The use of band-selectable digital filtering in magnetic resonance image enhancement.

Manipulation of the data describing two-dimensional magnetic resonance (MR) images can be used to zoom an image, decrease image noise and artifacts by modeling, or emphasize object edges in the field of view. In this paper, a two-dimensional band-selectable digital filtering (2D-BSDF) technique is detailed. This can be used to decrease the computational burden and increase algorithm stability associated with such data manipulation. Many display devices have the ability of expanding an image by pixel or linear interpolation. Application of the efficient zooming fast Fourier transformation algorithms provides a superior quality sinc-function interpolated image. In 2D-BSDF, the ideal rectangular windows used in sinc-function interpolation are replaced by windows with a more gradual roll-off. This gradual roll-off results in a slight degradation of the image edges but substantially reduces the computation time. Modeling of MRI data has been attempted to remove noise and artifacts from the image. These algorithms are computationally expensive and frequently unstable because of the high model orders required. The 2D-BSDF can be used to prepare a reduced data set, without loss of information. A lower order model may be applied to the subset and computation times approaching that required for normal fast Fourier transform algorithms result. The absence of noise and signal from objects outside the region of interest can considerably enhance the stability of the modeling algorithms. The use of BSDF is equally applicable when used in association with the modeling of 2D NMR spectroscopy data or with edge enhancement or any other data manipulation of magnetic resonance imaging images. In this paper an explanation of 1D-BSDF is provided and an algorithm for 2D-BSDF is developed. A comparison of filter designs and computational times is given when applying the technique to zooming and modeling of MR images. Images from medical MRI data are provided.