The kinetics of [18F]FDG uptake over a more prolonged period has been studied in normal tissues and a number of different diseases in humans with dynamic and multi-time point static acquisitions [9]. Some of the longest delayed acquisitions were performed in glioma patients where tumour-to-background contrast continued to increase up to 8 h post-injection [3]. The potential clinical uses of delayed time-point imaging are improved sensitivity for lesion detection, specificity for malignancy and differentiation of low- and high-grade tumours. Small studies have investigated the use of delayed imaging in multiple organs and cancer types with mixed results. A meta-analysis of using delayed [18F]FDG PET for differentiating benign and malignant pulmonary nodules failed to demonstrate a benefit for delayed imaging [10]. Conversely, a meta-analysis of 758 patients concluded that the diagnostic accuracy of [18F]FDG PET for mediastinal lymph node metastases was improved by using the retention index of [18F]FDG calculated from dual time point imaging [11].

In mouse imaging there have been recent efforts to standardise image acquisition [12, 13]. Tumour uptake typically shows a period of rapid uptake before a plateau. Imaging during the plateau phase would be beneficial for reproducibility and repeatability as small variations in the timing of static imaging will have a minimal effect on the result. However, the majority of pre-clinical studies perform static PET acquisitions no later than 60 min post [18F]FDG injection when most tumours, particularly subcutaneous tumours which are poorly perfused, are likely be still in the rapid uptake phase. Furthermore, the dynamic [18F]FDG kinetics in mice over a prolonged period have not been defined. For example, in a dual time point study tumour SUV’s increased between 45 and 90 min but, to our knowledge, no previous study has looked at [18F]FDG biodistribution beyond this time in small animals [14].

In this study prolonged dynamic [18F]FDG-PET acquisitions were performed in three different mouse tumour models, including both subcutaneous and autochthonous models, with the primary aim of defining the kinetics for a range of subcutaneous and autochthonous models under different animal handling conditions. Fasting suppressed myocardial uptake and increased tumour [18F]FDG uptake in EL4 tumour-bearing mice, corroborating previous studies [5]. Eµ-Myc mice were not fasted as the accompanying clinical features of the model e.g., ascites, pleural effusion, airway compression and cardiac tamponade, rendered them too unwell to undergo fasting and prolonged anaesthesia. The period of fasting prior to mouse studies was aimed to be a minimum of 6 h [7], with variations up to 10 h primarily due to delays in scanner and [18F]FDG availability. None of the models demonstrated significant [18F]FDG efflux.

Vivariums are typically regulated at approximately 21–25 °C, temperatures that are comfortable for clothed humans. However, this temperature is below the thermoneutral temperature for mice, which is 29–31 °C, and therefore thermogenesis is activated to maintain body temperature and activation of brown fat has been visualised using [18F]FDG imaging [5, 15]. Previously it has been demonstrated that warming during the uptake period minimises brown fat [18F]FDG uptake. Warming immediately prior to the uptake period has also been proposed for clinical [18F]FDG PET imaging. We failed to demonstrate any significant difference in tumour [18F]FDG uptake in mice that were warmed or at room temperature prior to injection, and visually there was no significant brown fat uptake in either group. However, warming prior to injection did facilitate tail vein cannulation.

Previous studies of the repeatability of [18F]FDG-PET in fasted mice reported a 60 min CV in tumours of between 11.4 and 15.4% for the % injected dose per gram (15.1% for SUVmean) [16, 17]. In our study the in-group reproducibility (CV at 60 min) of tumour SUVmean in fasted mice (EL4, 2 groups; colo205) ranged from 7 – 13%. In the previous studies, test–retest scans were performed 6 h apart. Therefore, the conditioning and Circadian timing differed significantly between scans, which may account for why reproducibility in our study is comparable or better. Using the same methodology repeatability of [18F]FLT and [18F]FMISO, which are less dependent on animal conditioning, had lower CV’s (4–6%). In EL4 tumour bearing mice that were fed and warmed and Eµ-Myc mice the 60 min CV was higher at 20 and 39%, respectively.

