Monte Carlo-based compensation for patient scatter, detector scatter, and crosstalk contamination in In-111 SPECT imaging

We have incorporated Monte Carlo (MC)-based estimates of patient scatter, detector scatter, and crosstalk into an iterative reconstruction algorithm, and compared its performance to that of a general spectral (GS) approach. We extended the MC-based reconstruction algorithm of de Jong et al. by (1) using the “Delta scattering” method to determine photon interaction points, (2) simulating scatter maps for many energy bins simultaneously, and (3) decoupling the simulation of the object and detector by using pre-stored point spread functions (PSF) that included all collimator and detector effects. A numerical phantom was derived from a segmented CT scan of a torso phantom. The relative values of In-111 activity concentration simulated in soft tissue, liver, spine, left lung, right lung, and five spherical tumors (1.3–2.0 cm diam.) were 1.0, 1.5, 1.5, 0.3, 0.5, and 10.0, respectively. GS scatter projections were incorporated additively in an OSEM reconstruction (6 subsets×10 projections×2 photopeak windows). After three iterations, GS scatter projections were replaced by MC-estimated scatter projections for two additional iterations. MC-based compensation was quantitatively compared to GS-based compensation after five iterations. The bias of organ activity estimates ranged from −13% to −6.5% (GS), and from −1.4% to +5.0% (MC); tumor bias ranged from −20.0% to +10.0% for GS (mean±std.dev.=−4.3±11.9%), and from −2.2 to +18.8% for MC (+4.1±8.6%). Image noise in all organs was less with MC than with GS.