Attenuation map segmentation in low-dose PET/CT

In PET/CT studies, artifacts can occur in CT/AC scans when parameters are adjusted to reduce CT radiation exposure to very low levels, and these can in turn cause PET errors. Scanning at 80 kVp instead of 120 or 130 kVp can result in attenuation correction errors due to photon starvation, especially in large patients, and to beam hardening in the skull. We addressed this problem by segmenting CT images before they were used to generate attenuation coefficients. We used this method in F-18 PET studies of the RSD striatal brain phantom, in cardiac Rb-82 PET/CT, and in whole-body 18F-FDG PET/CT. In the brain phantom study, high-dose 120 kV CT was compared with low-dose 120 kV CT and even lower dose 80 kV CT. In 7 cardiac studies, low-dose 120 kVp CT scans (0.29 mSv) and ultra-low dose 80 kVp scans (0.12 mSv) were acquired in the same study, and the 80 kV CT images were segmented. PET images were reconstructed with AC by all three methods, and the stress polar maps were compared. In a data set from a whole-body study, the 120 kV CT image series acquired with typical clinical settings was transformed into an 80 kV image series and segmented, and again PET reconstruction was performed with all three methods, and SUV levels were compared. Low-dose 120 kV scans of the brain phantom were suitable for CT/AC even without segmentation. Segmentation improved the PET image uniformity. 80 kV CT/AC is not recommended for brain imaging, with or without segmentation, since some beam-hardening artifacts remained. In cardiac perfusion imaging, 80 kV CT/AC without segmentation resulted in an accurate diagnosis for non-obese patients and we observed no clinical advantage to 80 kV segmentation. In whole-body PET/CT, low-dose CT artifacts were seen at the level of the shoulders and pelvis. SUV values were hardly affected by the CT artifacts or by segmentation.