Imaging during the initial rapid uptake phase means that relatively small differences in acquisition time will result in large variations in measured uptake. This can be seen in the same patient when multiple acquisitions are performed in different bed positions, a problem that will be alleviated by total body PET [18]. In most tumours and tissues other than brain and liver, the rate of dephosphorylation (k4) of [18F]FDG-phosphate is negligible and there is irreversible label trapping. After the initial bolus injection and rapid uptake, K1 and k2 rates start to equilibrate and k3 decreases resulting in a relative plateau in the tissue concentration of [18F]FDG. Small animal imaging can be performed in a single bed position and imaging in the plateau period would alleviate the need for precise image timing. In this study, the SUVmean increased in all tumours up to 60 min, with a mean increase from 30 – 60 min between 10.4 ± 4.5%. Beyond 60 min the percentage change in SUVmean reduced to 2.6 ± 3.8% from 60 – 90 min and 1.1 ± 6.7% from 90 – 170 min. These data therefore support performing static tumour imaging in mice at > 60 min post-injection.

In implanted subcutaneous tumours TBR continued to increase up to at least 60 min, an effect that was prolonged in fasted mice, for at least 170 min p.i. in Colo205 and EL4 tumours. In studies where maximising tumour-to-background contrast is paramount then delayed imaging would be appropriate, particularly when imaging tumours that are located within or adjacent to normal tissues that have low rates of uptake and signal is dominated by the blood pool e.g., liver, or significant rates of dephosphorylation [18F]FDG-P and washout e.g., brain [3, 19]. The relative insensitivity of SUV values, compared to TBR, to reductions in plasma [18F]FDG concentration over time is due to the imaging of unphosphorylated [18F]FDG in the tissue compartment [19].

This study used a previously validated method to derive an image derived input function (IDIF) from the inferior vena cava / aorta and generate influx rate constants for tumours (Ki) [20]. We examined the ROI’s in the left ventricle and carotid / internal jugular veins as alternative sites to define the blood pool, but both of these suffered significantly from partial volume effects and spill over from adjacent structures. A saline flush was not performed following [18F]FDG injection to avoid a double peak in the input function. Reconstructing the list-mode data with high temporal resolution to define the peak input function revealed that in time frames < 2 min there was a significant underestimation of activity in a [18F]FDG phantom included in the acquisitions. The need to define the peak of the arterial input function was obviated by using Patlak analysis, which is relatively insensitive to low temporal resolution [21].

There was good correlation between tumour SUVmean and Ki indicating that the tumour [18F]FDG concentration was dominated by irreversible uptake in the 20–60 min period used for Patlak analysis [22]. Given the increases in SUVmean that occurred over time we hypothesised that SUVmean measured at earlier times would underestimate Ki but this was not the case.

There was good agreement between tumour Ki’s calculated from the PBIDIF and individual IDIF’s. This would permit use of a PBIDIF to derive kinetic data where scanning of the whole uptake period cannot, or has not, been performed, or where a ROI cannot be defined over a suitable vessel. To be applied more generally this result will need to be externally validated, particularly if a PBIDIF were to be used with data generated on instruments from other vendors [12].

With prolonged scanning there were three considerations for the injected dose, (a) injecting doses that did not cause initial detector saturation, (b) maintained sufficient count rates for accurate quantification and image quality late in the acquisition when over two half-lives had elapsed and (c) not exceeding recommended injection volumes [13]. The dose injected was similar to that in with previous studies using the NanoScan PET/CT and detector linearity was confirmed using a [18F]FDG phantom in the field of view of each study [23].

This study had several limitations. The low numbers of mice in some of the groups limited some of the statistical conclusions that could be made. The lack of blood sampling meant that the study had to use the image derived input function method of Lanz et al. [20]. Kinetic analysis was further restricted to Patlak analysis because short time frames could not be reconstructed, resulting in underestimation of the arterial input function. With the exception of the brain, the ROIs included in the study are defined from the PET imaging and includes all tissues that demonstrated significant irreversible uptake (tumour, myocardium, Harderian gland) and those that demonstrated particularly high perfusion (liver, brain) and excretion (kidneys). The unenhanced CT acquired for the purposes of attenuation correction had poor soft tissue contrast and reliably differentiating organs in the abdomen was challenging, therefore these were not included. Other challenges to defining ROIs included spill-in effects from nearby organs e.g., lungs contaminated by spill-in from the heart and abdominal organs from the urinary tract.