Capitalizing on the efficiency of non-Cartesian imaging and iterative reconstruction, we sought an improved 3D iNAVs acquisition with isotropic resolution, to facilitate whole-heart motion correction with translational or more advanced models... In this work, we propose a method providing 3D motion correction on a per-heartbeat basis using a variable-density 3D cones iNAV acquisition... However, using a variable-density design, decreasing the sampling density from 1.0 at the k-space origin to 0.26 at kmax, a 32-readout trajectory was achieved corresponding to an acceleration factor of 9 and acquisition time of 175 ms. 3D iNAVs were first reconstructed with gridding which served as a starting point for ESPIRiT... Whole-heart images with 1.25 mm isotropic resolution were reconstructed with gridding... To test the feasibility of the 3D iNAVs, translational motion correction was applied using linear phase modulation... Figure 1 shows the initial 3D iNAV reconstruction with gridding and the corresponding reconstruction with ESPIRiT... ESPIRiT significantly reduced the aliasing artifacts, revealing contrast and features useful for motion estimation... Figure 2 shows motion estimates derived from the 3D iNAVs, in this case translational, and the resulting sharpening of the right coronary artery after motion correction... Acquiring 3D iNAVs every heartbeat, 3D motion of the heart was measured during a free-breathing coronary MRA acquisition to provide 100% respiratory efficiency and retrospective motion correction... Future work includes applying more advanced models based on the 3D iNAVs for improved correction.