Abstract
1678
Objectives: To improve contrast-to-noise ratio for brain imaging with constrain imposed by the total scan duration, and elimination of sampling and truncation artifacts using custom-made cone-beam collimators (CBCs).
Methods: Four different acquisition orbits have been studied for dual-headed gamma camera with two CBCs: single-circle, helical, helix-plus-arc, and CBC combined with parallel-beam collimator (PCB) in a single circular orbit. Analytical projection datasets were created for a numerical mini-Defrise phantom in a 64×64 matrix with CBC focal length = 70 cm and the radius of rotation = 32 cm. Furthermore, Tc-99m HMPAO brain SPECT scan was simulated by Monte Carlo package (SimSET) for the numerical Zubal brain phantom with the same collimators. The detector response, scatter and Poisson noise were considered in Monte Carlo simulations, but not in analytical projections. All projections were reconstructed using ordered subsets expectation maximization with a volumetric system model and exact attenuation correction (OS = 5, up to 100 iterations).
Results: Both qualitative and quantitative evaluations indicate that axial distortion artifacts are always found in reconstructed images acquired in the single-circle orbit since it cannot satisfy the Tuy’s sampling sufficiency condition. These artifacts could be suppressed by using other orbits. However, the helix-plus-arc orbit requires a large number of views ergo long imaging time and more iterations to eliminate artifacts.
Conclusions: For an object with the relatively short axial extent (such as human brain) and CBC with medium focal length (>50 cm), the improved contrast-to-noise and artifacts-free reconstructed images are obtained in SPECT with helical orbit with two CBCs or in combined CBC-PBC SPECT with a single circle orbit.
- Society of Nuclear Medicine, Inc.