Abstract
2014
Objectives DISA myocardial perfusion SPECT has a number of advantages but suffers from crosstalk contamination. We have developed methods that effectively compensate for crosstalk, but residual noise affects image quality. Thus, reducing crosstalk is essential. Here we investigate the optimization of energy windows and Tc-99m injected activity using the IO estimated using Markov Chain Monte Carlo techniques as a means to reduce crosstalk.
Methods We used a realistic digital phantom that models a medium sized male patient. We separately simulated low noise Tc-99m and Tl-201 projection data for heart, liver, lungs+remainder, allowing the use of post-simulation scaling and summing to efficiently model uptake variations. Data were simulated into 1 keV wide energy windows to allow creation of various energy windows. We simulated anterolateral and inferior 25%-contrast perfusion defects. We compared performance on a defect detection task for acquisition energy windows varying from 55-89 to 67-77 keV in 4 keV width increments. We created ensembles of 600 images with no, fixed, and reversible defects. We investigated Tc-99m sestamibi activities of 0, 6, 12, and 18 mCi and a 4 mCi injection of Tl-201. Variations in uptake in various organs were based on an ensemble of 30 patient studies. The area under the ROC curve was estimated for each energy window and injected activity.
Results The overall optimal energy window (35%) was wider than the 20% often recommended by manufacturers and resulted in statistically significant increases in the AUC (p<<1e-5). There was little change in the optimal energy window as a function of Tc activity and location; changes in corresponding AUC values were small and not statistically significant. AUC increased with decreasing Tc activity.
Conclusions The optimal energy window was wider (35%) than conventional windows. Reducing the Tc activity increased Tl defect detectability