Treatment of Compton scattering in maximum-likelihood, expectation-maximization reconstructions of SPECT images

This paper studies the extent to which lesion contrasts and signal-to-noise ratios in maximum-likelihood, expectation-maximization estimates of SPECT images can be improved by considering Compton scattering when calculating the photon detection probability matrix. Matrices are generated using a Monte Carlo code that realistically models a SPECT imaging system. For cold lesions having true contrasts of one, thirty-six combinations of three lesion sizes, three lesion locations, and four projection-data count levels are considered. Both hot and cold, 2-cm-radius lesions having various values of true contrast are studied at a few count levels and lesion locations. With scatter in the matrix, scatter effects are completely removed from the central regions of lesions whose radius is greater than the full width at half maximum of the spatial resolution. Resolution recovery is initially slower but eventually superior. Percent improvements in contrast are greater for lower-contrast cold lesions and higher-contrast hot lesions. In most cases, signal-to-noise ratios are comparable or better. Correct absolute activity levels are obtained.