Abstract
1739
Purpose: Nowadays the use of integrated PET-CT scanners has gained widespread acceptance proving to be consolidated in the clinical setting. Radionuclide imaging requires adequate correction for the attenuation effect caused by different body tissues. This correction process is essential for reliable quantitative assessment as well as improved image quality. CT data give opportune information on tissue density to obtain the attenuation map, besides providing anatomical referencing of the PET findings. Main requirement to obtain a good CT-based attenuation correction is the alignment of transmission and emission images, in order to guarantee a proper matching between the attenuation map and functional images. While image alignment and attenuation correction accuracy are extremely important, there is not yet a standard methodology for assessing these on hybrid systems. International guidelines prescribe to perform the alignment test by using specific manufacturer phantoms combined with closed source analysis software packages supplied by the vendor. The purpose of this work was to propose a full independent tool for evaluating CT and PET images alignment by scanning a commercially available phantom.
Methods: PET-CT data were acquired on a Discovery-IQ system (GE Healthcare, Milwaukee,USA) using a dedicated phantom (PET-CT Phantom, Data Spectrum Corporation), equipped with inserts visible in both modalities: 5 coplanar spheres and three aluminum rods fillable with radioactive solution. A Matlab code was written to perform image analysis. To evaluate translations, the spheres filled with 18F-FDG radioactive solution were considered through these steps: segmentation of the spheres using a region growing algorithm, calculation of the centre of mass for each sphere and comparison of its coordinates in PET and CT set of images. Rotations were assessed by considering the positioning of 3 aluminium rods filled with radioactive solution in PET and CT: the angles for swivel, roll and tilt rotations were calculated in sagittal, coronal and axial views. A series of nominal translations (3-5-8-10 mm in the three spatial directions) and rotations (swivel, tilt and roll, 3.2° up to 6.8°) was simulated to validate the method in terms of sensibility, reproducibility and linearity.
Results: The sensibility of the method turned out to be less than 1 mm for translations and between 1.0° and 1.7° for rotations. The reproducibility (coefficient of variation over 5 identical acquisitions) was less than 1% both for translations and for rotations. Linearity considered the detected translation/rotation as a function of the nominal one: the angular coefficient for the linear regression ranged from 0.87 (Y-axis translation) to 1.35 (tilt rotation). Conclusion: The method proposed represents an alternative to the use of vendor specific phantoms and software analysis, providing a general validity across each type of scanner. Further applications of this work comprehend a similar developement in conventional nuclear medicine (SPECT-CT).