• Michael D R Thomas, Dale L Bailey, and Lefteris Livieratos.
• Joint Department of Physics, Institute of Cancer Research, Royal Marsden Hospital, Sutton, UK.
• Phys Med Biol. 2005 Aug 7; 50 (15): N187-94.

AbstractRoutine quality control (QC) and optimization of image quality of reconstructed images in single photon emission computed tomography (SPECT) and positron emission tomography (PET) remains a relatively qualitative exercise. With the advent of combined SPECT/CT and PET/CT devices, and accurate post hoc co-registration algorithms, the potential exists to utilize high resolution structural information for QC evaluation in addition to their use for anatomical correlation in clinical studies. The aim of this work was to explore, in principle, the uses of x-ray CT data of QC phantoms used in SPECT and PET to develop more objective assessments of performance of the emission tomographic (ET) devices and reconstructed data. A CT reconstruction of a novel ET QC phantom was segmented into the various compartments it contained. Using software, the voxel values in the different compartments were then altered to correspond to the concentration of the radioactivity in the actual scan of the same phantom on the SPECT system. This produces a high resolution version of a 'perfect' ET scan. Image co-registration techniques were then used to spatially align the synthetic high resolution SPECT scan to the measured SPECT scan. Various parameters can then be objectively derived from the registered data, for example, image contrast, spatial resolution, spatial non-uniformity, etc. In this study, we have used this approach to estimate spatial resolution (full width at half maximum, FWHM) and recovered contrast in reconstructed images of a SPECT phantom. Two independent methods were used to measure spatial resolution, obtaining excellent agreement. In conclusion, the ability to produce high resolution synthetic phantoms in emission tomography QC affords an objective approach to assessing system performance and optimizing protocols which is readily automated and quantifiable.

27 keywords...

hide…