Rezension

Since a few years there is the possibility of tomographic imaging with a C-Arm-system in addition to the conventional X-ray-computed tomography. By the use of a flatpanel detector the C-Arm-CT offers a high isotropic resolution. Besides the reduction of dose the improvement of image quality is on the top of the user´s list of wishes. To improve the image quality at constant dose or allow dose reduction at changeless image quality methods of noise reduction are used in conventional CT-imaging. To reduce overall measurement- and reconstruction-time so-called on-line-compliant systems are developed which start reconstruction before the measurement is competed. The aim of this work is the development of algorithms for noise reduction in projection data which shall be applied especially to flatpanel-CT and fit in into online-compliant systems. Among the so far known noise reduction methods are the convolution based multidimensional adaptive filtering by Kachelries, Watzke and Kalender (MAF{sup KWK}) and the spline and statistic based filtering by La Riviere and Billmire (SSAF{sup RB}). The former can not be applied for on-line-reconstruction, the latter can be applied to one-dimensional data only. Both methods are developed further to overcome these restrictions. In addition a hybrid method from a combination of a convolution based and the spline and statistic approach is developed. The impact of the algorithms to noise and resolution is characterized using so-called {sigma}-FWHM-curves from simulated and measured one- and two-dimensional data, respectively. The change in noise impression and structure is considered by means of slices. Examples of the application to clinical data rounds out the comparison. The results of this work are a new convolution based adaptive filtering (CAF), which is on-line-compliant, a spline and statistic based filtering for two-dimensional data (SSAF{sup B2d}) and a hybrid method (Hybrid{sup CAF}). These new adaptive algorithms for noise reduction conform with the given requirements. For all examined adaptive algorithms applies: The more isotropic the noise becomes by the filtering the more anisotropic becomes the resolution. The methods differ only little in terms of the characteristics of the {sigma}-FWHM-curves. They separate in function and the resulting noise impression. The results yield recommendations for the choice of a filtering method: If the projection data is complete before the reconstruction starts and a discontinuous transition from modified to unmodified parts of the slices is tolerable, the use of the MAF{sup KWK} gives the possibility to set the maximum fraction of modified data. Is the reconstruction to be started before the acquisition is complete, one can choose from the following algorithms according to the additional needs: If the transition from modified to unmodified parts of the slices has to be as smooth as possible or if only one parameter should be given the SSAF is the method of choice for time-uncritical reconstructions. If the fraction of modified projection data is to be as small as possible, the Hybrid{sup CAF} should be used. For time-critical reconstructions there is the CAF{sup Tanh} with two parameters available. This theses, within which new filtering methods for noise reduction were developed, the properties of the methods were shown, and hence usage recommendations were derived, broadens the basis for clinical as well as technical advancements in the field. (orig.)