23Na lung MRI is a potential tool e.g. for investigating tumor viability and treatment response. The acquired data is, however, influenced by respiratory motion since breath-hold acquisitions are not possible due to the long measurement times. Here, we present a method for full flexibility in retrospective gating of 3D-radial data, combined with a Compressed-Sensing reconstruction. Up to 14 frames could be reconstructed from one in-vivo dataset, corresponding to a temporal resolution of 0.27 s. The temporal resolution of 23Na lung MRI could potentially lead to a more accurate quantification of the Tissue Sodium Concentration in the human torso.