Contributions to the improvement of image quality in CBCT and CB?CT and application in the development of a CB?CT system

Resumen

During the last years cone-beam x-ray CT (CBCT) has been established as a widespread imaging technique and a feasible alternative to conventional CT for dedicated imaging tasks for which the limited flexibility offered by conventional CT advises the development of dedicated designs. CBCT systems are starting to be routinely used in image guided radiotherapy; image guided surgery using C-arms; scan of body parts such as the sinuses, the breast or extremities; and, especially, in preclinical small-animal imaging, often coupled to molecular imaging systems. Despite the research efforts advocated to the advance of CBCT, the challenges introduced by the use of large cone angles and two-dimensional detectors are a field of vigorous research towards the improvement of CBCT image quality. Moreover, systems for small-animal imaging add to the challenges posed by clinical CBCT the need of higher resolution to obtain equivalent image quality in much smaller subjects. This thesis contributes to the progress of CBCT imaging by addressing a variety of issues affecting image quality in CBCT in general and in CBCT for small-animal imaging (CB?CT). As part of this work we have assessed and optimized the performance of CB?CT systems for different imaging tasks. To this end, we have developed a new CB?CT system with variable geometry and all the required software tools for acquisition, calibration and reconstruction. The system served as a tool for the optimization of the imaging process and for the study of image degradation effects in CB?CT, as well as a platform for biological research using small animals. The set of tools for the accurate study of CBCT was completed by developing a fast Monte Carlo simulation engine based on GPUs, specifically devoted to the realistic estimation of scatter and its effects on image quality in arbitrary CBCT configurations, with arbitrary spectra, detector response, and antiscatter grids. This new Monte Carlo engine outperformed current simulation platforms by more than an order of magnitude. Due to the limited options for simulation of spectra in microfocus x-ray sources used in CB?CT, we contributed in this thesis a new spectra generation model based on an empirical model for conventional radiology and mammography sources modified in accordance to experimental data. The new spectral model showed good agreement with experimental exposure and attenuation data for different materials. The developed tools for CB?CT research were used for the study of detector performance in terms of dynamic range. The dynamic range of the detector was characterized together with its effect on image quality. As a result, a new simple method for the extension of the dynamic range of flat-panel detectors was proposed and evaluated. The method is based on a modified acquisition process and a mathematical treatment of the acquired data. Scatter is usually identified as one of the major causes of image quality degradation in CBCT. For this reason the developed Monte Carlo engine was applied to the in-depth study of the effects of scatter for a representative range of CBCT embodiments used in the clinical and preclinical practice. We estimated the amount and spatial distribution of the total scatter fluence and the individual components within. The effect of antiscatter grids in improving image quality and in noise was also evaluated. We found a close relation between scatter and the air gap of the system, in line with previous results in the literature. We also observed a non-negligible contribution of forward-directed scatter that is responsible to a great extent for streak artifacts in CBCT. The spatial distribution of scatter was significantly affected by forward scatter, somewhat challenging the usual assumption that the scatter distribution mostly contains low-frequencies. Antiscatter grids showed to be effective for the reduction of cupping, but they showed a much lower performance when dealing with streaks and a shift toward high frequencies of the scatter distributions. --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------