The quantitative accuracy of determining in vivo radionuclide activities using single photon emission computed tomography (SPECT) is influenced by the physical and computational processes involved in the acquisition and reconstruction processes. Among the most important factors are the following: 1) detection of scattered photons, 2) attenuation of primary photons, 3) finite spatial resolution and geometric collumator response with distance, and 4) inadequate utilization of the available detector area. Although first-order methods have been developed to compensate for some of these factors, these methods have not been adquately evaluated for more complex source geometries such as occur in clinical imaging. The long term objectives include the assessment of the quantitative accuracy of SPECT and the development of new and improved approaches to minimize the factors limiting SPECT quantification. Specifically, we propose to develop a unified mathematical reconstruction process that takes into consideration intrinsic inadequacies in the acquisition of SPECT projecton data. This algorithm is based on the use of Inverse Monte Carlo (IMOC) techniques to generate the probability matrix required to relate the measured projection data to the reconstructed image. General classes of nonuniformly attenuating and scattering geometries will be quantitatively characterized using accurate Monte Carlo simulations that allow for nonuniform attenuation and three dimensional scattering geometries. These simulations will be validated using carefully executed phantom experiments. The effectiveness of first- and second-order compensation methods will be determined. Improved SPECT collimator concepts including axially converging and cone beam geometries will be evaluated using computer modeling and feasibility experiments. Three dimensional reconstruction methods will be developed for the cone beam geometry. Clinical applicaitons will include evaluations of ventricular volumes using multigated SPECT blood pool images, and SPECT distributions of I-123 labeled hydroxyiodopropyldiamine (HIPDm) within the brain. The proposed investigations offer a balanced approach for long term advancements in SPECT acquisition and reconstruction methods, as well as practical improvements in quantitative clinical applications.
Showing the most recent 10 out of 84 publications
