ACCELERATION OF MAXIMUM-LIKELIHOOD RECONSTRUCTION, USING FREQUENCY AMPLIFICATION AND ATTENUATION COMPENSATION

Algorithms that calculate maximum likelihood (ML) and maximum a poster iori solutions using expectation-maximization have been successfully a pplied to SPECT and PET. These algorithms are appealing because of the ir solid theoretical basis and their guaranteed convergence. A major d rawback is the slow convergence, which results in long processing time s. This paper presents two new heuristic acceleration methods for maxi mum likelihood reconstruction of ECT images. The first method incorpor ates a frequency-dependent amplification in the calculations, to compe nsate for the low pass filtering of the backprojection operation. In t he second method, an amplification factor is incorporated that suppres ses the effect of attenuation on the updating factors. Both methods ar e compared to the one-dimensional line search method proposed by Lewit t. Ah three methods accelerate the ML algorithm. On our test images, L ewitt's method produced the strongest acceleration of the three indivi dual methods. However, the combination of the frequency amplification with the line search method results in a new algorithm with still bett er performance. Under certain conditions, an effective frequency ampli fication can be already achieved by skipping some of the calculations required for ML.