MR DATA ACQUISITION AND RECONSTRUCTION USING EFFICIENT SAMPLING SCHEMES

Magnetic resonance imaging is most often performed with a combination of magnetic gradients and digital sampling that yields a regular rectangular grid of data points in K-space. This rectangular pattern is inefficient for data which are isotropically band limited, as is typical of most medical images. More efficient sampling patterns require fewer data points and less acquisition time. In two dimensions, a hexagonal sampling pattern is most efficient, requiring up to 13.4% fewer data points than standard methods to obtain the same resolution without aliased image overlap. In three dimensions, body-centered cubic sampling is most efficient, with an efficiency gain of 29.3%. A modified FFT algorithm for hexagonal sampling has been derived and implemented. Images have been reconstructed with no loss of resolution on a 256 x 256 square pixel format from 224 x 256 hexagonally sampled raw data points with no interpolation. As expected, these images demonstrated aliasing in a hexagonal pattern around the primary image. This technique can be used to reduce scan time, data throughput, and data storage requirements while maintaining resolution or to improve resolution while maintaining scan time. © 1990 IEEE