Iatrogenic air embolisms play an often underestimated role in the occurrence of postinterventional cerebral infarcts, potentially resulting in neurological dysfunction, in various cerebral and cardiac endovascular interventions. The brain is particularly vulnerable to air embolisms due to its hypoxic sensitivity and poor collateralization. In this dissertation, a model is developed to produce and inject air bubbles of defined and reproducible size and number. This model is based on a microfluidic system that allows the number and size of air bubbles to be recorded and controlled in real time immediately prior to injection. Hereby, a systematic investigation of air emboli in a small animal model is possible - by MR imaging studies in vivo as well as by histological examinations post mortem. Thus, it is possible to investigate the influencing factors in air embolism and the pathophysiology of the resulting damage in more detail. The first step is to establish the model. It could be shown that air bubbles of defined size can be generated precisely and reproducibly (median: 85.5 ?m, SD 1.9 ?m). Automatic counting and measurement of the air bubble size allow real-time control of the injected bubbles. The ability to vary the size of the microfluidic channel and to adjust the pressure in the microfluidic system allows to produce air bubbles of different sizes. Thus, the model provides the basis for systematic studies on air embolisms, such as the influence of bubble size, bubble number, and application site on the infarct pattern. In most cases of air embolisms, the air is accidentally injected into the vasculature through a catheter. Therefore, in a second step, air bubbles proximal and distal to a catheter are measured and thus the influence of a microcatheter on the air bubbles is investigated. It is found that after passage through the catheter, the mean diameter of the bubbles hardly change (median: 86.6 ?m), but passage through the catheter results in a larger dispersion of the air bubble size (SD 29.6 ?m) and in a lower number of air bubbles (60.1 % of the injected air bubbles). In a third step, the influence of the number of air bubbles and the site of application on cerebral injury is investigated. This confirms the hypothesis that a higher number of air bubbles leads to a higher number of infarcts with a larger total infarct volume. Interestingly, it can be shown that when the same number of air bubbles with uniform bubble size is injected into the aorta at the level of the aortic valve, a similar number of cerebral infarcts can be detected as when the bubbles are injected directly into the carotid artery (5.5 vs. 5.5 (medians); p=0.769). Also the pattern of infarcts is comparable at both sites of application. However, infarcts are larger when air bubbles are injected into the carotid artery (median infarction volume: 0.41 mm³ vs. 0.19 mm³; p<0.001). This result has high clinical relevance as it shows that the risk of air embolism from cardiac interventions can be similar to that from neuroradiological interventions. When injected into the right atrium - as a model for venous air embolism - no infarcts are detectable. This shows that paradoxical air embolism does not occur with the total amount of air used in this model, most likely because the filtering capacity of the lung is not exceeded. The model established in this dissertation is highly relevant for further studies on pathophysiology, prophylaxis, and therapy of iatrogenic air embolism because the innovative setting (air bubble size, air bubble number, total air volume, transfemoral catheterization) largely corresponds to the simulated situation in catheter-based interventions.