Scattered radiation reduction in digital volume tomography for dental applications

Cone-beam computed tomography (CBCT) typically suffers from scattered X-rays leading to inaccuracies of CT numbers and artifacts, such as cupping or streaks. A vast variety of scattered radiation estimation and compensation techniques exist. The improved primary modulator scatter estimation (iPMSE) method is based on a spatial modulation of the primary intensity, shifting this desired signal towards higher spatial frequencies. In contrast, the scattered radiation generated in the object is composed of mainly low spatial frequencies. With the modulator in the beam path, seperation of the two signals is possible in either the Fourier- or the image space. In this thesis, the iPMSE method has been implemented and tested on a CBCT device for dental applications. While the iPMSE method has many advantages (e.g. robustness, short computation time), its patch-wise constant approach for the scattered radiation profile may lead to inaccuracies close to sharp object edges in the projection images. To avoid those, a novel convolution-based ansatz was developed in this thesis, in which the scattered radiation profile is estimated by convolving a scatter potential derived from measured intensities with a physically motivated kernel, yielding a hybrid iPMSE (hiPMSE) method. The method was implemented and tested on both simulation and measurement data and showed promising results in both cases where scatter artifacts could be substantially reduced.