Monte Carlo-based down-scatter correction of SPECT attenuation maps

Combined acquisition of transmission and emission data in single-photon emission computed tomography (SPECT) can be used for correction of non-uniform photon attenuation. However, down-scatter from a higher energy isotope (e.g. Tc-99m) contaminates lower energy transmission data (e.g. Gd-153, 100 keV), resulting in underestimation of reconstructed attenuation coefficients. Window-based corrections are often not very accurate and increase noise in attenuation maps. We have developed a new correction scheme. It uses accurate scatter modelling to avoid noise amplification and does not require additional energy windows. The correction works as follows: Initially, an approximate attenuation map is reconstructed using down-scatter contaminated transmission data (step 1). An emission map is reconstructed based on the contaminated attenuation map (step 2). Based on this approximate Tc-99m reconstruction and attenuation map, down-scatter in the 153Gd window is simulated using accelerated Monte Carlo simulation (step 3). This down-scatter estimate is used during reconstruction of a corrected attenuation map (step 4). Based on the corrected attenuation map, an improved Tc-99m image is reconstructed (step.5). Steps 3-5 are repeated to incrementally improve the down-scatter estimate. The Monte Carlo simulator provides accurate down-scatter estimation with significantly less noise than down-scatter estimates acquired in an additional window. Errors in the reconstructed attenuation coefficients are reduced from ca. 40% to less than 5%. Furthermore, artefacts in Tc-99m emission reconstructions are almost completely removed. These results are better than for window-based correction, both in simulation experiments and in physical phantom experiments. Monte Carlo down-scatter simulation in concert with statistical reconstruction provides accurate down-scatter correction of attenuation maps.