The use of 3D whole-heart coronary MRA at 3T is still limited as whole-heart SSFP acquisitions remain challenging. To overcome that hurdle, a slow-infusion Gd injection in conjunction with a gradient-echo readout has been proposed (1) and successfully applied (2,3). To enable free-breathing acquisition, these methods make use of a diaphragmatic navigator (NAV) (4). Respiratory self-navigation (SN) has recently been introduced (5,6) as a promising alternative with respect to NAV; it enables the extraction of respiratory motion information from the imaging data directly, provides 100% scan efficiency, and alleviates the need of sophisticated plan scanning. Typically, a superior-inferior (SI) projection is acquired at the very beginning of each data segment (Figure 1a). One of these SI projections is then taken as reference, and all the other SI projections are compared to that reference in order to estimate and compensate for respiratory motion. The use of slow infusion throughout data acquisition, however, introduces transient signal variations between the reference profile and the subsequently acquired SI projections (Figure 1b). This may adversely affect both the extraction of SI displacement measurements and the efficacy of respiratory motion compensation. A recently introduced iterative approach to SN (IT-SN) operates by iteratively minimizing residual respiratory motion throughout an SI projection series and, thus, without requiring a specific SI projection as a reference profile (7). Here, we tested the hypothesis that IT-SN can provide 100% scan efficiency and improved image quality with respect to standard SN in whole-heart coronary MRA at 3T during slow-infusion.Data acquisition was performed in N=8 patients on a clinical 3T Scanner (Siemens Magnetom, Skyra) using a prototype 3D radial trajectory (8) specifically adapted for performing respiratory SN (Figure 1a). Data acquisition was performed during slow-infusion (Gadovist, 0.03 ml/sec, 0.1 ml/kg). Imaging parameters for the ECG-triggered, fat-saturated, GRE sequence were as follows: FoV (220mm)³, voxel size (1.15mm)³, radiofrequency excitation angle 15°, TE 2.9ms, TR 4.8ms, bandwidth 1002 Hz/pixel. For each data segment, 15-35 radial lines were acquired, over 377-800 heart-beats (~ 12’000 radial lines in total). Inversion recovery preparation was used (TI 290ms). First, the acquired data were reconstructed by using the standard (reference-based) SN and the first SI projection acquired at end expiration was used as the reference profile (9). Secondly, data were motion corrected and reconstructed using IT-SN. Finally, data were also reconstructed without any respiratory motion correction for comparison. Reconstructed datasets were analyzed by using the software tool described in (10); percentage vessel sharpness (%VS) and visible vessel length were computed along multiple coronary segments. Statistical significances were assessed using Bonferroni correction for multiple comparisons.Data acquisition amounted to ≈7 min for every subject, with 100% scan efficiency. All the quantified endpoints are summarized in Table 1. An improvement in coronary delineation is not only observed between uncorrected and corrected datasets (white arrows in Figure 2), but also when IT-SN corrected datasets are compared with datasets corrected by using standard SN (blue arrows in Figure 2). Furthermore, transient signal and contrast variations among the SI projections were observed during slow infusion (Figure 1a, Figure 3a, 3d) and led to failure of the conventional SN algorithm to identify and adequately correct for respiratory displacement of the heart. However, IT-SN remained successful in all cases and a significantly improved visible vessel length and %VS was observed for most coronary segments as compared to SN or non-corrected data. These findings are corroborated by the coronary MRA displayed in Figures 2 & 3.We successfully demonstrated that IT-SN enables self-navigated whole-heart coronary MRA acquisitions during slow infusion of a contrast agent at 3T. As such, indirect estimation of heart motion and sophisticated sequence planning can be avoided; furthermore, 100% scan efficiency is enabled even in the presence of slow-infusion. Simultaneously, and due to the fact that the choice of a specific reference SI projection is no longer needed, both motion detection and the accuracy of the thus related correction may be less affected by transient signal and contrast changes of the SI projections. Preliminary results demonstrate that IT-SN improves image quality with respect to standard SN in the presence of slow infusion at 3T. Studies in larger patient cohorts and in comparison to the gold standard x-ray coronary angiography are now warranted.