In parallel imaging, under-sampled non-Cartesian trajectories help to achieve faster scan times; however the resultant images may have aliasing artifacts. Pseudo-Cartesian GRAPPA ¹ along with GRAPPA Operator Gridding (GROG) ² has been used in the recent past to eradicate these artifacts. In this paper, the application of ESPIRiT ³ (an eigenvectors based reconstruction algorithm) in combination with GROG ² is proposed to get good reconstruction results from highly under-sampled radial data.GROG ² provides mapping of the acquired non-Cartesian data points in k-space to the nearest Cartesian locations. GROG exploits self-calibrated weight sets to grid the non-Cartesian data and leaves some empty spaces in the gridded k-space ². Pseudo-Cartesian GRAPPA ¹ employs several patterns to estimate the coil-by-coil weight sets from auto-calibration signal (ACS) lines to fill the empty points of the GROG ² gridded k-space. These suitable pseudo-Cartesian GRAPPA weight estimating patterns are difficult to determine ¹. ESPIRiT uses eigenvector based multiple sets of the sensitivity maps to get un-aliased image. In ESPIRiT, multiple sets of sensitivity maps are estimated from a small calibration area using eigenvalue decomposition approach ³. The eigenvectors corresponding to the most significant eigenvalues provide sensitivity maps. This paper proposes the application of ESPIRiT ³ with GROG ² to recover MR images from under-sampled radial data. The first step in the proposed method is to apply GROG on the acquired accelerated radial data which provides an estimate of the Cartesian k-space. This is followed by the application of ESPIRiT on this GROG gridded data. Figure 1 provides block diagram of the proposed method. The performance of the proposed method and pseudo-Cartesian GRAPPA ¹ is evaluated using Artifact Power (AP) ⁴ and Root Mean Square Error (RMSE).¹The proposed method is tested on the short-axis cardiac MRI radial data which has been acquired using 3T Skyra (Siemens, Case Western Reserve University, Cleveland) scanner at TR=2.94ms with real-time, free-breathing and no ECG gating. The radial data set contains 30 channel cardiac coils, 80 measurement frames, 256 readout points and 144 projections. This short-axis cardiac radial data is resized to using temporal averaging method on the measurement frames.⁵ Figure 2 shows the reconstruction results of the proposed method as well as pseudo-Cartesian GRAPPA ¹ for different acceleration factors (AF= 4, 6 and 9). Figure 2A shows the reconstruction results of the proposed method and Figure 2B shows the results of pseudo-Cartesian GRAPPA.¹ The results confirm that the proposed method provides significant improvement (e.g. 81% in term of AP, 57% in term of RMSE improvement at AF=4) in the reconstructed images as compared to pseudo-Cartesian GRAPPA.¹The application of GROG followed by ESPIRiT is proposed to reconstruct the MR images from under-sampled radial data. The results show that the proposed method provides significant improvement in reconstruction results as compared to conventional pseudo-Cartesian GRAPPA.