Clinical evaluation of iterative reconstruction (ordered-subset expectation maximization) in dynamic positron emission tomography: Quantitative effects on kinetic modeling with N-13 ammonia in healthy subjects

Abstract

The purpose of this study was to investigate the quantitative properties of ordered-subset expectation maximization (OSEM) on kinetic modeling with nitrogen 13 ammonia compared with filtered backprojection (FBP) in healthy subjects. Cardiac N-13 ammonia positron emission tomography (PET) studies from 20 normal volunteers at rest and during dipyridamole stimulation were analyzed. Image data were reconstructed with either FBP or OSEM. FBP- and OSEM-derived input functions and tissue curves were compared together with the myocardial blood flow and spillover values. The late area under the OSEM input functions during dipyridamole is overestimated by 30% (P < .0001) relative to FBP. Conversely, the area under the late part of the OSEM tissue curves is underestimated by 20% (P < .0001) compared with FBP during both rest and dipyridamole. These differences in tissue and input functions cause the resting myocardial blood flow to be underestimated by 15% (P < .0001). During dipyridamole, the OSEM flow is underestimated by 25% (P < .0001) relative to FBP, causing the myocardial flow reserve to be underestimated by 10% (P < .0001). Large inter-regional differences in FBP and OSEM flow values were observed with a flow underestimation of 45% (rest/dipyridamole) in the septum and of 5% (rest) and 15% (dipyridamole) in the lateral myocardial wall. OSEM reconstruction of myocardial perfusion images with N-13 ammonia and PET produces high-quality images for visual interpretation. However, compared with FBP, OSEM is associated with substantial underestimation of perfusion on quantitative imaging. Our findings indicate that OSEM should be used with precaution in clinical PET studies.