Brain blood flow using [15O]H2O and PET

The brain receives about 15% of the cardiac output and uses about 20% of oxygen used by the body. The cerebral metabolic rate for oxygen, CMRO2, and cerebral blood flow (CBF, perfusion) are tightly coupled (Raichle et al., 1976a). Young women have higher perfusion than men in the cortical grey matter whole, although men have higher synaptic density than women and CMRO2 is similar in both sexes (Aanerud et al., 2017). Cerebral perfusion decreases with ageing in women but not in men, and at age 65 no differences in perfusion remains between sexes (Aanerud et al., 2017). Caffeine affects cerebral perfusion, and both caffeine consumption and withdrawal are confounding variables in brain perfusion studies (Field et al., 2003).

CBF in cortical grey matter is ∼50 mL×100 g-1×min-1 (Aanerud et al., 2017), and the ratio of grey and white matter blood flow is ∼2.8 (Østergaard et al., 1999). Physiological partition coefficient (p) of water for the whole brain is 0.95 mL/mL, and for grey and white matter 1.03 and 0.86 mL/mL, respectively (Herscovitch & Raichle, 1985). Depending on the data analysis, p may need to given in alternative units: 0.90, 0.98, and 0.82 mL/g, or, 0.96, 1.04, and 0.87 g/g, for the whole brain, grey matter, and white matter (Herscovitch & Raichle, 1985). Normal vascular volume fraction for the brain grey matter is 5.2%, for white matter 2.7%, and for cerebellum 4.7% (Leenders et al., 1990).

Analysis methods

The analysis method of blood flow (perfusion) in the brain (CBF) is based on the specific one-tissue compartment model for [15O]H2O. [15O]H2O is freely diffusible and metabolically inert. Repeated measurements with [15O]H2O or other tracers can be performed within one scanning session because of the short halflife of 15O. [15O]H2O and [15O]O2 studies are often performed together to quantify CBF, OEF, and CMRO2.

Perfusion reserve (usually 30-40%) can be measured by assessing CBF in baseline and after acetazolamide challenge, either by two bolus PET studies, or during one constant-infusion radiowater PET (Weber et al., 2004).

Blood flow can be estimated from dynamic PET data after a bolus infusion of [15O]H2O or [15O]CO2, either from regional tissue time-activity concentration curves (TTACs), or from dynamic PET image to produce perfusion map, using kinetic model fitting methods, or, also from static imaging using ARG method. These quantitative methods use arterial blood curve (BTAC) as their arterial input function (AIF).

The first quantitative CBF measurements with PET and [15O]H2O were not performed with dynamic scanning, but using steady-state technique. Since modern PET scanners enable collection of dynamic data with good accuracy, and dynamic and ARG methods lead to lower radiation dose, the steady-state method is seldom used.

Alternatively, static or summed dynamic PET image data can be used without blood sampling in brain activation studies (Evans et al., 1992); only an index of regional perfusion changes can then be observed. Pixel values can be normalized with global mean, reference region, or reference cluster (Borghammer et al., 2009). Population-based input function can be used to improve the quantitation of regional blood flow changes in brain activation studies.

Semi-quantitative perfusion estimates can be calculated using double-integration method (Koopman et al., 2019).

Input function

Gold-standard input function needed for absolute quantification of CBF is arterial blood curve, preferably obtained using on-line sampling system.

Image-derived input function is difficult to obtain in brain PET studies because the carotid arteries are so thin that spill-in and spill-out effects need to be accounted for. Dose-optimization with TOF-enabled PET/MR enables the measurement of angiograms using both PET and MR, and based on those the AIF can be calculated (Khalighi et al., 2018).

In theory, reference tissue could be used to cancel out the arterial input function from the equations, but in case of radiowater, the perfusion in the reference tissue should be known. Mejia et al (1994) used whole brain as reference, assuming that its perfusion is 50 mL/(dL*min), and assuming that the partition coefficient p is 1 in all brain regions. Lammertsma (1994) noted that fitting both perfusion and p would provide perfusion estimates that would not be biased by variable p and tissue heterogeneity. Watabe et al (1995) validated the method further, using weighted integration method, and generalized it to assess perfusion in two brain regions, serving as reference region for each other. This idea has since then expanded to model-based input function methods for cerebral [15O]H2O data analysis (Watabe et al., 1996; Kudomi et al., 2002; Treyer et al., 2003; Kudomi et al., 2016). Integrated PET-MRI systems have allowed simultaneous measurement of brain perfusion using radiowater PET and phase-contrast MRI. MRI-based perfusion estimate in the brain can be used as the reference tissue perfusion for the PET method (Ssali et al., 2018). Fiestra et al. (2018) validated fMRI assessment of cerebrovascular reactivity (CVR) against PET-derived cerebral blood flow reserve, measured using the method by Treyer et al (2003). The method of Treyer et al (2003) is based on a population average AIF, which is used to calculate parametric images of K1 and k2, and the K1 image, having better quality, is scaled to correct level, representing CBF, based on cortical average of K1 and k2 values. The method essentially assumed that the partition coefficient p is 1 in all brain regions.

