Current catheter tracking in the x-ray catheter laboratory during coronary interventions is performed using 2D fluoroscopy. Although this features real-time navigation on high-resolution images, drawbacks such as overlap and foreshortening exist and hamper the diagnosis and treatment process. An alternative to fluoroscopy-based tracking is device tracking by means of a magnetic tracking system (MTS). Having measured the 3D location of the interventional device, its position can be reconstructed on 3D images or virtual roadmaps of the organ or vessel structure under examination. In this paper, a method is presented which compensates the interventional device location measured by the MTS for organ motion and thus registers it dynamically to a 3D virtual roadmap. The motion compensation is accomplished by using an elastic motion model which is driven by the ECG signal and a respiratory sensor signal derived from ultrasonic diaphragm tracking. The model is updated during the intervention itself, thus allowing for a local refinement in regions which bear a complex geometric structure, such as stenoses and bifurcations. The evaluation is done by means of a phantom-based study using a dynamic heart-phantom. The mean displacement caused by the overall motion of the heart is improved from 10.4+/-4.8 mm in the uncompensated case to 2.1+/-1.2 mm in the motion compensated case.