Kinetic modeling of PET data without blood sampling

In positron emission tomography (PET) imaging, application of kinetic modeling always requires an input curve (IC) together with the PET data. The IC can be obtained by means of external blood sampling or, in the case of cardiac studies, by means of a region-of-interest (ROI) drawn on the blood pool. It is, however, very unsuitable to withdraw and to analyze blood samples, and in small animals, these operations become difficult, while ICs determined from ROIs are generally contaminated by emissions from neighboring sites, or they are underestimated because of partial volume effect. In this paper, we report a new method to extract kinetic parameters from dynamic PET studies without a priori knowledge of the IC. The method is applied in human brain data measured with fluorodeoxyglucose (FDG) human-brain and in cardiac-rat perfusion studies with ¹³N-ammonia and ¹¹C-acetate. The tissue blood volume (TBV), usually fitted together with the rate constants, is extracted simultaneously with the tissue time activity curves for cardiac studies, while for brain gray matter, TBV is known to be about 4% to 7%. The shape of IC is obtained by means of factor analysis from an ROI drawn around a cardiac tissue or a brain artery. The results show a good correlation (p<0.05) between the cerebral metabolic rate of glucose, myocardial blood flow, and oxygen consumption obtained with the new method in comparison to the usual method. In conclusion, it is possible to apply kinetic modeling without any blood sampling, which significantly simplifies PET acquisition and data analysis.