ROC analysis of ordered subset expectation maximization and filtered back projection technique for FDG-PET in lung cancer

The purpose of this study was to evaluate the image quality and discriminative ability of FDG-PET images reconstructed by filtered back projection (FBP) and ordered subset expectation maximization (OSEM) in patients with lung cancer. Thirty-six subjects including normal controls and lung cancer patients underwent whole body PET scan (emission 3 min, transmission 1 min) using a GE Advance scanner after the administration of approximately 370 MBq of FDG. The raw data were then reconstructed by conventional FBP and OSEM. The PET images were interpreted by two observers with random exposure of normal and diseased cases for transverse and coronal sections. The results were evaluated using receiver operating characteristic (ROC) curve analysis. The optimal parameters for reconstructing by OSEM involved 2 iterations and 16 subsets. The image quality of thirty-six subjects including normal controls and lung cancer patients reconstructed by OSEM was generally superior to FBP in terms of visual analysis. Though FBP had poor image quality, it had better image contrast than OSEM. The ROC curve plotted for the identification of lung cancer in coronal sections showed better results than in transverse sections by FBP and OSEM, except for the OSEM result of observer 2. ROC analysis showed that OSEM performed better than conventional FBP in terms of its discriminative ability of lung cancer with FDG-PET, although there was no significant difference in the area under the ROC curve (p-value > 0.05). The results suggest that the simultaneous use of FBP and OSEM images is helpful for the FDG-PET diagnosis of lung cancer.