Development of Novel Cardiac CT Methods

Cardiovascular diseases (CVD) are the number one killer in US and worldwide. In 2005, 1 in every 2.7 deaths was associated with CVD. Cardiac CT has become a very valuable tool for non-invasive CVD diagnosis due to enabling technologies like helical scanning, fast rotation, multi-row detector, and advanced image reconstruction. However, challenges like cone-beam artifact reduction, temporal resolution improvement and x-ray dose reduction must be met to improve and broaden cardiac CT applications. To address these issues, the methods of interior tomography, compressive sensing based algorithms, knowledge-based dynamic volumetric cardiac CT, and controlled cardiac CT were developed along with novel trajectories like saddle curves and composite-circling scanning. Interior tomography permits reconstruction of a region of interest (ROI) using only the x-rays through the ROI, provided that we have appropriate prior knowledge such as the intensity of a small sub-region within the ROI. We evaluated the dose and scatter reduction capabilities of this method. Our results indicate that x-ray dose and scatter can be reduced respectively by ~18 to ~57% and ~19 to ~59% for SFOVs of 21 to 9 cm as compared to the 50 cm SFOV. We evaluated the possibility of using blood in aorta and bone in vertebra as known information. To enable accurate interior tomography reconstruction, we also developed an extended projection onto convex sets (POCS) based method and studied the contributions of individual constraints. Using the compressive sensing theory, it has recently become possible to reconstruct a piece-wise constant ROI using interior tomography. The ability of a total-variation minimization based algorithm was then demonstrated with a limited number of projections. Since the circular cone-beam trajectory does not satisfy the data completeness condition, which results in artifacts, we extended it to the composite-circling trajectory. We obtained promising results, for the thorax phantom, using just 50 views. Knowledge-based dynamic volumetric cardiac CT algorithm is capable of reconstructing multiple-states of a beating heart based on the relationship between the state (e.g. volume) and phase (e.g. ECG). This algorithm was originally proposed using circular trajectory, which does not allow exact reconstruction. We extended it to a saddle-curve trajectory for exact reconstruction. Promising results were obtained with this knowledge-based approach for accurate reconstruction. Controlled cardiac CT proposes instantaneous change in the x-ray source velocity to acquire the required projections in a minimum time. The fast acquisitions with this approach result in a non-uniform distribution of x-ray source positions along a scanning trajectory. The implementation of analytical filtered-backprojection reconstruction algorithms used involves multiplication with a uniform angular discretization term. This could lead to reduced image quality especially if fewer views are used. We used triangle based linear interpolation, quadrilateral based linear interpolation and nearest neighbor interpolation to estimate projection data on uniform grid. We performed various weighting based reconstructions. We concluded that the triangle based linear interpolation provides the best image quality while the quadrilateral based linear interpolation provides the best noise performance.

Grantor dc:publisher

Wake Forest University

Year dc:date.issued

2010

Author dc:creator

• Bharkhada, Deepak

• Computed Tomography

Language dc:language.iso

en_US

Handle dc:identifier.uri

http://hdl.handle.net/10339/14902

OAI identifier oai:identifier

oai:wakespace.lib.wfu.edu:10339/14902