Comparison between dual-time-window protocol with other simplified quantifications in dynamic total-body 18F-FDG PET imaging

Introduction: The kinetic parameters derived from dynamic total-body 18F-FDG PET imaging could provide additional value in disease characterization, staging, and treatment response evaluation. However, its associated long scan time often results in less patient comfort and lower scan efficiency. In this study, we implemented a dual-time-window protocol to obtain kinetic parameters from a total-body FDG scan. Comparison between the dual-time-window protocol and several existing simplified quantification methods was performed. The potential clinical usage of these simplified quantifications was therefore suggested.

Methods: The study included 28 scans performed on an uExplorer PET/CT (United Imaging Healthcare, Shanghai) at Henan Provincial People’s Hospital, China. Twenty subjects (9 males, 11 females, 55.3±9.2 years) were used for quantification of normal tissues and 10 subjects (5 males, 5 females, 58.6±11.3 years) were used for quantification of malignant lung tumors. Region-of-interests (ROIs) were drawn on normal tissues (cerebral cortex, liver and muscle) and tumors. The dual-time-window protocol composes of two short dynamic scans: an early scan performed for 10 mins post-injection and a late scan performed after a break with the end scan fixed at 60 min; different late scan durations (5, 10, 20 mins) were evaluated. According to characteristics of the tissue activity curves (TACs) in different regions, either linear or non-linear fittings (3rd degree rational function) was applied to estimate the missing data. Subsequent kinetic modelling was performed for these complete TACs by assuming irreversible two-tissue compartment model (2TCM). Fast non-linear least-square fitting was applied to estimate the micro- and macro-parameters (K1, k2, k3, Ki and Vb), with the hybrid input functions combined the image-derived input function (from acceding aorta) and a population-based input function (PBIF). ROI-based quantification and voxelized parametric images were generated and analyzed. Besides, other existing simplified quantifications, i.e. Patlak Ki (30-60 min), SUV (50-60 min) and FUR (fractional uptake ratio at 50-60 min) were computed for comparison. The correlation between each quantification with the reference Ki (2TCM with 60 min data) was assessed.

Results: The dual-time-window protocol (with 10-min early and 5-min late scan) can produce Ki and K1 with good consistency to the reference in multiple regions (Ki correlation - 0.975, 0.923, 0.985, 0.984, K1 correlation - 0.866, 0.976, 0.959, 0.795, in cerebral cortex, liver, muscle and tumor, respectively). The correlation coefficients for each simplified quantifications at ROIs were summarized in Table.1. Figure.S1 shows correlation analysis in tumors. Dual-time-window protocol produced the highest correlation of Ki estimation when compared with the reference, followed by FUR and Patlak while SUV has the worst correlation. In terms of the parametric image, Ki generated by dual-time-window protocol was more consistent with the reference Ki image than the one by Patlak analysis (Figure.S2a). K1 image derived from dual-time-window protocol also exhibited good consistency with the reference (Figure.S2b).

Conclusions: This study has shown that, with a dual-time-window protocol, it is possible to reduce the dynamic total-body FDG scan to 15 min. Accurate Ki and K1 quantification and acceptable visual quality of parametric images can be achieved. While considering the additional time and the complexity of implementation, other existing simplified quantifications, e.g. FUR, could be more appropriate in certain applications. We suggest when the clinical task is lesion detection that requires reliable visual assessment or quantifying micro-parameters such as K1, dual-time-window protocol is preferred; on the other hand, when the clinical task is to quantify the regional metabolic rate, with known lesion position or organs of interested, FUR with PBIF is more feasible as a surrogate of Ki which only requires regular scan time.

• Download figure
• Open in new tab
• Download powerpoint