The effect of intrinsic spatial resolution on the quantitative accuracy of SPECT imaging

Due to the advent of 511 keV coincidence detection for sodium iodine based gamma cameras, there has been a trend towards using increasing crystal thickness. Since these cameras typically provide the capability for imaging both low-energy single photon emitters and positron emitters, the choice of optimal crystal thickness is unclear. In this paper, we measure how the change in intrinsic spatial resolution R_i, resulting from an increased crystal thickness would affect quantitative accuracy in SPECT imaging. In order to objectively assess quantitative accuracy achieved with cameras of differing R_i , we evaluated performance of a multi-parameter estimation task; estimating the amplitude and size of a small Gaussian function embedded within two different, realistic anthropomorphic phantoms. The fundamental performance of this task was evaluated using the Cramer-Rao bound on unbiased estimates of the signal parameters. Results suggest that when the imaging agent is Tc^99m, the difference in quantitative accuracy between cameras with R_i of 3.5 mm and 5.5 mm is very small when using a low-energy high-resolution collimator with a camera radius of rotation typical of chest imaging. A larger difference is observed when using a low-energy ultra-high-resolution collimator and a camera radius of rotation typically used in brain imaging. For higher energy single photon emitters such as In¹¹¹ and I¹³¹, the advantage of increased detection efficiency with thicker crystals far outweighs the loss of estimator performance resulting with increased R_i