Preprocessing arterial blood data

Arterial blood data, collected using on-line sampling system, must be calibrated and corrected for physical decay, dispersion and time delay. It is recommended that regional tissue TAC from pancreas is used in time delay correction.

  1. Process the count-rate data, or draw ROI on the whole brain and calculate tissue TAC to be used as “head curve” in time delay correction (below). Make sure that TAC files do not contain ROIs of large blood pools!
  2. On MS Windows PC in TPC network, do the corrections for blood data using water_input script. Alternatively, these corrections can also be done using a series of low-level commands.
  3. Verify visually that the corrected blood TAC is fine and that time delay correction has moved it to start to rise at the same time as the tissue TACs. Previous water_input command made a graph of these curves. Alternatively you can create the plot by yourself. Sampler blood TAC often contains close-to-zero values in the end, which should be removed with a text editor, or left out in analysis when setting the fit or integration time.

Calculation of CBF

Follow the general instruction for analysis of radiowater data to calculate CBF from either regional TTAC data or to compute perfusion maps using kinetic method, or instructions for ARG method to compute perfusion maps with ARG technique.

Kinetic method can provide also an estimate of the arterial plus capillary blood volume, which may be more reliable hemodynamic parameter than the total blood volume (including venous volume) resulting from [15O]CO PET study (Okazawa et al 2001).

Partial volume effect

Since the cerebral cortex is only ∼2-4 mm thick, and the cortex of cerebellum is even thinner, cortical PET data suffers from substantial partial volume effect (PVE). Since the ratio of grey and white matter blood flow is ∼2.8, it is obvious that measured CBF is highly dependent on the relative contributions of grey and white matter to the image voxel or volume-of-interest.

The basic radiowater model assumes tissue homogeneity, which assumption is broken in case of tissue heterogeneity, whether it is caused by biological tissue heterogeneity or PVE. In compartmental model analysis the tissue heterogeneity causes bias especially in partition coefficient (p=K1/k2) estimate, but affects also the CBF estimates. To demonstrate the effect of partial volume effect on the radiowater model, the results of a very simple simulation are first shown below. The data and codes used in the simulations on this page are available in GitLab.

Simulated radiowater TACs
Figure 1. Simulation on the effect of PVE.
Tissue time activity curves (TTACs) representing cerebral grey matter (GM) and white matter (WM) are simulated using a one-tissue compartmental model and arterial blood input function (purple). CBF (K1) of 50 and 20 mL×100 mL-1×min-1 was used for GM (black) and WM (green), respectively. For simplicity, vascular volume fraction is ignored, and partition coefficient (p=K1/k2) is set to 1 for both regions.
PVE is simulated by mixing the TTACs of GM and WM in ratio 1:1 (blue); if the PVE is assumed to affect CBF linearly, the estimated CBF from this region should be (50+20)/2 = 35 mL×100 mL-1×min-1. For comparison, also a TTAC with this CBF was simulated (red).
Concentration units are SUVs.

Simulated radiowater TACs fitted using compartmental model Simulated radiowater TACs fitted for 180 s using compartmental model
Figure 2a. Compartmental model fits of the simulated TTACs.
When compartmental model is fitted (420 s and 180 s) to the simulated TTACs, the fits are excellent and model parameters are well reproduced, except for the mixed GM & WM TTAC (green): the fit is not optimal, although with noisy data and especially with shorter fit time this might not be noticed. When 420 s of data were used in the fitting (left), the estimates of K1, k2, and p for the mixed TTAC were 34, 0.37, and 0.92, respectively. When fit time was constrained to 180 s (right), results were 35, 0.39, and 0.88, respectively. Thus, PVE causes bias especially to the p=K1/k2. Data weighting has limited effect on the results.

If the simulated TTACs (Figure 1) are analyzed using ARG, the CBF is well reproduced, except for the mixed TTAC where CBF estimate is only 30 mL×100 mL-1×min-1, if p is set to 1, that was used to simulate the TTACs. If p in ARG calculation is set to 0.92, the result from compartment model fit to the mixed TTAC, then CBF estimate 34 is obtained, which is closer to the true value, but results from TTACs representing homogeneous tissue are then highly overestimated.

Yokoi plot of simulated radiowater TACs
Figure 2b. Yokoi plot of the simulated TTACs.
Yokoi plots of the simulated homogeneous TTACs are linear, with slight curvature in the plot of the mixed GM and WM TTAC. When line is fitted to data collected between 30 and 300 s, correct model parameters are obtained from the homogeneous TTACs, and for the mixed TTAC the estimates of K1, k2, and p were 34, 0.38, and 0.90, respectively.

