The Thresholding MLEM Algorithm
The maximum likelihood expectation maximization (MLEM) algorithm has several advantages for image reconstruction over the conventional filtered backprojection (FBP). However, the slow convergence rate and the high computation cost for a practical implementation have impeded its clinical applications. In this study, we propose the incorporation of thresholding technique into MLEM to speed up the convergence rate. Owing to the fact that the reconstruction time is proportional to the total number of pixels (voxels), the thresholding technique that nullifies the value of a pixel if it falls below a threshold, can effectively remove the non-active pixels and significantly speed up the reconstruction. Preliminary tests on simulated PET data show that the thresholding technique speeds up the convergence rate and reduce error in the reconstructed image.