Differences in the quantitative parameter of lung lesions in dynamic chest FDG PET/CT imaging: framing protocol, reconstruction algorithm, software processing

Abstract

Objective This study compares kinetic parameters (K1, k2, k3, Ki) from dynamic chest PET/CT for lung lesions across different framing protocols, reconstruction algorithms, and software processing, aiming to evaluate quantification consistency and support multi-center standardization. Methods This study enrolled 107 patients with 124 lung lesions who underwent 65-min dynamic 18F-FDG PET/CT. Dynamic data were compared across three factors: framing protocols (28 frames vs. 39 frames), reconstruction algorithms (the Block Sequential Regularized Expectation Maximization [BSREM] group vs. the third-order Hermite interpolation [Pitch-In] group), and software platforms (the Matlab group vs. the PMOD group). Kinetic parameters were derived based on an irreversible two-tissue compartment model. Subjective image quality was graded by two experts using a 5-point Likert scale.We performed Wilcoxon signed-rank tests with Bonferroni correction, intraclass correlation coefficients (ICCs), Bland–Altman analysis, and post-hoc power analysis (n=125) to evaluate statistical significance, consistency, and agreement. Results The Pitch-In group achieved scores >3 points and demonstrated significantly superior subjective image quality over the BSREM group (P < 0.05), except for comparable lesion discrimination. For kinetic parameter quantification, no statistically significant systematic differences were observed between the Pitch-In group and the BSREM group for K1, k3, and Ki (P > 0.05; ICCs: 0.795–0.955), though Pitch-In yielded higher k2 values (P = 0.003). No significant differences were detected between the Matlab group and the PMOD group for K1 and Ki (P > 0.05), achieving high absolute agreement (ICCs: 0.854 and 0.978; Ki bias: 0.000). Between 28 frames and 39 frames protocols, no significant differences were found for all parameters (P > 0.05; ICCs: 0.885–0.957), with a minimal Ki bias of 0.001 (95% LoA: -0.010 to 0.011). Conclusions The primary quantitative endpoint, the Ki value, remains highly robust across all evaluated framing protocols, software platforms, and reconstruction algorithms. This strong methodological equivalence facilitates cross-center data harmonization and clinical standardization for multi-center dynamic 18F-FDG PET/CT.