Development, evaluation and validation of a novel kinetic model approach using an image-derived arterial input function for absolute micro-PET quantification

In modern molecular imaging absolute quantification of biochemical interaction mechanisms is a crucial prerequisite to foster an in-depth understanding of biological processes. In this context, model-driven methods using compartmental models of interaction of the radiotracer with the target tissue have been described. In this context the assessment of the arterial input function (AIF), which represents the tracer concentration in the blood plasma over time, is a crucial prerequisite for the reliable application of compartment modelling [1].
 
Material&Methods
Imaging was performed on Sprague-Dawley rats (n=6) with an injury of the peroneal nerve in the right hind limb, using a micro-PET/CT scanner (Siemens Inveon, Siemens Medical Solutions, USA). 45MBq of [18F]FDG was injected in the tail vein and dynamic PET acquisition (70min) was performed. PET images were reconstructed using an OSEM 3D/OP-MAP. Image quantification was performed using PMOD 3.7 (PMOD Technologies Ltd., Switzerland). VOIs comprising the m. tibialis anterior in the right and left hind limb were outlined on the micro-CT and transferred to the PET images of the individual time frames. TACs were calculated and expressed as SUV (g/mL). The AIF was extracted using a VOI located in the tail vein and the glucose metabolic rate (MRGlu.) was derived assuming a glucose level of 5mmol/l.

Results/Discussion
The mean value of the SUV was found to be 0.379±0,049 g/mL for the healthy limb and 0.631±0.103 g/mL for the injured limb respectively. The corresponding MRGlu revealed mean values for the healthy hind limb of 2.196±1.942 μmol/min/100g and 5.547±3.880 μmol/min/100g for the injured side. Although, differences between the healthy and injured limb of the dedicated rats were statistically significant for both, the SUV (p = 0.0080) and the MRGlu (p=0.0194), the application of the novel kinetic model approach determining the MRGlu was shown to be able to depict individual differences of each animal with higher reliability and validity. The calculated MRGlu were found to be significant higher for the injured limb of all six individual animals compared to the healthy limb, contrasting the side-specific evaluation using the SUV, in which five animals revealed a higher uptake value for the injured limb and for one animal paradoxically the SUV was found to be higher for the healthy side.

Conclusion:
Herein we successfully presented a novel approach for determining the MRGlu using an image-derived arterial input function. Although the gold standard for identifying the AIF is manual blood sampling after bolus administration of the dedicated radiotracer, it was demonstrated that the assessment of MRGlu could be accomplished by using an image-derived input function, even when the left ventricle or carotids were outside the field of view.