The influence of tomograph sensitivity on parameter estimation in small animal imaging studies

Our aim is to design a PET/SPECT scanner with sufficient sensitivity to support reconstruction of parameter estimates at high spatial resolution in small animals. Therefore, we studied the influence of tomograph sensitivity on parameter estimation at the pixel level. A rat brain section was segmented into striatum and cerebral cortex, kinetics were assigned to each region and the images forward projected. Tracers simulated included: (a) one with high striatal uptake and rapid exchange between plasma and the free compartment, and (b) one with moderate striatal uptake and slower exchange between plasma and the free compartment. Both tracers were simulated assuming a typical injected dose of 10 MBq and a reduced dose of 1 MBq, providing peak striatal uptake ranging from 0.05-1% injected dose/g. Sinograms were scaled to realistic count rates based on Monte Carlo simulation and a component-based count rate model. The EM parametric image reconstruction algorithm was used to form images of binding potential (BP) and bias and variance were calculated as a function of effective sensitivity (ES). ES of 2% produced low bias (<5%) estimates of BP for both tracers with an injectate of 10 MBq. When the injectate was reduced to 1 MBq for tracer B, an ES of 3% was required to reduce the bias below 5%. We conclude that ES of 3% is an appropriate design goal for a small animal PET scanner