Effect of including detector response in SPECT quantification of focal I-131 therapy

With a regularized strip-integral (1D) SAGE reconstruction, circular-orbit SPECT estimates of phantom focal 131-I activity vary with changes in the level of uniform background. They also vary with changes in image resolution due to different settings of the radius of rotation. To solve these problems, we investigated the effect of employing two different depth-dependent detector-response models. A regularized plane-by-plane (2D) SAGE algorithm reduced dependence of the counts-to-activity conversion factor on relative background concentration by 37% compared to the 1D SAGE. With unregularized multi-plane (3D) OSEM reconstruction, initial results showed: 1) a conversion factor that was independent of relative background concentration, and 2) a recovery coefficient that was approximately 1 for any sphere volume down to 20cc. We conclude that using a 3D detector-response model has the potential to eliminate bias problems. For a patient, the preliminary activity-estimate changes using 3D OSEM compared to 1D SAGE were: 1) +16% for a large tumor, and 2) -35% for a small tumor for which recovery-coefficient-based-correction-factor errors can be large.