Space-Alternating Generalized EM (SAGE) algorithm with attenuation correction but without detector-response modeling

We calibrated a scintillation camera for the counts-to-activity conversion factor, CF, by measuring a phantom consisting of a sphere containing a known 131-I activity placed within an elliptical cylinder. Within a 3D OSEM reconstruction algorithm, we employed a depth-dependent detector-response model based on smooth fits to the pointsource-response function. Using the ultra-high-energy (UHE) collimator and 100 iterations, the recovery coefficient, RC, appeared to be 1 for any sphere volume down to 20 cm3. The CF changed only a small amount as the background-over-target activity concentration ratio, b, increased for both UHE and highenergy (HE) collimation. Tests of activity quantification were carried out with an anthropomorphic phantom simulating a 100cm3 spherical tumor centrally located inferior to the lungs. With 3D OSEM reconstruction, using the global-average CF and no RC-based correction, mean bias in the simulated-tumor activity estimate over 20 realizations was -7.4% with UHE collimation, and –9.4% with HE collimation. For comparison, with 1D SAGE reconstruction, using the CF corresponding to the experimental estimate of b and RC-based correction, the mean bias was worse, -10.7% for UHE collimation, but better, -4.3%, for HE collimation.