| In the past few years, numerous explicit and implicit methods have arisen to deal with the problem of intra-fractional, more specifically, respiration-induced lung tumor motion. Regardless of the method of choice (explicit, implicit, standard), all planning computed tomography (CT) data sets, with the exception of breath-hold acquisitions, are acquired while the target is non-stationary and are thus subject to the presence of motion artifacts. While the presence of these motion-induced aberrations is generally acknowledged, there exists a paucity of published literature concerning their quantification. In the current pursuit to achieve better treatment outcomes in conformal lung radiotherapy, a better understanding must be gained of the detrimental effects of respiration-induced tumor motion upon CT imaging. In this doctoral work, the principles of motion artifacts in CT, the functional dependence of motion artifacts associated with lung tumor motion and the impact of respiration-induced tumor motion upon CT image integrity and target delineation were investigated. A filtered backprojection (FBP) computer model developed in MATLAB and experimentally validated under both static and dynamic conditions served as the main investigational tool. Overall, it was found that respiration-induced motion projects the reconstructed object in directions displaced from its velocity vector. This is in contrast to the smearing of mobile objects in planar radiography which is always along the line of motion. It was also found that spatial extent of a mobile object is distorted from its true shape and location and does not accurately reflect the total volume occupied by the mobile object during the extent of motion captured. Finally, the presence of motion during data acquisition leads to density distributions that are altered from their true physical time-averaged density distributions. A byproduct of this research was the development of a slice-specific imaging technique which provides direct assessment of the extent of free-breathing tumor motion without recourse to gating techniques. The methodology, better known as phase sequence image, was experimentally evaluated using a three-dimensional lung tumor motion phantom and found to accurately estimate the internal target volume at a contrast threshold level of about 20%. |
