The ability to measure metabolic activity, as well as the ability to discriminate among different tissue types, make SPECT a potentially medical imaging tool. However, the low resolution and poor image quality that characterize SPECT images limit the clinical usefulness of SPECT. Mathematical Technologies Inc. believes that mathematical methods are available which will significantly improve SPECT reconstruction techniques. The principle investigator has developed an iterative reconstruction technique based on Markov random field image models in a Bayesian framework. Experiments with this technique indicate that dramatic improvements in SPECT image quality may be possible. The proposed research would address the issues relevant to the clinical viability of the approach: 1. the imaging characteristics of a state of the art machine are to be modeled; 2. Bayesian reconstructions are to be systematically compared to the best available reconstruction techniques; 3. the feasibility of optimizing and/or modifying the computational algorithms to produce the reconstruction within clinical time limits will be explored; 4. the possibility of exploiting this framework to obtain simultaneous estimates of attenuation and isotope concentration will be explored. (Current methods must assume a priori the attenuation structure). If successful, this fourth research objective would result in a significant advance in the SPECT technology.
