Image quality for advanced iterative reconstruction schemes with SPECT/CT

Abstract

Background Advanced iterative reconstruction methods are now widely used for SPECT-CT. The aim of this study was to evaluate the gain in image quality for clinical settings currently used for brain and body SPECT studies. Material and methods Data from the IEC image quality head and body phantoms were obtained with different SPECT/CT systems. Spheres (inner diameters varying from 10 to 37 mm) were filled with Tc-99 m for spheres to background contrasts of 4:1 and 6:1 and the ‘body’ filled with 150 MBq. Camera settings were those used routinely for brain and body SPECT studies except pixel size that was adjusted as close as 3 mm. SPECT data were reconstructed with clinical parameters for OSEM (number of iterations and subsets, post-filter) using CT-based attenuation correction and scatter compensation based on multiple windowing estimates and nonuniform collimator response compensation for recovering isotropic resolution. Image quality was assessed following NEMA NU2-2007 definitions for hot spheres percent contrasts and background variability for the 37 mm sphere for 24 background ROIS (8 ROIs on three slices). Results Visual analysis exhibited that all spheres were detected in brain studies whereas in body studies, in one case the smaller sphere was not visible for the 4:1contrast. Spatial distorsions due to incomplete collimator response correction were present in all body transaxial images. Image noise was substantially decreased with resolution recovery correction. Despite different reconstruction parameters depending on manufacturers recommendations, hot sphere contrasts were quite similar from one system to the other and exhibited strong variations with sphere size (from 10% for 10 mm to 66% for 37 mm). Contrast recovery was mainly affected by post filtering. Conclusion We proposed an image quality test for evaluation of image quality and contrast recovery in SPECT-CT. Advanced iterative reconstructions including collimator modelling demonstrated important spatial distorsions for body like studies and limitations of contrast recovery.