A new interpolation technique for orthogonal sets of tomographic medical images

A spatially iterated, multidimensional Fourier interpolation method (SIMFI) is presented. The method interpolates a multidimensional object from one or more sets of parallel biplanes. It may be applied in medical tomographic imaging where a full-resolution, three-dimensional reconstruction is desired, but, due to time and radiation constraints, the data acquired from the scanner are limited to one or more sets of planar slices with large interspatial domain information. The data sets of parallel hyperplanes are used to compute an estimate of the Fourier domain of the object being reconstructed. After inverting the Fourier domain estimate into the spatial domain, the principle of projections onto convex sets (POCS) is used to further improve the reconstruction. This method is applied to a three-dimensional phantom and to a two-dimensional MRI image of the brain. The Fourier interpolation provides a good initial estimate for he iterative POCS technique, which produces sharper images with significant improvement in homogeneous regions. The effect of aliasing in three-dimensional reconstructions from typical MRI images is significant. This effect can be reduced using the proposed method.