Compared with conventional Cartesian kspace sampling, the nonCartesian sampling can enable higher coverage effeciency of kspace, more efficiently make use of the gradient system performance, and reduce dB/dt to prevent to cause the undesirable human physiological reactions. The combination of nonCartesian kspace sampling and parallel imaging can further accelerate imaging speed, however the artifact pattern in image domain would become much more complicated, which introduces a lot of technical difficulties to nonCartesian parallel MRI reconstruction. In this article, several typical nonCartesian parallel imaging reconstruction techniques including SENSE, CGSENSE, nonCartesian GRAPPA, SPIRiT and newlyemerging compressed sensing are reviewed, their technical details, advantages and disadvantages are discussed. SENSE and CGSENSE can achieve optimal reconstruction results theoretically, but both of them are restricted by the accurate measurement of coil sensitivity distribution. NonCartesian GRAPPA doesn’t rely on coil sensitivity measurement, but can only perform approximate calculation for specified nonCartesian sampling mode. SPIRiT combines the advantages of SENSE and GRAPPA, and can obtain satisfactory result by using iterative optimization algorithm. Taking the advantage of sparse transform characteristic of images, compressed sensing cooperating with existing iterative optimization parallel imaging method can further improve reconstructed image quality, and it will be a hotspot in the future study.