Kinetic analysis of ¹⁸F-FMISO dynamic PET using shortened datasets

Objectives To investigate the feasibility of kinetic analysis of ¹⁸F-FMISO dynamic PET (dPET) using shortened datasets.

Methods ¹⁸F-FMISO PET measurements were performed in (i) 10 HT29 colorectal carcinoma-bearing nude rats (0-90 min dPET) using the R4 or Focus 120 microPET, and (ii) 6 patients with head and neck cancer (0-45 min dPET followed by two 10-min static scans at 96±5 and 159±16 min post-injection) using the GE DSTE PET/CT. Each patient underwent a prior ¹⁸F-FDG PET/CT study. PET images were reconstructed using (i) 3D MAP (rats) or (ii) OS-EM (patients) algorithms. Whole-tumor volumes of interest (wVOIs) were delineated (i) manually (rats) or (ii) on the ¹⁸F-FDG images using 50% threshold (patients). Voxels within wVOI were subclassified into 4 clusters (cVOI) based on their kinetic behavior using k-means clustering. The time activity curves (TACs) were determined from the dynamic images. The input functions were derived using the hottest (i) 150 voxels within the abdominal aorta (rats) or (ii) 20 voxels within the internal carotid artery (patients). Kinetic analysis was carried out for each of the following datasets: (i) Full dynamic dataset (FD; reference standard), (ii) the first 90min of data (patients only; SD1), and (iii) the first 45min of dynamic data (SD2), using an irreversible 2-tissue compartment model.

Results In rats, Pearson’s r between k₃ [K_i] for FD-SD2 correlations as calculated from wVOI (cVOI)-derived TACs were 0.98 (0.98) [0.98 (0.98)]. In patients, the corresponding values were 0.95 (0.92) [0.98 (0.95)] for FD-SD1, and 0.88 (0.71) [0.85 (0.75)] for FD-SD2 correlations. r for v_B, K₁ and K₁/k₂ were ≥0.98 for all correlations (both rats and patients).

Conclusions Our preliminary results show that kinetic analysis of ¹⁸F-FMISO PET is feasible using shortened dPET datasets of up to 45min.

Research Support This work was partially supported by NIH grant 1U01CA157442-01A1.