Cerebral blood flow and glucose metabolism in healthy volunteers measured using a high-resolution PET scanner

Abstract

Background: Positron emission tomography (PET) allows for the measurement of cerebral blood flow (CBF; based on [15O]H2O) and cerebral metabolic rate of glucose utilization (CMRglu; based on [18 F]-2-fluoro-2-deoxy-d-glucose ([18 F]FDG)). By using kinetic modeling, quantitative CBF and CMRglu values can be obtained. However, hardware limitations led to the development of semiquantitive calculation schemes which are still widely used. In this paper, the analysis of CMRglu and CBF scans, acquired on a current state-of-the-art PET brain scanner, is presented. In particular, the correspondence between nonlinear as well as linearized methods for the determination of CBF and CMRglu is investigated. As a further step towards widespread clinical applicability, the use of an image-derived input function (IDIF) is investigated.

Methods: Thirteen healthy male volunteers were included in this study. Each subject had one scanning session in the fasting state, consisting of a dynamic [15O]H2O scan and a dynamic [18 F]FDG PET scan, acquired at a high-resolution research tomograph. Time-activity curves (TACs) were generated for automatically delineated and for manually drawn gray matter (GM) and white matter regions. Input functions were derived using on-line arterial blood sampling (blood sampler derived input function (BSIF)). Additionally, the possibility of using carotid artery IDIFs was investigated. Data were analyzed using nonlinear regression (NLR) of regional TACs and parametric methods.

Results: After quality control, 9 CMRglu and 11 CBF scans were available for analysis. Average GM CMRglu values were 0.33 ± 0.04 μmol/cm3 per minute, and average CBF values were 0.43 ± 0.09 mL/cm3 per minute. Good correlation between NLR and parametric CMRglu measurements was obtained as well as between NLR and parametric CBF values. For CMRglu Patlak linearization, BSIF and IDIF derived results were similar. The use of an IDIF, however, did not provide reliable CBF estimates.

Conclusion: Nonlinear regression analysis, allowing for the derivation of regional CBF and CMRglu values, can be applied to data acquired with high-spatial resolution current state-of-the-art PET brain scanners. Linearized models, applied to the voxel level, resulted in comparable values. CMRglu measurements do not require invasive arterial sampling to define the input function.

Trial registration: ClinicalTrials.gov NCT00626080.

Figures

Figure 1

Correlation between Patlak- and NLR-derived Kivalues. Data of all 16 gray matter regions and a white matter brain region of nine healthy subjects are presented. Data points for each individual subject are shown with a separate symbol. The solid line indicates the line of identity.

Figure 2

Correlation of average Kivalues derived using parametric and regional Patlak analyses. Data of all 16 total gray matter regions and a white matter brain region are presented for nine subjects. Parametric values represent the mean of all voxels within an ROI. The solid line indicates the line of identity. Results for both images without smoothing (black dots) and those smoothed with a 6-mm Gaussian filter (white dots) are shown.

Figure 3

Representative parametric images of a single subject. The CBF image (upper panel) and the CMRglu image (lower panel) of the same subject are presented. The parametric CBF image was generated after smoothing with a 6-mm Gaussian filter.

Figure 4

Correlation of CBF values derived using parametric (basis function method) and regional (NLR) analyses. Data of 16 gray matter regions and a white matter region are shown for 11 subjects. Parametric values represent the mean of all voxels within an ROI. Data points for each individual subject are shown with a separate symbol. The solid line indicates the line of identity.

Figure 5

Correlation between IDIF- and BSIF-based NLR-derived Kivalues. The correlation is for 16 gray matter regions and a white matter region. Data points for each individual subject (n = 9) are shown with the same symbol. The solid line indicates the line of identity.

Figure 6

Correlation between IDIF- and BSIF-based Patlak-derived Kivalues. The correlation is for 16 gray matter regions and a white matter region. Data points for each individual subject (n = 9) are shown with the same symbol. The solid line indicates the line of identity.