An optimized rotator for use in iterative reconstruction [SPECT]

For implementations of iterative reconstruction algorithms that rotate the image matrix, the characteristics of the rotator may affect the reconstruction quality. Desirable features of the rotator include: (a) preservation of image counts, (b) accurate count positioning, (c) a uniform and predictable amount of blurring due to the rotation, and (d) rapid computational speed. A new rotation method was developed, entitled Gaussian rotation, in which counts in each pixel in the origin array are deposited in a Gaussian distribution centered over a fractional pixel location in the destination image. This method was compared to standard rotation techniques, and was shown to be superior in terms of the above four features. For example, bi-linear interpolation resulted in up to a 23% error in total image counts; such failure to preserve image counts is clearly unacceptable for image reconstruction. By comparison, Gaussian rotation resulted in no variation in image counts. The computational cost of Gaussian rotation was demonstrated to be only slightly more than bi-linear interpolation, and substantially less than that of bi-cubic polynomial or cubic spline interpolation. These features make Gaussian rotation preferable for iterative reconstruction.