摘要：AbstractThe goal in dynamic SPECT studies is to estimate the time-activity functions describing the tracer activity changes in the observed tissue. Dynamic studies can be created as a series of static projection images recorded in several subsequent time frames. By applying a static reconstruction algorithm for each images independently and fitting the parameters of a kinetic model to the activity on the reconstructed images we can estimate the function of the tracer activity changes. However, due to the limited data collection time images in dynamic studies often suffer from poor statistics influencing the accuracy of the kinetic function estimation. According to previous studies improved function reconstruction can be achieved by applying reconstruction algorithms that estimate kinetic parameters directly from the measured projection data. In this study our previously developed Monte Carlo based maximum likelihood expectation maximization (ML-EM) SPECT reconstruction method - originally verified for static SPECT studies - is extended for dynamic studies. An iterative dynamic reconstruction algorithm is developed that estimates directly the temporal function of the tracer activity in the image voxels in each iterations and uses the static ML-EM iteration afterwards. The method of sieves has been applied for regularization of the estimated kinetic parameters. The estimation accuracy of the suggested dynamic reconstruction algorithm was evaluated in a mathematical phantom study with different time functions.
关键词：Keywordsdynamic SPECTdirect reconstructionmethod of sievescompartmental models
