Towards quantitative SPECT: Error estimation of SPECT OSEM with 3D resolution recovery, attenuation correction and scatter correction

摘要

In this study we systematically investigate biases relevant to quantitative SPECT if OSEM with isotropic (3D) depth dependent resolution recovery (OSEM-3D), attenuation and scatter correction is used. We focus on the dependencies of activity estimation errors on the projection operator, structure size, pixel size, count density and reconstruction parameters. We use Tc-99m to establish a base line. Four Siemens low energy collimators (Low Energy Ultra High Resolution, Low Energy High Resolution, Low Energy All Purpose, Low Energy High Sensitivity) with geometric resolution between 4.4 mm and 13.1 mm at 10 cm distance and sensitivity between 100 cpm/μCi and 1020 cpm/μCi are tested with simulations of spheres with diameters between 9.8 mm and 168 mm in background. Pixel sizes and total counts are varied between 2.4 mm and 9.6 mm and 0.125 and 32 million counts. Images are reconstructed with OSEM-3D (Flash3D) with attenuation and scatter correction. Emission recovery is quantitatively measured for different reconstruction parameter settings. In addition, physical measurements of standard quality control phantoms are performed using an actual SPECT/CT system (Symbia® T6). Cross calibration of the imaging system with a well counter and results from simulations are used to quantitatively estimate the true activity concentration in the physical phantoms. Results show variations of emission recovery between 13.8% and 104.5% depending on sphere volume and number of OSEM-3D updates. After correction for the emission recovery errors and cross calibration of the imaging system the errors in absolute quantitation using the physical sphere phantom are between +0.01±0.61% for the largest (16 ml) and −5.87±1.00% for the smallest (0.5 ml) sphere. As a conclusion, the emission recovery varies over a wide range and is highly dependent on imaging parameters when using OSEM-3D reconstruction. Accurate quantitation in phantoms is possible gi--ven that errors at the specific imaging operation point can be estimated. In a clinical setup this is a nontrivial task, and perhaps too cumbersome for routine clinical use.