Improving Glucose Metabolic Rate Estimates without Arterial Blood Sampling
Abstract
Purpose
Two key challenges in pharmacokinetic modeling (PKM) of dynamic PET are metabolite contamination of the image-derived arterial input function (AIF) and the presence of these metabolites in tissue time-activity curves (TACs). We propose a method to correct for both issues without arterial blood sampling, using brain [18F]-FDG studies as an example, where the metabolite is [18F]-FDG trapped in red blood cells (RBCs).
Methods
Five dynamic FDG-PET brain scans from four patients were analyzed. High-resolution images were reconstructed using QClear (GE Healthcare) to minimize partial volume effects. AIFs were averaged from maximal points in 10-15 slices of the extracranial internal carotid artery. The approach used iterative simultaneous estimation (SIME), fitting four brain TACs with different kinetics and a corrected AIF. Our critical modification to the original SIME framework, which enables convergence, was to iteratively define the difference between the measured and fitted AIF as the RBC AIF and subtract its estimated contribution from the tissue TACs. At convergence, the image-derived AIF is successfully decomposed into the corrected plasma AIF and the RBC AIF.
Results
The algorithm converged using only image-derived data. Patlak graphical analysis with the corrected plasma AIF produced a glucose metabolic rate (MRglu) of 41.6 ± 14.6 µmol/min/100g, compared to 24.7 ± 7.7 µmol/min/100g using the uncorrected AIF. These MRglu estimates align with literature values but reveal a potential 1.7-fold underestimation when using the uncorrected, RBC-contaminated input function.
Conclusion
Our study demonstrates that a modified SIME (mSIME) method can estimate a metabolite-corrected plasma AIF without arterial blood sampling. This correction significantly impacts PKM, potentially improving the quantitative accuracy of MRglu estimates by addressing RBC contamination. Future work will involve validation using large animals with arterial blood sampling.