IMPROVED QUANTITATIVE ACCURACY OF SPECT-PET IMAGING

点击翻译按钮获取中文摘要

DESCRIPTION (Adapted from Applicant's Abstract): The recent commercial development of coincidence detection systems that can be added on to existing SPECT cameras, as well as a pharmacy network to supply FDG to hospitals which do not have a cyclotron on-site, has provided the potential for a vast increase in the number of imaging studies performed using positron emitters such as FDG. The goal of this research proposal is to investigate techniques for improving the quantitative accuracy of FDG-oncological imaging using these new coincidence detection systems. A major limitation of current gamma camera coincidence imaging system is a limited detector count-rate. An important goal of this research plan will be to investigate a number of different data collection strategies for improving the coincidence rate. The applicants proposed to evaluate and optimize these different strategies using the recently developed uniform Cramer-Rao (CR) bound on estimator variance which provides a measure of the best possible performance for a specific estimation task that is achievable using the measured projection data. To simulate various data collection strategies, they proposed to modify a Monte Carlo simulation packaged (SIMIND) which currently models SPECT imaging, to allow for the tracking of annihilation photons and their detection in coincidence using a two or three headed rotating gamma camera. In addition to investigating different data collection strategies for coincidence detection systems, they proposed to evaluate reconstruction approaches of varying computational complexity, and the extent to which compensations for attenuation, detector response and scatter can affect image quality. Two figures of merit will be used to objectively assess performance of clinically relevant estimation tasks using both simulated and clinical data. To compute these task-based figures-of-merit and to evaluate their accuracy and precision, methods to analytically compute population statistical quantities, such as the mean, covariance matrix, and the probability density function of images obtained with both non-iterative and accelerated iterative reconstruction approaches will be developed.