Current indications for aortic surgery are solely based on maximum aortic diameter and have proven to be nonreliable. There is an urgent need of pertinent information to aid physicians assess surgical risk-benefits and develop an adequate treatment for the patient as the morbidity and mortality risks for these interventions are considerably high. In this paper, we present an algorithm that semi-automatically creates a personalized 4D model (3D + time) of the patient's aorta from MR images. This model is the first mandatory step towards a quantification of aortic deformation and wall stress analysis. The developed approach complements the information of the oblique sagittal cine images with the anatomical images (oblique coronal, oblique sagittal and pure transverse). A curvilinear structure detector locates the descending aorta on a user-defined sagittal anatomical image. The ascending aorta is detected on the transverse anatomical images through a grey-scale adapted Hough transform, due to regional anatomic complexity, the aortic arch is manually identified on the coronal anatomical images. A 3D initial model is generated and projected onto the cine images using affine transformations. The cine images are enhanced: a Kalman-like filtering technique reduces blood flow artifacts and phase congruency emphasizes the edges. The segmentation stage is carried on using a hybrid level-set approach. Manual editing is conducted on the first temporal volume to compensate errors due to image ambiguities. Temporal segmentation uses the same level-set algorithm. Qualitative and quantitative evaluations of the results obtained from several patient's imaging studies show that our algorithm provides promising results.