The signal suppression of unwanted lipid resonances is increasingly challenging at high magnetic field strengths due to increased B0 and B1 inhomogeneities. In a number of specific applications, e.g. coronary MRA, fat suppression is of vital importance, since vessels are embedded in lipid tissue layers and fat suppression needs to be homogeneous within a relatively large volume. Not only may residual off-resonance lipid signals hinder the correct anatomical visualization of the small-caliber coronary vessels, but the presence of such signal from static structures (e.g. chest) can also be a major issue when advanced motion compensation techniques are used. In particular respiratory self-navigation [1, 2] relies on the precise detection of the left-ventricular blood pool in a projection image of the whole thorax [3] to perform motion correction and achieve a 100% scan efficiency in free-breathing coronary MRA. When fat signal is not completely suppressed, the blood pool can no longer be isolated and correct tracking becomes increasingly challenging. The aim of this study was therefore to develop and test a novel lipid insensitive binomial off-resonant excitation (LIBRE) pulse for respiratory self-navigated whole-heart coronary MRA at 3T, and to quantitatively assess the resulting image quality in comparison to more commonly used lipid-nulling schemes.Coronary MRA was performed in (n=6) healthy adult volunteers on a 3T clinical scanner (PRISMA, Siemens Healthcare, Erlangen, Germany). Data were acquired using a prototype respiratory-1D-self-navigated ECG-triggered free-breathing 3D radial gradient-recalled-echo (GRE) imaging sequence [3], preceded by an adiabatic T2 preparation module (T2Prep) to improve blood myocardium contrast [4]. All datasets were 1D motion-corrected, using the superior-inferior (SI) projection image of the thorax, and reconstructed at the scanner. Three different scans were performed in each volunteer: 1) a conventional lipid nulling using spectral presaturation of the lipid resonances (fat-sat, or FS), 2) a conventional 1-1 binomial water excitation scheme (WE), 3) the novel lipid insensitive excitation using LIBRE pulses. The pulse design was based on binomial rectangular pulses which are both frequency and phase modulated. Imaging parameters were: field-of-view (220 mm)³, matrix size 192³, TE_T2-Prep = 40 ms, RF excitation angle 18°, TE/TR = 2.5 ms/5.6 ms, with 24 to 32 radial readouts per heartbeat for a total of ~15k k-space lines. To compare the efficacy of the three different lipid nulling methods, the signal-to-noise ratio (SNR) of blood, myocardial and fat tissue, the contrast-to-noise ratio (CNR) between blood and myocardium, and between blood and fat tissue, and the vessel sharpness of the right coronary artery (RCA) were calculated using SoapBubble [5]. Statistics were computed via two-tailed student’s t-test for paired data corrected for multiple comparisons. Finally, the tracking signals of the algorithm for respiratory self-navigation were compared by visual inspection. Data were successfully acquired and reconstructed in all volunteers. Representative MRA images show a clear visual difference between the three methods (Fig. 1, A to C). Both RCA and LAD could be clearly visualized using the LIBRE method (Fig. 1, A1 to A3). In all volunteers, the LIBRE method showed improved lipid nulling in the whole acquired volume when compared to conventional spectral fat suppression and water excitation techniques (Fig. 1). The value of all quantitative endpoints was significantly improved for the LIBRE method, with respect to the two other acquisition schemes. SNR and CNR measurements showed significant improvements comparing the LIBRE water excitation method with conventional water excitation and fat pre saturation (Fig. 2A and 2B). The vessel sharpness of the RCA was improved by 30% on average (Fig. 2C). These results are most likely due to the absence of ubiquitous streaking artifacts caused by unsuppressed chest fat. The effects of the different fat suppression techniques can also be appreciated in sagittal images (Fig. 3A-C), where the LIBRE excitation pulses achieve a near complete suppression of the chest fat, thus generating a clean tracking signal for the self-navigation algorithm (Fig. 3D). Conversely, the suboptimal fat suppression of the other two techniques resulted either in a dampened tracking due to the superimposition of static tissue (Fig. 3E), or to unreliable motion detection due to the lack of sharpness in the delineation of the blood pool (Fig. 3F). The suboptimal motion detection and correction in FS and WE also contribute to the lower final image quality.These preliminary findings demonstrate that novel LIBRE excitation pulses achieve improved fat saturation for coronary imaging at 3T, higher CNR of blood/myocardium and increased vessel sharpness, leading to improved visualization of the coronary arteries. Respiratory self-navigation may particularly benefit from the improved efficacy of this novel pulse.