Investigation of methods to correct for distance-dependent resolution in single-photon-emission computed tomography imaging

In single photon emission computed tomography there are number of limitations that effect the quality of images. The distance dependent spatial resolution is one of the limitations. It is a direct consequence of non-zero range of directions of the arriving photons.; One method that can be employed to compensate for distance dependent resolution is restoration filtering based on frequency distance relationship. Another method that can be employed to correct for spatially varying resolution is to incorporate the blurring model in the projector/backprojector pair of the iterative algorithm.; In this thesis a systematic comparison between both the methods of correction for distance dependent resolution is given on the basis of extent of resolution recovery vs noise tradeoffs and processing times. The source distribution consisted of 3D Gaussian point sources embedded in a flat background (which was 10% of the maximum counts in the Gaussian sources).; Simulations indicate that the extent of resolution recovery of a source distribution in the presence of the background activity gets slowed down. A new algorithm (AB-EMML) was implemented for the acceleration of the "low" contrast point source distribution. Again the source distribution consisted of the 3D Gaussian point sources embedded in the flat background which was 150% of the maximum counts the Gaussian sources. Comparison between various reconstruction strategies was made in terms of the extent of resolution recovery as a function of iterations.; Further, systematic comparison for both the methods of correcting for DDR was given for an extended source distribution as obtained from the mathematic cardiac torso phantom. It models the organs in the thorax region realistically. The comparison was made on the basis of the uniformity of the heart wall counts and the average wall thickness.; The DDR correction by modeling the blurring model gave slightly better image metrics than the restoration filtering for both the source distributions. Also, the AB-EMML algorithm can speed up resolution recovery for "low" contrast source with properly chosen bounds.