An intrinsic 3D Wiener filter for the deconvolution of spatially varying collimator blur

The authors have developed distance-dependent deconvolution of collimator blur for 3D SPECT reconstruction. The collimator model is based on the frequency-distance principle (FDP), and the authors have applied it to the circular harmonic transform (CHT) SPECT reconstruction algorithm. The filter model used is a 3D Wiener (FT stochastic) filter. Since the filter is applied between the premultiply and backprojection phases of reconstruction, it is an intrinsic filter, rather than a pre or post filter. The authors compare the new 3D filter to (i) the 3D post filter followed by attenuation correction as implemented in a commercial nuclear medical imaging system, (ii) the 3D collimator model and a standard Wiener filter formulation, and (iii) the “gold standard” or quantitative CHT SPECT protocol that utilizes a 2D Wiener prefilter. In all cases, the new 3D filter yielded the best images with minimal noise amplification