Quantitative single photon emission compute tomography (SPECT) has been limited mainly by the inadequate number of detected photons in the projections, attenuation within the body, the inclusion of scattered photons in the projections, and variations in collimator response with distance. The long term objectives of this project are: 1) to increase the number of detected photons by using new acquisition geometries, and 2) to develop and evaluate a unified reconstruction methodology for SPECT using a statistical Bayesian formulation. Cone beam collimation (CB) can result in a three-fold increase in the number of detected photons, compared with a parallel beam (PB) collimator having similar resolution characteristics. Bayesian image processing (BIP) can integrally include the a priori source distribution entropy information, the Poisson nature of the measured projection data, as well as other a priori information concerning the shift-variant point spread function. Filtered backprojection (FBP) is widely used since it offers the practical advantage of fast computational speed; however, BIP offers the potential for a more accurate reconstruction. Other statistical estimators, such as maximum likelihood (ML) and maximum entropy (MAXENT) are directly derivable from the Bayesian formulation. This project will evaluate cone beam SPECT using FBP and BIP approaches. A validated Monte Carlo model of the acquisition geometry will be used to determine the system response matrix. Receiver operator characteristic (ROC) curves will be used to evaluate lesion detectability for CB, fan beam (FB), and PB SPECT. The convergence characteristics of BIP will be quantitatively compared with ML and MAXENT, as a function of iteration number. Simulated and experimentally acquired projections of phantoms will be used to evaluate BIP for quantification of activity, activity ratios, resolution, and noise cone beam SPECT. Patient scans acquired with CB, FB, and parallel hole collimation will be compared. Early results indicate that cone beam collimation offers the potential for improved SPECT imaging. The application of a unified BIP algorithm that simultaneously compensates for attenuation, scatter, and collimator variations should offer further improvements in the accuracy of the reconstruction, and provide the basis for fundamental advances in SPECT quantification.