INTRODUCTION As an auto-calibrated k-space parallel imaging method, GRAPPA (1) has shown its advantages in some applications when accurate coil sensitivity maps are difficult to obtain. However, there are two main drawbacks for GRAPPA, especially under large imaging acceleration. First, accurate data calibration requires many ACS (Auto-Calibration Signal) lines, which essentially lowers the actual reduction factor. Second, at high reduction rates, missing data are synthesized using data acquired at a far distance which have low correlation with the missing data. The low correlation is a result of the intrinsic narrow-banded sensitivity profiles (3). To overcome these shortcomings, we propose an iterative k-space-based Parallel Reconstruction Using Null Operations (PRUNO). In PRUNO, some local null operators are applied on all k-space locations within a neighborhood rather than only acquired lines. By using these null operators, the reconstruction problem is formulated as estimating missing data from available k-space data by solving a system of linear equations. We also demonstrate that it can be solved efficiently and accurately with a conjugate gradient method. Our preliminary simulation and in vivo results suggest that PRUNO can significantly improve the accuracy of image reconstruction compared to GRAPPA.