The effect of mis-calibration of the CT on quantification of PET/CT images

Abstract

1730

Objectives: The advent of the dual modality PET/CT scanner has allowed the use of co-registered CT images to be converted into the attenuation correction map used during data reconstruction in PET. The CT image is typically transformed into the attenuation correction map through bi- or tri-linear scaling with the value for water at 0 Hounsfield units (HU) being the reference point. The aim of this study was to determine the effect of mis-calibration of the 0 HU point on the final reconstructed PET image in terms of SUV and activity concentration changes. Methods: The NEMA NU-2 2001 chest phantom with inserts removed was filled with water and 2.7 mCi of FDG and uniformly mixed. The phantom was then scanned on a GE Discovery ST PET/CT machine, using 140 kV and 300 mAs CT image and a 15 min PET emission scan. The CT attenuation correction (CTAC) map was then calculated for different HU values of water (0, ±1, ±2, ±3, ±5, ±10, ±25, ±100). The PET data was then reconstructed with each of the CTAC files. Regions of interest were drawn at the center of the phantom, at the top edge and at the lower left corner. The mean and maximum activity concentrations were measured in each ROI and used to calculate the SUVmean and SUVmax. These values were evaluated as the percent change from 0 HU. Results: There was an approximately linear relationship between the activity concentrations or SUV values and the HU of water. Activity concentrations or SUV values determined from the different ROIs were very similar. There was a ±2% change in values for a ±5 HU change in water and a ±4% change in values for a ±10 HU change in water. Conclusions: PET/CT quality control should include routine verification of the CT number accuracy of water to ensure that there are no gross mis-calibrations of the CT. The results here indicate that provided the CT number falls within ±5 HU for water, there will be less than 2% error in the SUV values measured in a chest-sized object. Somewhat larger errors may be possible for obese patients or if the arms are positioned in the field of view.