An Investigation of Markov Random Fields for Bayesian Reconstruction of Single Photon Emission Computed Tomography

This research investigates the use of Markov random fields for Bayesian reconstruction algorithms to be used with high-resolution and high-sensitivity SPECT systems for small animal imaging. It extends previous research on mechanical models for Bayesian image reconstruction by using a three-dimensional nonconforming finite element model and linear elasticity concepts to derive minimum potential energy functionals which regularize the reconstruction process. It combines dual collimator SPECT projection data by using high-resolution data to penalize lower-resolution data. It compares the new three-dimensional penalized reconstruction technique with existing penalized techniques through the use of modulation transfer and contrast discrimination functions