Comparison of two simulation approaches for a first-generation transmission type gamma computed tomography

Abstract

A first-generation transmission type computed tomography setup was simulated using two different approaches, each corresponding to a different rotation method used in actual experimental setups. Two digital phantoms were used, one of which is the Shepp-Logan phantom, a standard phantom used to test image reconstruction algorithms. The sinogram data from the simulations were compared by plotting the normalized absolute difference in a colormap. It was observed that regions near the rotation angles 0, 90, and 180 degrees as well as regions swept by edges of shapes from the original phantoms exhibit high degrees of difference. In actual ideal experiments, the sinogram data from the two different rotation methods are expected to be the same. The discrepancy between the simulation data for the two different simulation approaches was pinpointed to be due to the image rotation function used in one of the methods which fails to preserve the pixel values of the digital phantoms upon rotation. This is undesirable because it physically translates to a change in the material's attenuation coefficients in those regions which does not happen in an actual setup. The simulation approach where the phantom is not rotated is preferable since the pixel values are guaranteed to be unchanged. Differences in the reconstructed images arising from the difference in simulation approaches are yet to be explored.