Objective: There is much controversy about the correlation between the degree of 2-[(18)F]fluoro-2-deoxy-D: -glucose (FDG) uptake and clinical aggressiveness of non-Hodgkin's lymphoma (NHL). In this study, we investigated whether partial volume corrected FDG uptake based on count recovery coefficient in 3D-positron emission tomography (PET)/computed tomography (CT) correlates with the clinical aggressiveness of NHL and improves diagnostic accuracy.
Methods: Forty-two patients with NHL underwent FDG-PET/CT scans (26 aggressive NHLs and 16 indolent ones). Count recovery curve was obtained using NEMA 2001 body phantom. Scan protocol and reconstructive parameters in the phantom study were the same as those in a clinical scan except for emission time. Relative recovery coefficient (RC) was calculated as RC = A/B (A, maximum pixel count of each hot sphere; B, maximum pixel count of greatest sphere). Partial volume corrected maximum count of standardized uptake value (PVC-SUV) was calculated as PVC-SUV = NC-SUV/RC (NC-SUV: non-corrected maximum count of SUV). Three parameters (NC-SUV, PVC-SUV, and size) between aggressive and indolent NHLs were compared.
Results: Significant differences were shown in all parameters between aggressive and indolent NHLs. Means +/- SD of NC-SUV, PVC-SUV, and size was as following: NC-SUV (15.3 +/- 6.9, 8.7 +/- 7.0; P < 0.01), PVC-SUV (18.2 +/- 8.1, 12.7 +/- 7.8; P < 0.05), and size (mm, 32.4 +/- 18.3, 21.9 +/- 10.3; P < 0.05). When an NC-SUV of 9.5 was the cutoff for aggressive NHL, the receiver-operating-characteristic (ROC) analysis correctly identified 21 of 26 aggressive ones. Sensitivity and specificity were 81% each, and the positive and negative predictive values were 88% and 72%, respectively. When a PVCSUV of 11.2 was the cutoff, the ROC analysis revealed 81% sensitivity, 63% specificity, and positive and negative predictive values of 78% and 67%, respectively. At a cutoff for aggressive NHL of a size of 27 mm, the ROC analysis revealed 50% sensitivity, 81% specificity, and positive and negative predictive values of 81% and 50%, respectively. The comparison of area under the curve in ROC analyses indicated that NC-SUV showed the greatest diagnostic accuracy (NC-SUV 0.84, PVC-SUV 0.72, and size 0.69).
Conclusions: Diagnostic accuracy of PVC-SUV was inferior to that of NC-SUV. These results suggest that NC-SUV, which contains information on both size and FDG density, provides better differentiation between aggressive and indolent NHLs than PVC-SUV.