We propose a 3D+t/2D+t registration strategy to relate preoperative CTA to intraoperative X-ray angiography for improved image guidance during coronary interventions. We first derive 4D coronary models from CTA and then align these models both temporally and spatially to the intraoperative images. Temporal alignment is based on aligning ECG signals, whereas the spatial alignment uses a similarity metric based on centerline projection and fuzzy X-ray segmentation. In the spatial alignment step we use information from multiple time points simultaneously and take into account rigid respiratory motion. Evaluation shows improved performance compared to a 3D/2D approach with respect to both registration success and reproducibility.