To improve the image quality at high accelerations for radial MRI sampling.The radial trajectory is useful for MRI applications such as ultra-short echo time MRI¹. However, radial sampling is more time consuming because the number of readout lines is $$$\pi$$$/2 times that of a Cartesian acquisition. The long acquisition times in radial imaging can be shortened by acquiring fewer lines and using parallel imaging reconstruction. Additionally, recent advances using controlled aliasing in parallel imaging (CAIPI²) can further improve image quality at higher accelerations. CAIPI has recently been further extended to Wave-CAIPI³ that applies wave gradients during readouts to introduce spatially varying convolutions in the image domain. However, existing studies have been limited to Cartesian sampling. In this work we present preliminary results that apply Wave-CAIPI to radial sampling for improved non-Cartesian parallel imaging efficiency.In vivo brain data were acquired on a Siemens (Erlangen, Germany) TimTrio 3T scanner with a 32-channel head coil using a 3D Gradient Echo FLASH sequence (TE/TR=5/50ms, FA=15°, 2mm isotropic resolution). Radial and wave-radial data were then synthesized by retrospectively sampling the fully sampled FLASH k-space data. The wave amplitude was approximately 5% of the separation dk_z and the phase range was from 0 to 2$$$\pi$$$. Figure 1 shows an example of a 3D wave-radial trajectory. The reconstruction was performed using the iterative SENSE method with NUFFT from the Advanced Reconstruction Toolbox⁴. The images reconstructed from the fully sampled data were used as ground truth to compare to images from the undersampled radial and wave-radial data. Reduction factors (in-plane R_xy=4 and through-plane R_z=2) were used.Figures 2 a, b, and c show four axial images reconstructed from fully-sampled radial, undersampled radial, and undersampled wave-radial sampling, respectively. Even though the sensitivities of the 32-channel coil provide encoding power for good overall image quality, close inspection shows through-plane aliasing in the regions annotated by the dashed cyan circles in b. These artifacts do not appear in the images from the undersampled wave-radial data in c. The differences between the fully sampled image and the undersampled radial and wave-radial are shown in d and e, respectively. Note the larger difference in the undersampled radial images (regions encircled by the cyan dashed lines) than in the undersampled wave-radial images.This abstract shows that Wave-CAIPI is also useful for improving image quality at high accelerations in non-Cartesian sampling such as radial sampling. Future work will be to find optimal wave parameters (the amplitude and frequency of the sine/cosine function) for the highest image quality and minimum geometry factor.