Quantitative cardiac 15O-water PET : Assessment of left-ventricular function, remodeling, and impact of patient motion

Abstract: International guidelines advocate the use of noninvasive cardiac imaging as the initial diagnostic test for coronary artery disease, the global leading cause of death according to the world health organization. Within the wide spectrum of cardiac imaging, 15O-water PET is the gold standard for noninvasive quantification of myocardial blood flow (MBF). However, because 15O-water is a metabolically inert and freely diffusible tracer, the net retention of 15O-water in the myocardium is zero and there is no contrast between the myocardial wall and the cavity in a standard uptake image of 15O-water. The lack of contrast poses difficulties for the measurement of cardiac function and remodeling, paramount assessments for coronary artery disease evaluation along with MBF. Part one of the aim of this thesis is the development and evaluation of methods for assessment of cardiac function and remodeling in terms of left-ventricular (LV) volumes and ejection fraction (EF), LV mass (LVM), and LV wall thickness (WT). Part two is focused on patient motion, which occurs frequently in all cardiac PET studies and represents a possible source for induced error in the quantification of MBF. The feasibility of LV volumes and EF calculations was shown in paper I, where cardiac-gated parametric blood-pool images and first-pass images were imported into a commercially available software for SPECT. The method was, however, too laborious for clinical practice but served as an important proof-of-concept. In paper II, LV volumes and EF calculations were performed using first-pass images in the same software used for standard analysis of 15O-water PET and MBF assessment. The results were improved compared to paper I and the method was feasible for clinical implementation. In paper III, LVM and WT calculations were performed using segmentation of perfusable tissue fraction (PTF) images. The results showed high accuracy compared to cardiac magnetic resonance (MR) imaging, and the method was highly automated, allowing for ready clinical implementation. In papers IV-V, the impact of patient motion on the quantitative accuracy of 15O-water PET was investigated. Simulations showed a minimal impact of PET-CT misalignment on MBF, but did show that impact of dynamic motion during PET acquisition was more pronounced. Visual inspection of clinical scans showed frequent motion, but at a small amplitude with generally limited impact according to the simulations. An automated motion detection algorithm was developed which was highly accurate in detecting larger types of motion. A clear pattern of motion-induced artifacts were discovered, which may help improve their visual detection.