For high qualityMR imaging of time-varying objects such as beating heart or brain hemodynamics, we need to reduce signal acquisition time without sacrificing the spatial resolution. Considerable efforts have been made to achieve this goal especially in the area of parallel imaging and temporal filtering techniques. Recently, researchers have tried to combine temporal filtering with parallel imaging so that any spatial residual artifact from parallel imaging can be further suppressed by temporal filtering. This paper extends the idea and
proposes a new algorithm called the x-f SENSE (x-f domain SENSitivity Encoding) that optimally combines the parallel imaging and temporal filtering using the powerful lattice sampling theory. Unlike the conventional hybrid methods such as TSENSE and UNFOLD-SMASH, the x-f SENSE algorithm is optimal in the sense that it achieves the minimal sampling limit whereas the reconstruction procedure is as simple as that of the conventional SENSE. Theoretical analysis and simulation results demonstrate that x-f SENSE achieves theoretically optimal performance limit without any artifact.