Impact of attenuation and scatter correction in SPECT for quantification of cerebral blood flow using 99mTc-ethyl cysteinate dimer

Attenuation and scatter of photons are the main causes that binder the quantification of regional cerebral blood flow (rCBF) value by single photon emission computed tomography (SPECT), using Tc-99m-Ethyl cysteinate dimer (ECD). We investigated the effects of attenuation correction and scatter correction on rCBF values with Tc-99m-ECD SPECT. In particular, the applicability of uniform attenuation maps (mu maps) was evaluated in terms of errors on the estimated CBF values and the optimal threshold levels for extracting brain contours. SPECT scans were performed on seven subjects, in the presence of Tc-99m-ECD. Quantitative K-1 images were computed using the reconstructed images and the input function obtained with the frequent arterial blood sampling method. The images were reconstructed by the ordered subset expectation maximization (OSEM) reconstruction in which uniform and segmented mu maps were used for attenuation correction with and without scatter correction. The transmission-dependent convolution subtraction technique was utilized for scatter correction. Segmented and uniform mu maps were generated from magnetic resonance (MR) images. We also produced uniform mu maps using ECD images obtained at various threshold levels and mu values (0.11, 0.15, and 0.172 cm(-1)). Scatter correction improved the image contrast dramatically. There were no significant differences between K-1 images with attenuation and scatter corrections assuming a uniform mu map (not 0.15 but 0.172 cm(-1)) and those corrected with segmented mu maps for most regions. However, in the former images, values were overestimated for deep structures (e.g., overestimation of 9.5% in the striatum and 7.3% in the central semi oval). This small but significant error was also observed in phantom studies and Monte Carlo simulations. We show that the overestimation using uniform mu maps is due to the weight of the path length in the bone. Absolute K-1 values were sensitive to the threshold level when the edge of the brain was determined from the ECD images, but the variation of the estimated K-1 was +/-9.0% when the optimal threshold level was selected. This study suggests that the use of uniform attenuation mu maps provides reasonable accuracy, despite a small but significant error in deep structure regions, and that uniform mu maps may be provided from the emission data alone in this patient population.