Evaluation of projection pixel-dependent and pixel-independent scatter correction in SPECT

Dual energy window scatter subtraction methods in single photon emission computed tomography (SPECT) to compensate for the detection of Compton-scattered photons are based on the assumption that the number of scattered photons detected by each projection pixel in the primary (photopeak) energy window is proportional to the number of photons detected in a secondary (scatter) energy window. The reliability of this assumption is tested using Monte Carlo modeling of photon detection kernels for photon emission by Tc-99 m within a water-filled cylinder. The primary window scatter kernels and the secondary window kernels are correlated highly, and projection pixel-dependent and pixel-independent scatter correction factors are obtained from least squares fits of these kernels to each other. In a numerical experiment with noise-free projection data, average reconstructed intensities in regions of interest are nearly as well-estimated for both pixel-independent and pixel-dependent scatter-corrected projection data reconstructed using the approximate generalized matrix inverse (GMI) of the nonscatter photon detection matrix as for primary window projection data reconstructed with the approximate GMI of the photon matrix which includes scatter.<>