xml | |
| dc.format.extent | xxxiv, 318 pages | |
| dc.format.medium | electronic | |
| dc.identifier.other | Thesis 9114 | |
| dc.identifier.uri | https://unbscholar.lib.unb.ca/handle/1882/14328 | |
| dc.language.iso | en_CA | |
| dc.publisher | University of New Brunswick | |
| dc.rights | http://purl.org/coar/access_right/c_abf2 | |
| dc.subject.discipline | Mechanical Engineering | |
| dc.title | Region-of-interest image reconstruction in computed tomography | |
| dc.type | doctoral thesis | |
| thesis.degree.discipline | Mechanical Engineering | |
| thesis.degree.fullname | Doctor of Philosophy | |
| thesis.degree.grantor | University of New Brunswick | |
| thesis.degree.level | doctoral | |
| thesis.degree.name | Ph.D. |
Files
Original bundle
1 - 1 of 1