Photon-counting x-ray computed tomography (CT) is the latest advancement in the field of medical x-ray imaging, and this new technology is now adopted in hospitals. Our research is centered around a silicon-based photon-counting CT scanner prototype based on a detector developed by our research group, which promises high spatial resolution, improved noise performance and better energy resolution compared to today. To produce the best possible image based on the data that can be acquired with this prototype scanner, it is necessary to simulations of data acquisition with this new scanner studies and use this simulated data to understand the factors affecting image quality and developing improved imaging hardware and image reconstruction methods. In this project we will simulate data acquisition with photon-counting detectors and CT scanners using CATSim (General Electric Company). This will result in simulated projection data, and this data will be used for two main purposes. The first is to evaluate the image quality achievable with this new detector technology, by calculating contrast-to-noise ratio and more sophisticated metrics of image quality both in the raw data and in the reconstructed image. The benefit of doing this in simulations, as a complement to measurements, is that this provides a greater degree of flexibility in terms of varying parameters or inserting simulated features of known composition in the imaged object. The second main purpose is to generate simulated training data for deep-learning image reconstruction. By generating thousands of simulated images, based on models of human anatomy or on actual patient scans, we can generate input and label images for training deep neural networks for different image enhancement tasks, such as denoising, artifact correction and radiotherapy planning. We plan on using CPU-based resources on Dardel, as a complement to two other GPU-based allocations, NAISS 2024/22-220 (Highly accurate photon-counting CT for cancer imaging with deep-learning image reconstruction) and Berzelius-2024-61 (Deep-learning data processing for photon-counting CT), wherein these simulated images will be used to train deep neural networks. The CPU- and GPU-based compute allocations can thereby complement each other in the development of future medical x-ray imaging equipment and ensuring that the new hardware can be translated to substantial patient benefit. The proposed project is a continuation of the past allocation NAISS 2023/22-485. In the previous allocation, we performed simulations of computed tomography images and used these as training data for a novel deep-learning motion artifact correction method. We also simulated novel detector designs for ionizing radiation imaging. In this allocation we will continue the motion-artifact correction project with more realistic simulated data and investigate the possibility of not only correcting for motion in a single image frame but also creating a motion-corrected movie of a beating heart. We will also use this allocation for investigating the effects of nonideal effects such as pulse piluep and intradetector cross-talk on image quality.