Scatter correction in single-photon imaging using multiple energy bins and minimum-risk estimators

We are investigating methods of scatter correction that use multiple (typically, 6-15) energy bins for scatter correction in single photon imaging. We have previously reported preliminary results of this work with simple phantoms. In this article, we present new experimental results using the Rollo phantom, a complex phantom designed for evaluation of quantitative imaging techniques. We have refined our correction techniques by allowing the use of Poisson weighting and by developing a technique for progressive basis function refinement. We have studied a new class of methods, the 'minimum risk' estimators, which are specifically designed to minimize the mean-square error (MSE) of primary and scatter estimates (much like the goal of the Wiener filter, but in the energy domain). The performance of minimum-risk estimators and our earlier constrained minimization techniques both improve on that of simple deconvolution, particularly with respect to noise sensitivity. We have applied these techniques to the removal of scatter from images of the Rollo phantom filled with Tc-99m, I-131, and Tl-201; studies with variable-depth scatterers and other isotopes are currently underway.