Cone beam computed tomography (CBCT) and magnetic resonance imaging (MRI) are two medical imaging modalities capable of producing time-resolved images. Reconstruction of the acquired imaging data and subsequent processing do not always "take advantage" of the temporal dimension however. In CBCT reconstruction for example, the acquired data is often sorted into temporal bins that are subsequently reconstructed separately and thus not striving for temporal consistency. Similarly, many image segmentation techniques for image sequences are repeated frame-by-frame - treating the temporal dimension notably different from the spatial dimensions.

A recurring topic in this dissertation is integration of temporal image registration into the contexts of medical image reconstruction and segmentation to explore potential improvements in image and segmentation quality.

For both the reconstruction and segmentation task a dedicated cost function is formulated such that minimization hereof is likely to provide the desired result.

For CBCT image reconstruction we propose a new technique that provides better overall results compared to current state-of-the-art algorithms based on total variation regularization and prior image constrained compressed sensing. We have furthermore proposed and thoroughly evaluated a number of new regularization terms in the cost function formulation. That part of our work is intended to bridge the gap between current-practice CBCT reconstruction algorithms applied in the clinic and more advanced algorithms mostly used for research.

On the topic of image segmentation we propose a new technique for segmentating a sequence of MRI images. We combine registration and segmentation to segment the entire image stack concurrently.