Comparative assessment of different energy mapping approaches in CT based attenuation correction: A patient study

Attenuation correction of PET emission data using spatially correlated CT images is fast and precise yielding a noise-free attenuation map (μmap) in comparison with radionuclide transmission scanning (TX). However, it is essential to convert the linear attenuation coefficients obtained from CT scans to those corresponding to 511 keV. Several conversion strategies have been developed including scaling, segmentation, hybrid and bilinear methods. The aim of this study is to compare the accuracy of different energy mapping methods for generation of 511 keV μmap using clinical studies. The procedure for generation of attenuation map from CT images using different energy mapping methods was assessed using clinical studies and the results compared to the TX image derived using Ga-68 rod sources acquired on the Discovery LS PET/CT scanner, were used as gold standard in this study. A region of interest analysis was performed at different locations of the μmaps. It was shown that for soft tissues, the relative difference of scaling, segmentation, hybrid and bilinear methods compared to TX technique were 11.3%, 9.2%, 11.3% and 10.8% respectively (no major difference). For bony structures, the quantitative analysis showed that the scaling method produces a substantial relative difference (31%). The relative difference of segmentation, hybrid and bilinear methods compared to TX were 29%, 14% and 18% respectively. However these results for lung tissue were 4%, 13%, 4% and 4% respectively for scaling, segmentation, hybrid and bilinear methods with a great difference for segmentation method. It can be concluded that for soft tissues all energy mapping methods give satisfactory results. For bone, the scaling and segmentation methods yield substantial relative differences but the other 2 methods give acceptable results. For lung tissue the results are approximately close to each other except for segmentation method.