Interpolation algorithms for 3-D reconstruction of magnetic resonance images

This investigation evaluates various interpolation algorithms for generating 'missing' data between multiple 2-D (two-dimensional) magnetic resonance (MR) image planes (slices). Slices when stacked parallel and displayed, represent a 3-D (three-dimensional) volume of data usually with poorer spatial resolution in the third dimension (slice direction). Interpolation algorithms have been developed to determine the missing data between the image slices and to compute a 3-D volume data array representing equal spatial resolution in three dimensions. Known, 3-D volume data arrays (volume phantoms) were used in a computer simulation of MR imaging to produce multiple 2-D slices to evaluate the performance of the interpolation algorithms. Of the interpolation algorithms, the slope-weighted method reproduces the original data set most accurately. The MR imaging simulation results demonstrate that for a given number of image planes, the data are most accurately reproduced by interpolation when the selected image slices are as thin as possible, as opposed to selecting the slices to be thicker and contiguous which results in the worst reproduction of the original data set. One of the goals of this project is to develop the best interpolation scheme for improved generation of 3-D reconstructions of MR images and for visualizing images in planes other than the acquisition plane (reformatting).<>