"Image-guided radiotherapy using cone-beam computed tomography (CBCT) provides volumetric imaging before treatment to reduce geometric uncertainties. However, respiratory-induced motion in thoracic and abdominal regions can blur CBCT scans. Four-dimensional (4D) CBCT reduces this blur by sorting projection data into motion-free subsets, but this results in streak artifacts due to fewer projections per subset. This thesis developed new methods to mitigate these under-sampling streak artifacts and evaluated their clinical relevance, exploring both analytic and algebraic reconstruction approaches. Firstly, a directional sinogram interpolation (DSI) method was introduced to increase the number of projections by extracting directional features from the sinogram and interpolating new projection images. DSI demonstrated fewer streaks and less blur than existing methods, enhancing the similarity between CBCT and planning CT scans. Secondly, DSI was extended to 4D CBCT through a motion-weighted reconstruction (MWR) approach that combines regular 3D CBCT with interpolated 4D CBCT. MWR effectively reduced streak artifacts and improved image quality in both static and moving regions compared to conventional methods. Thirdly, an algebraic reconstruction method utilizing a cost function with spatial and temporal regularization was proposed. Despite varying optimal parameter values, this method reduced streak artifacts compared to Feldkamp-Davis-Kress (FDK) methods. Finally, the clinical relevance was assessed by evaluating delineation accuracy on CBCTs reconstructed using the proposed methods. The algebraic CBCT showed improved delineation accuracy over conventional FDK CBCT in head and neck and breast cancer patients, suggesting potential for adaptive radiotherapy. Overall, the developed methods effectively reduced streak artifacts in sparsely acquired CBCT images, enhancing image quality and potentially improving clinical outcomes in radiotherapy."--