Ensuring the long-term reliability and safety is the major challenge faced for geological CO2 sequestration projects. Since the wellbore of injection and monitoring wells are the most probable leakage path of CO2 into overlying formations, aquifers or even to the surface, well integrity is of utmost importance. Poor zonal isolation caused by failure of Portland-based cements used as annular sealant in these corrosive downhole environments needs therefore be thoroughly investigated. Since the carbonation of this material in presence of CO2 and water results in cement degradation. Therefore, innovative testing equipment in the form of the computed tomography (CT)-scannable in-situ test cell is used to determine the long-term effect of supercritical CO2 on the cement matrix under permanently maintained simulated downhole conditions. This master thesis presents the results of this long-term exposure test of a cement system to a corrosive downhole CO2-environment, with focus put on the evaluation procedure for the CT-scans conducted during the experiment. This procedure is based on application of artificial neural networks (ANN) and algorithms to assign the individual pixels of the CT-images to certain phases and afterwards conducting precise pixel-based measurements. The results show significant alteration of the tested cement if exposed to a corrosive downhole CO2-environment. Additionally, the initial function as gas migration pathways and the later auto-sealing capability of formed cracks within the cement core are pointed out. Especially the respective influence of these two occurrences, with a sharp increase in initial stages and gradual decrease at later stages, on the measured propagation rate of the carbonation front was characterized by the conducted measurements.