Title :
Study of the convergence properties of the EM algorithm-a new stopping rule
Author :
Kontaxakis, George ; Tzanakos, George
Author_Institution :
Dept. of Biomed. Eng., Piscataway, NJ, USA
Abstract :
The authors have studied convergence properties of the expectation-maximization (EM) algorithm for image reconstruction in positron emission tomography (PET) using simulated phantoms. The problem of the deterioration of the image quality after a certain number of iterations has been addressed. The feasibility criterion as a stopping rule has been studied, and it was found that the number of the measured counts and the structure of the image pattern have a considerable effect on the feasibility of the reconstructed images. A new stopping rule of the iterative procedure is proposed, based on the statistical properties of the updating coefficient for each image pixel. Preliminary results on this stopping rule from simulation studies are presented
Keywords :
computerised tomography; image reconstruction; medical image processing; radioisotope scanning and imaging; convergence properties; expectation maximization algorithm; feasibility criterion; image pattern; image pixel; image reconstruction; iterative procedure; measured counts; positron emission tomography; reconstructed images; simulated phantoms; simulation studies; statistical properties; stopping rule; structure; updating coefficient; Computational modeling; Convergence; Detectors; Image quality; Image reconstruction; Image storage; Iterative algorithms; Pixel; Positron emission tomography; Testing;
Conference_Titel :
Nuclear Science Symposium and Medical Imaging Conference, 1992., Conference Record of the 1992 IEEE
Conference_Location :
Orlando, FL
Print_ISBN :
0-7803-0884-0
DOI :
10.1109/NSSMIC.1992.301047
