Discriminating inflammation from malignancy with short-dynamic patlak parametric 18 F-FDG PET/CT

Differentiating malignant from inflammatory uptake on ¹⁸F-FDG PET/CT remains a major diagnostic challenge, as standardised uptake value (SUV) lacks specificity. This study evaluated whether parametric imaging from short-duration dynamic FDG PET/CT provides complementary information beyond SUV for distinguishing malignancy from inflammation.

Twenty-eight patients undergoing oncologic PET/CT (breast, lung, lymphatic or gastrointestinal cancer) were included, yielding 68 lesions (43 malignant, 25 inflammatory). Short dynamic acquisitions (20 min) were motion-corrected and used to generate influx rate (Ki) and distribution volume (Vd) maps. Lesions were segmented on SUV images (40% SUVmax), and radiomic features were extracted from SUV, Ki, and Vd maps, including core and peritumoral regions. Classification performance was assessed both using logistic regression and Random Forest model.

Malignant lesions exhibited higher mean values than inflammatory lesions for both SUV (6.0 vs. 2.7 g/ml) and Ki (2.0 vs. 0.8 ml/min/100 ml), while Vd values largely overlapped between classes (50 vs. 47%). No single parameter provided a reliable discriminative threshold. Restricting the analysis to equivocal SUV values (< 5.2 g/ml), multivariate regression combining SUV Mean, Ki/Vd Variance, and Ki Entropy achieved an accuracy of 0.86 (pseudo-R² = 0.42), outperforming SUV Mean alone (accuracy 0.77, pseudo-R² = 0.26). Core-peritumoral analysis revealed as Ki/Vd Variance the most statistically significant features between the two classes (4.35 ± 3.36 malignant vs. 1.93 ± 1.34 inflammatory; p = 0.006). Random Forest classification confirmed superior performance of parametric features (ROC AUC 0.86 ± 0.10) compared with SUV-only models (0.83 ± 0.08).

Short dynamic Patlak FDG PET/CT improves differentiation of malignant from inflammatory uptake beyond SUV alone. Decomposing FDG uptake into metabolised (Ki) and unmetabolised (Vd) fractions, components provide physiologically meaningful insights, revealing steeper core-to-peritumoral metabolic gradients in malignancy and more homogeneous tracer distribution in inflammation. These findings support the added value of parametric imaging and motivate prospective validation in larger clinical cohorts.