Optimization technique for SPECT reconstruction

Tomographic reconstruction of images is an important problem in nuclear medicine. Most commercial systems perform the task of reconstruction using CT techniques that assume additive and independent noise. In general, these techniques do not take into account any particular model for image formation. The quality of reconstructed images is, however, strongly dependent on the models assumed during processing. These models must reflect an "a priori" information on the process of image formation. This work presents an alternative solution using a Poisson noise model that produces better results than those obtained from conventional techniques currently used in commercial systems.<>