Quantitative simultaneous cardiac SPECT using MC‐JOSEM

Abstract

Simultaneous rest cardiac SPECT imaging has the potential to replace current clinical ‐Sestamibi rest/stress imaging and therefore has great potential in the case of patients with chest pain presenting to the emergency department. Separation of images of these two radionuclides is difficult, however, because their emission energies are close. The authors previously developed a fast Monte Carlo (MC)‐based joint ordered‐subset expectation maximization (JOSEM) iterative reconstruction algorithm (MC‐JOSEM), which simultaneously compensates for scatter and cross talk as well as detector response within the reconstruction algorithm. In this work, the authors evaluated the performance of MC‐JOSEM in a realistic population of studies using cardiac phantom data on a Siemens e.cam system using a standard cardiac protocol. The authors also compared the performance of MC‐JOSEM for estimation tasks to that of two other methods: standard OSEM using photopeak energy windows without scatter correction (NSC‐OSEM) and standard OSEM using a Compton‐scatter energy window for scatter correction (SC‐OSEM). For each radionuclide the authors separately acquired high‐count projections of radioactivity in the myocardium wall, liver, and soft tissue background compartments of a water‐filled torso phantom, and they generated synthetic projections of various dual‐radionuclide activity distributions. Images of different combinations of myocardium wall/background activity concentration ratios for each radionuclide were reconstructed by NSC‐OSEM, SC‐OSEM, and MC‐JOSEM. For activity estimation in the myocardium wall, MC‐JOSEM always produced the best relative bias and relative standard deviation compared with NSC‐OSEM and SC‐OSEM for all the activity combinations. On average, the relative biases after 100 iterations were 8.1% for and 3.7% for with MC‐JOSEM, 39.4% for and 23.7% for with NSC‐OSEM, and 20.9% for with SC‐OSEM. The relative standard deviations after 30 iterations were 0.7% for and 1.0% for with MC‐JOSEM, as compared to 1.1% for and 1.2% for with NSC‐OSEM and 1.3% for with SC‐OSEM. Finally, the authors compared the relative standard deviation after 30 iterations with the minimum theoretical variance on activity estimation, the Cramer–Rao lower bound (CRB), and with the biased CRB. The measured precision was larger than the biased bound values by factors of 2–4, suggesting that further improvement could be made to the method.