To evaluate the clinical performance of a 3D-isotropic self-gated radial volumetric interpolated breath-hold examination (VIBE) for late-phase MR imaging of the liver.70 patients with suspected intrahepatic mass were included and underwent liver MRI at 1.5T (Magnetom Aera, Siemens Healthcare, Germany). Depending on the clinical indication, either Gd-EOB-DTPA (35 patients) or Gadobutrol (35 patients) were administered. During late-venous (Gadobutrol) or hepatocyte-specific phase (Gd-EOB-DTPA), a self-gated 3D-isotropic radial VIBE prototype sequence was acquired. The acquisition principles of the underlying k-space readout using the stack-of-stars scheme have been described recently in detail [1-3]. In short: Cartesian k-space sampling is only performed along the slice direction (kz) whereas in-plane (kx; ky) readout is performed along rotated spokes. The resulting spoke overlap in the center facilitates a more motion-robust data acquisition. Two series were reconstructed from the data set: 1) Using the complete data set (r-VIBE-100) and 2) prototype algorithm with implementation of a retrospective self-gating approach that allows sorting the data into bins according to the respiration phase. The self-gating signal was extracted from the k-space center, which is crossed by every spoke in the central k-space partition. For image reconstruction, 40% of the acquired data at the most consistently visited respiratory position, typically close to end-expiration, were used (r-VIBE-40) to further reduce motion artifacts. Schemes of the stack-of-stars trajectory and the self-gating algorithm are provided in Figure 1. A standard Cartesian VIBE (std-VIBE) served as reference. Images were assessed qualitatively (overall image quality, sharpness, lesion detection, artifacts due to streak or motion, biliary tree) on a 5-point Likert-scale (5=excellent) and quantitatively (coefficient of variation (CV); contrast-ratio) in axial and coronal reformations by two readers independently. Statistical analyses were performed using SPSS (Version 22, IMB, Armonk, USA). Non-parametric testing using Friedman´s ANOVA was conducted for qualitative analysis. For quantitative values ANOVA analysis and post-hoc testing were performed. P-values <0.05 were considered to indicate significance.In eight cases (11%) only radial VIBE provided diagnostic image quality to detect relevant findings such as portal venous thrombosis or liver lesions (Figure 2/3) whereas std-VIBE was considered non-diagnostic. Image quality (p<0.05) and sharpness (p<0.001) of r-VIBE-40 were rated significantly superior in GD-EOB-DTPA-enhanced scans and coronal reformations as compared to std-VIBE. Lesion detection was significantly improved (p<0.05) in coronal reformations of Gd-EOB-DTPA-enhanced r-VIBE-40 scans in comparison to std-VIBE. In axial planes streak artifacts were graded significantly higher (p<0.001) in r-VIBE-40 as compared to r-VIBE-100 and std-VIBE. In coronal multiplanar reformations, image scores for motion artifacts were significantly lower (p<0.001) in r-VIBE-40 than in r-VIBE-100 and std-VIBE. No differences could be observed in the depiction of the biliary tree. CV was significantly higher in r-VIBE-40 as compared to the other sequences (p<0.01). Highest contrast-ratios were found in r-VIBE-40 and std-VIBE with significant differences to r-VIBE-100 (p<0.001) in Gadobutrol-enhanced scans, whereas no differences were found in Gd-EOB-DTPA-enhanced examinations. This study demonstrates that 3D-isotropic self-gated radial VIBE significantly improves image quality, especially in coronal multiplanar reformations and late-phase Gd-EOB-DTPA-enhanced scans as compared to non-gated radial VIBE and standard Cartesian VIBE. This has a significant clinical impact, as patients with liver disease frequently are unable to properly hold their breath. Therefore, based on the results of this study we consider 3D-isotropic self-gated radial VIBE as a promising approach to reduce the number of non-diagnostic examinations due to motion-induced impairment of image quality. Moreover, in 11% of the cases only radial VIBE provided sufficient image quality to detect relevant pathologies, thus having a direct implication for patient care. The improved robustness of the assessed self-gated 3D radial prototype sequence also allows acquiring data during free breathing beyond an individual patient´s breath-hold capability. This is especially useful for the assessment of discrete anatomic and pathologic details. Furthermore, the possibility of reconstructing arbitrary reformations of the 3D-isotropic data sets without additional reconstruction time poses a viable strategy to further increase diagnostic accuracy in liver imaging. A current disadvantage of radial VIBE is the relatively long acquisition time, due to reduced sampling efficiency and the lack of parallel imaging. However, to facilitate diagnostic image quality in dynamic contrast-enhanced MRI, especially important for the diagnosis of hepatocellular carcinoma [4], different image reconstruction approaches have been developed including k-space-weighted image contrast reconstruction (r-VIBE-KWIC) [5] or extra-dimensional golden-angle radial sparse parallel MRI (XD-GRASP) [6]. Thus, radial k-space sampling combined with iterative reconstruction algorithms seems a valuable approach for abdominopelvic and dynamic liver MRI.Utilization of self-gated 3D-isotropic radial VIBE provides significantly superior image quality as compared to the standard Cartesian VIBE, especially in multiplanar reformations and Gd-EOB-DTPA-enhanced late-phase examinations.