PET with /sup 18/F-fluoride: effects of interative versus filtered backprojection reconstruction on kinetic modeling

High focal uptake in patients poses particular problems in PET imaging. Filtered backprojection (FBP) introduces disturbing streak artifacts, adversely affecting the identification of structures and delineation of regions. Iterative reconstruction methods (MLEM) provide images of superb quality, however, the accuracy of quantitative results obtained from MLEM images has not been established for clinical data. Dynamic images were acquired over 1 hr with PET and /sup 18/F-fluoride in 6 patients with an old unilateral hip fracture. FBP and MLEM reconstruction was performed. Since the bladder was in the FOV and filled up with fluoride, FBP produced streaks hampering region delineation. Bone blood flow (kl) and fluoride influx rate (Ki) were estimated with a 3 compartment model. Analyzed regions (n=190) showed correlation coefficients between FBP and MLEM: 0.88 for kl and 0.97 for Ki. Affected and normal femoral head regions (n=30) yielded r=0.89 for kl and r=0.95 for Ki. Variations up to 46% were seen in individual data. It is concluded that in patients MLEM provides superior images at the expense of an increased reconstruction duration. The authors' procedure appeared acceptable in clinical routine. Quantitative estimates obtained with kinetic modeling from MLEM data were reliable and correlated highly to those obtained with the standard, validated FBP algorithm.