Skip to Main Content
The expectation-maximization maximum-likelihood (EM-ML) algorithm belongs to a family of algorithms that compute positron emission tomography (PET) reconstructions by iteratively solving a large linear system of equations. The authors describe a preprocessing scheme for automatically focusing the attention, and thus the computational resources, on a subset of the equations and unknowns. Experimental work with a CM-5 parallel computer implementation using a simulated phantom as well as real data obtained from an ECAT 921 PET scanner indicates that quite significant savings can be obtained with respect to both time and space requirements of the EM-ML algorithm without compromising the quality of the reconstructed images.
Date of Publication: April 1997