Bayesian image reconstruction in SPECT: correction for photon attenuation and spatial resolution

A novel method based on a Bayesian model has been developed for SPECT (single photon emission computed tomography) image reconstruction. This model uses a Poisson noise model with a priori information, introduces line sites with continuous values, and incorporates the effects of depth-dependent photon attenuation and point spread function to improve the image quality. Images reconstructed by means of this extended Bayesian approach exhibit improved signal-to-noise ratio and better reconstructed resolution than those produced by other methods. In general, the extended Bayesian approach improves the residual root mean square errors and the image contrast values of the image by about 10 to 40% over those obtained with other algorithms