论文部分内容阅读
Purpose:Low count PET studies suffer from quantitative bias in iterative reconstruction methods such as MLEM or OSEM.This work aims at reducing PET image bias and enhancing lesion detectability particularly in low activity regions.Methods:The CT-guided Negative Maximum Likelihood(NML)algorithm is developed.Firstly,The CT image is used to segment low density regions such as lungs and stomach.These regions are assumed to have low activity in PET scan.The NML algorithm takes this segmentation results as prior information and incorporates it in the bias-removal iterative reconstruction algorithm.Phantom studies and clinical data are tested in the experiments for this algorithm.Results:The convergence rate in the low activity regions is significantly enhanced.The positive bias in low activity regions in the PET image such as the lungs and stomach is reduced by as much as 10~20%,under the same number of reconstruction iterations.The contrast recovery of small hot structures is enhanced by >5%in the study.Phantom studies reveals the same observation.Conclusions:The CT-guided NML algorithm reduces positive bias in low activity regions in PET images.It also provides better contrast recovery for small structures.This CT-guided NML algorithm could assist doctors in more informative diagnosis particularly in low count PET studies.