Practical iterative reconstruction methods for quantitative cardiac SPECT image reconstruction

We proposed a practical iterative reconstruction method for cardiac SPECT imaging. The method consisted of four major parts. First, the first order Chang reconstructed image with attenuation compensation for the chest region was obtained as the initial estimate. The iterative weighted least square-conjugate gradient (WLS-CG) reconstruction algorithm, known for its fast convergence rate, was then applied. A projector/backprojector which modeled the nonuniform attenuation and the spatially variant collimator-detector response was used with the WLS-CG algorithm to provide accurate compensation for the effects of attenuation and collimator-detector response. Finally a smoothing filter was used to suppress noise amplification between iterations. The method was evaluated relative to conventional reconstruction methods in terms of quantitative accuracy, image noise and ventricular wall thickness. We conclude that the practical iterative reconstruction method has great potential for clinical use in providing cardiac SPECT images with high quantitative accuracy and low image noise within a few iterations.< >