Research project

The goal of this research project is to develop, improve and evaluate approaches for exact image reconstruction for computed tomography (CT) systems. The projection data is aquired along novel trajectories using a cone-beam setup. For data acquisition, modern C-arm systems are well suited. In addition to the selection and evaluation of the trajectories, the focus lies on the development of strategies for a correct handling of real data as well as on the reduction of reconstruction artifacts. The calibration of the system, which is needed to correctly account for deviations from the ideal geometry in the reconstruction step, is a most challenging problem. The project is supported by our industrial partner Siemens Medical Solutions who also provides access to the latest computed tomography devices in order to enable the development under realistic conditions. The project is sceduled over three years: In 2005, a prototype of the M-line algorithm for exact image reconstruction from real C-arm data has been developed for a circle-plus-arc data acquisition setup during a research visit at the research facility UCAIR (www.ucair.med.utah.edu) of the Univeristy of Utah. The work was supervised by Prof. Frederic Noo. In 2006, we concentrated on the evaluation of this prototype as well as on its improvement with respect to reconstruction artifacts. In a continuative collaboration with the research facility UCAIR (www.ucair.med.utah.edu) of the Univeristy of Utah, we were able to further improve the image quality of our approach. The results have been published at the 2006 IEEE MIC conference in San Diego (CA, USA) in two contributions where the robustness of our algorithms has been confirmed. In 2007, we continued our work and concentrated on the detailed analysis of the following topics: 1) Reconstruction with truncated projections. 2) Evaluation of a fundamental implementation step for accurate image reconstruction (derivative along the trajectory). 3) Acquisition and reconstruction of clinically relevant real cone-beam data. Topic 1) comprises the evaluation of artifacts resulting from axial and trans-axial truncated projections when applying the M-line algorithm for the circle-plus-arc trajectory. First results have been published at the 3D07 conference in Lindau (Germany). Topic 2) comprises the evaluation of a new, numerically robust method to compute the derivative along the trajectory as well as its comparison to alternative approaches. First results have been published at the 2007 IEEE MIC conference in Honolulu (Hawaii, USA). Topic 3) concentrates on the identification and elimination of cone-beam artifatcs on real and clinically relevant data. For that purpose, several data sets of a human thorax phantom have been acquired and reconstructed with the M-line algorithm. First comparisons with the Feldkamp algorithm, which is installed in todays clinical facilities, show promising results with respect to the elimination of cone-beam artifacts. Those results will be published in 2008 in one of the relevant journals. The work was supported by two further research visits at the Univeristy of Utah.