Comparing 511 keV Attenuation Maps Obtained from Different Energy Mapping Methods for CT Based Attenuation Correction of PET Data

The advent of dual-modality PET/CT scanners has revolutionized clinical oncology by improving lesion localization and facilitating treatment planning for radiotherapy. In addition, the use of CT images for CT-based attenuation correction (CTAC) decreases the overall scanning time and creates a noise-free attenuation map (µmap). CTAC methods include scaling, segmentation, hybrid scaling/segmentation, bilinear and dual energy methods. All CTAC methods require the transformation of CT Hounsfield units (HU) to linear attenuation coefficients (LAC) at 511 keV. The aim of this study is to compare the results of implementing different methods of energy mapping in PET/CT scanners. This study was conducted in 2 phases, the first phase in a phantom and the second one on patient data. To perform the first phase, a cylindrical phantom with different concentrations of K2HPO4 inserts was CT scanned and energy mapping methods were implemented on it. For performing the second phase, different energy mapping methods were implemented on several clinical studies and compared to the transmission (TX) image derived using Ga-68 radionuclide source acquired on the GE Discovery LS PET/CT scanner. An ROI analysis was performed on different positions of the resultant µmaps and the average µvalue of each ROI was compared to the reference value. The results of the µmaps obtained for 511 keV compared to the theoretical values showed that in the phantom for low concentrations of K2HPO4 all these methods produce 511 keV attenuation maps with small relative difference compared to gold standard. The relative difference for scaling, segmentation, hybrid, bilinear and dual energy methods was 4.92, 3.21, 4.43, 2.24 and 2.29%, respectively. Although for high concentration of K2HPO4 the three methods; hybrid scaling/segmentation, bilinear and dual energy produced the lowest relative difference of 10.91, 10.88 and 5%, respectively. For patients it was found that for soft tissues all the mentioned energy mapping methods produce acceptable attenuation map at 511 keV. The relative difference of scaling, segmentation, hybrid, and bilinear methods compared to TX method was 6.95, 4.51, 7, and 6.45% respectively. For bony tissues, the quantitative analysis showed that scaling and segmentation method produce high relative difference of 26 and 23.2%, respectively and the relative difference of hybrid and bilinear in comparison to TX method was 10.7 and 20%, respectively. Discussion and Based on the result obtained from these two studies it can be concluded that for soft tissues all energy mapping methods yield acceptable results while for bony tissues all the mentioned methods except the scaling and segmentation yield acceptable results.