Value of Iterative Reconstruction, Attenuation Correction, and Image Fusion in the Interpretation of FDG PET Images with an Integrated Dual-Head Coincidence Camera and X-Ray-based Attenuation Maps

Abstract

To compare lesion detectability on 2-[fluorine-18]fluoro-2-deoxy-D-glucose (FDG) positron emission tomographic (PET) images obtained with a dual-head coincidence (DHC) gamma camera equipped with an integrated x-ray tube-based transmission system (a) with images reconstructed with filtered back projection (FBP) and those reconstructed with an iterative reconstruction algorithm based on coincidence-ordered subsets expectation maximization (COSEM), (b) with images reconstructed without and with attenuation correction (AC), and (c) with images reconstructed without and with image fusion for anatomic mapping. Thirty-five patients known or suspected to have malignancy underwent initial imaging with a dedicated positron emission tomography (PET) unit after injection of 10 mCi (370 MBq) of FDG. Transmission computed tomographic (CT) scans and FDG emission images were then obtained with the DHC camera. The proportion of lesions detected on the various sets of FDG DHC images was determined by using FDG PET as the standard of reference. Imaging findings were correlated with those from histologic examination and clinical follow-up, in consultation with the respective referring physicians. FDG PET depicted 78 lesions, 29 of which were equal to or less than 1.5 cm in diameter. FDG DHC depicted 52 of the 78 (67%), 59 of 78 (76%), and 61 of the 78 (78%) lesions, respectively, when image reconstruction was performed with FBP without AC, COSEM without AC, and both COSEM and AC. The detection rate of lesions 1.5 cm or smaller was better with COSEM and AC than with FBP (55% vs 34%, respectively). In addition, COSEM and AC allowed more confidence in the interpretation. None of these differences, however, were significant. Fusion of CT scans and FDG DHC images obtained with COSEM and AC allowed localization of lesions to the skeleton in three patients and to the liver versus adjacent bowel in three patients. Image fusion was especially helpful for localizing lesions in the neck in five patients. Anatomic mapping on fusion images was clinically relevant in 11 patients (31%). The COSEM reconstruction algorithm should replace FBP when available. Functional anatomic mapping improved lesion localization in one-third of the patients studied.