The paper introduces ESPIRiT, a new autocalibrating parallel MRI method that combines the advantages of SENSE and GRAPPA. It shows that both methods can be viewed as subspace methods, where the solution is restricted to a subspace spanned by coil sensitivities. ESPIRiT uses an eigenvalue decomposition to compute sensitivity maps directly from autocalibration lines in k-space, resulting in robust and high-quality maps. This method is more flexible and efficient than traditional SENSE and GRAPPA, as it allows for multiple sensitivity maps and relaxed constraints. The paper evaluates the performance of ESPIRiT on various data sets, demonstrating its effectiveness in reconstructing images with high quality and minimal artifacts. The method is implemented using an eigenvalue decomposition approach, which is computationally efficient and can be applied to both 2D and 3D data. The results show that ESPIRiT provides better image quality and is more robust to errors compared to SENSE and GRAPPA. The paper also discusses the theoretical foundations of the method, including the relationship between the calibration matrix and the null space, and the properties of the reconstruction operator. The results demonstrate that ESPIRiT is a promising new method for parallel MRI that bridges the gap between SENSE and GRAPPA.The paper introduces ESPIRiT, a new autocalibrating parallel MRI method that combines the advantages of SENSE and GRAPPA. It shows that both methods can be viewed as subspace methods, where the solution is restricted to a subspace spanned by coil sensitivities. ESPIRiT uses an eigenvalue decomposition to compute sensitivity maps directly from autocalibration lines in k-space, resulting in robust and high-quality maps. This method is more flexible and efficient than traditional SENSE and GRAPPA, as it allows for multiple sensitivity maps and relaxed constraints. The paper evaluates the performance of ESPIRiT on various data sets, demonstrating its effectiveness in reconstructing images with high quality and minimal artifacts. The method is implemented using an eigenvalue decomposition approach, which is computationally efficient and can be applied to both 2D and 3D data. The results show that ESPIRiT provides better image quality and is more robust to errors compared to SENSE and GRAPPA. The paper also discusses the theoretical foundations of the method, including the relationship between the calibration matrix and the null space, and the properties of the reconstruction operator. The results demonstrate that ESPIRiT is a promising new method for parallel MRI that bridges the gap between SENSE and GRAPPA.