Variance prediction in SPECT reconstruction based on the Fisher information using a novel angular blurring algorithm for computation of the system matrix

In this paper we propose a Gaussian decomposition approach to compute the SPECT imaging system matrix. The flexibility of our method allows the implementation of a wide range of imaging systems. With the system matrix, we predict the variance in reconstructed images using the Fisher information matrix and local block circulant approximation. We present results of noise prediction for 3 multi-pinhole systems and 3 multi-slit slit-slat systems, each of them being designed to be inserted in an MRI system. Results show that for a particular phantom 2×2 multi-pinhole and 2 slits slit-slat systems achieve lowest variance.