A convolution-subtraction technique to compensate for downscatter effects in simultaneous Tl-201/Tc-99m cardiac SPECT

One of the major concerns in simultaneous Tl-201/Tc-99m cardiac SPECT is the downscatter counts acquired in the 70/spl plusmn/10% keV window, contributed by the 140 keV Tc-99m photons. In this study, a new method was proposed to remove the downscatter counts. The downscatter distribution was obtained by convolving the 140 keV reconstructed image with a spatially variant 2-D scatter response function (SRF), and was then subtracted from the 70 keV projection data. The SRF was assumed to be a combination of either two Lorentz functions or two Gaussian functions. Their amplitudes and widths were determined by fitting the Monte Carlo (MC) simulated downscatter distributions. Furthermore, the amplitudes were assumed to be exponential functions and the widths to be linear functions of the depth of the source. The derived SRF depends on the shape of the object relative to the detector, so errors may occur when a SRF derived for an object at a particular view is applied to other objects or to other views even with the same object. To reduce the errors, a simple correction method was proposed. As a result, the downscatter distribution of a Tc-99m point source obtained using the derived SRF agreed reasonably with the MC results for the same object at all the views. The agreement becomes worse as the detector rotates away from the view where the SRF is derived. When applying to the phantom study, the quality of the dual-isotope 70 keV images is improved, resulting in a comparable accuracy as that of the Tl-201 only 70 keV images.