Algorithms for parameter estimation in dynamic tracer studies using positron-emission tomography

Several algorithms for estimating parameters in dynamic tracer studies using positron-emission tomography (PET) were developed and evaluated in this research. The algorithms are all based on a previously developed method that uses the EM algorithm to compute maximum-likelihood estimates of physiological parameters from list-mode data. This algorithm is extended to include the effects of attenuation, accidental coincidences, and non-impulsive input functions. Furthermore, variants are developed that can estimate histograms of activity levels vs. time for use with weighted least squares and that can utilize projection data as well as list-mode data. The algorithms can be decomposed into two parts, one of which estimates statistics for the list-mode or projection data; and one which estimates parameters from these statistics. The former algorithm computes statistics with temporal resolution that is superior to that of the most commonly used approach. The evaluation shows that parameters estimated using this algorithm are significantly more accurate than those found using the standard approach. Evaluations of the parameter-estimation algorithms show that they posses superior convergence characteristics and that they compute more accurate estimates than the most commonly used method of weighted least squares