Evaluation of two conjugate gradient based algorithms for quantitation in cardiac SPECT imaging

Correction attenuation in SPECT is a very important issue in cardiac imaging. Here, the authors evaluate two conjugate gradient (CG) based algorithms to reconstruct an attenuation compensated emission map. The first algorithm is a classic preconditioned CG approach, the second one uses the minimal residual (MR) algorithm. The authors discuss the emission reconstruction problem and the difficulty to get a good uniformity within the reconstructed cardiac wall. An evaluation on numerical and real phantoms shows that the point spread function (PSF) of the system strongly affects the uniformity. Thus, two essential improvements are made. First, the response of the reconstruction procedure is made isotropic by replacing the classic 1D apodized ramp reconstruction filter with a 2D filter. Second, to improve the spatial resolution, the authors apply a PSF deconvolution to the projections. A threshold parameter is used to prevent the deconvolution from amplifying high-frequency noise. Stabilization is also achieved by incorporating into the reconstruction algorithms a regularization, which is made spatially adaptive to preserve high contrast within the heart while smoothing the rest of the image. Results obtained on numerical and anatomic phantoms show that the proposed algorithms lead to fast, stabilized and more accurate attenuation compensation