Emission-Based Scatter Correction in SPECT Imaging

Scatter correction in single photon emission computed tomography (SPECT) has been focused on either using multiple-window acquisition technique or the scatter modeling technique in iterative image reconstruction. We propose a technique that uses only the emission data for scatter correction in SPECT. We assume that the scatter data can be approximated by convolving the primary data with a scatter kernel followed by the normalization using the scatter-to-primary ratio (SPR). Since the emission data is the superposition of the primary data and the scatter data, the convolution normalization process approximately results in the sum of the scatter data and a convolved version of scatter data with the kernel. By applying a proper scaling factor, we can make the estimation approximately equal to or less than the scatter data anywhere in the projection domain. Phantom and patient cardiac SPECT studies show that using the proposed emission-based scatter estimation can effectively reduce the scatter-introduced background in the reconstructed images. And additionally, the computational time for scatter correction is negligible as compared to no scatter correction in iterative image reconstruction.