Evaluation of three scatter correction methods based on estimation of photopeak scatter spectrum in SPECT imaging: A simulation study

• Two older scatter correction approximations are compared with a new purposed approximation.
• Four different simulated activity distributions are used for evaluation of the three approximations.
• The new approximation offers a better estimation of the area of the photopeak scatter spectrum.
• The new approximation improves the image contrast as compared to the two other methods.
• The purposed approximation increases slightly the image noise.

Three practical methods for scatter correction of Tc-99m SPECT images are evaluated. Among these, two methods, three-energy window (TEW) methods using the trapezoidal and triangular approximations, have been described previously by investigators, and a new approximation is offered in this work. The SIMIND (SIMulation of Imaging Nuclear Detectors) Monte Carlo program is used to simulate a line source placed at on-axis and 5 cm off-axis locations, a cold-sphere/hot-background phantom, a hot-sphere/cold-background phantom, and a more clinically realistic NCAT (Nonuniform Rational B-spline-based CArdiac-Torso) phantom. For evaluation of these methods, the scatter line-spread functions and scatter fractions for the on- and off-axis line source, image contrast, signal-to-noise ratio and relative noise for the cold spheres, and recovery coefficient for the hot spheres of different diameters are compared. For the NCAT phantom, a line profile through a slice of the reconstructed image is considered before and after scatter correction, and also image contrast defined by this profile is used to compare the correction methods. The results of this study indicate that for the line source simulation the scatter fractions obtained from the proposed method are a better estimation of true scatter fractions. Also, for both the sphere simulation and NCAT simulation, the proposed method improves the image contrast as compared to the two other methods.