The second simulation is based on more realistic assumptions: Grey and white matter are mixed in variable proportions (80:20, 60:40, 40:60, 20:80). Perfusion in grey matter is varied in the range 30-70 mL×100 mL-1×min-1 and white matter perfusion is still set to 20 mL×100 mL-1×min-1. Partition coefficient (p) is set to 1.03 and 0.86 mL/mL in GM and WM, respectively. VB is set to 5.2 and 2.7% in GM and WM, respectively, and arterial fraction of it is set to 30% in both GM and WM.
Simulated tissue data is analyzed with compartmental model fitting using fit_h2o (Figure 3a), Yokoi plot using yokoi with line fit to 36-300 s of the data (Figure 3b), autoradiographic method with 100 s integration time and p=0.8 (Figure 4), and perfusion ratio method using perfrat (Figure 5). Obviously, increasing the PVE by decreasing the volume fraction of GM and increasing the volume fraction of WM with lower perfusion leads to lower perfusion estimates in all analysis methods. If PVE (or heterogeneity) is similar in all cortical regions, then CBF estimates will be comparable although biased.

Compartmental model K1 estimated from simulated radiowater TACs Compartmental model k2 estimated from simulated radiowater TACs
Figure 3a. Comparison of K1 and k2 from compartmental model fitting to the true CBF used to simulate the TTACs.

K1 estimated from simulated radiowater TACs using Yokoi plot k2 estimated from simulated radiowater TACs using Yokoi plot
Figure 3b. Comparison of K1 and k2 from the Yokoi plots to the true CBF used to simulate the TTACs. Time range of the linear fit could be altered to get perfusion estimates that slightly better match the high or low perfusion tissue component in heterogeneous tissue.

Perfusion estimated using ARG from simulated radiowater TACs
Figure 4. Comparison of perfusion from ARG method to the true CBF used to simulate the TTACs.
Tissue heterogeneity (or PVE) leads to lower apparent partition coefficient (p) of water, and therefore p=0.8 is often used, also in this analysis. If tissue data consists of pure GM (blue), or contains very little WM, then this value is too low and leads to an overestimated CBF.

Data and codes used in the simulations on this page are available in GitLab.

Perfusion ratio from simulated radiowater TACs using region with highest perfusion
            as the reference region Perfusion ratio from simulated radiowater TACs using an average region as
            the reference region
Figure 5. Comparison of perfusion ratio to the true CBF used to simulate the TTACs.
Perfusion ratio is often used in animal studies where one side of the brain or other kidney can be used as normal reference region.
On the left, the region with the highest CBF is used as reference region (therefore the one missing point). On the right, region with GM perfusion 50 mL×100 mL-1×min-1, mixed with WM in 80:20 (red), is used as the reference region.
Perfusion ratio is biased in several ways, but can provide robust differences without blood sampling.

See also:



References:

Blomqvist G. On the construction of functional maps in positron emission tomography. J Cereb Blood Flow Metab. 1984; 4: 629-632. doi: 10.1038/jcbfm.1984.89.

Boellaard R, Knaapen P, Rijbroek A, Luurtsema GJ, Lammertsma AA. Evaluation of basis function and linear least squares methods for generating parametric blood flow images using 15O-water and positron emission tomography. Mol Imaging Biol. 2005; 7(4): 273-285. doi: 10.1007/s11307-005-0007-2.

Furmark T, Tillfors M, Marteinsdottir I, Fischer H, Pissiota A, Långström B, Fredrikson M. Common changes in cerebral blood flow in patients with social phobia treated with Citalopram or cognitive-behavioral therapy. Arch Gen Psychiatry 2002; 59: 425-433. PMID: 11982446.

Herscovitch P. (1995): Cerebral blood flow, volume, and oxygen metabolism. In: Principles of Nuclear Medicine. (Eds. Wagner HN Jr, Szabo Z, Buchanan JW) 2nd ed., W.B. Saunders Co, Philadelphia, 505-514.

Herscovitch P, Markham J, Raichle ME. Brain blood flow measured with intravenous H215O. I. Theory and error analysis. J Nucl Med. 1983; 24: 782-789. PMID: 6604139.

Iida H, Kanno I, Miura S, Murakami M, Takahashi K, Uemura K. Error analysis of a quantitative cerebral blood flow measurement using H215O autoradiography and positron emission tomography, with respect to the dispersion of the input function. J Cereb Blood Flow Metab. 1986; 6: 536-545. doi: 10.1038/jcbfm.1986.99.

Kanno I, Lammertsma AA, Heather JD, Gibbs JM, Rhodes CG, Clark JC, Jones T. Measurement of cerebral blood flow using bolus inhalation of C15O2 and positron emission tomography: description of the method and its comparison with the C15O2 continuous inhalation method. J Cereb Blood Flow Metab. 1984; 4(2): 224-234. doi: 10.1038/jcbfm.1984.31.

Kanno I, Iida H, Miura S, Murakami M, Takahashi K, Sasaki H, Inugami A, Shishido F, Uemura K. A system for cerebral blood flow measurement using an H215O autoradiographic method and positron emission tomography. J Cereb Blood Flow Metab. 1987; 7: 143-153. doi: 10.1038/jcbfm.1987.37.

Kudomi N, Maeda Y, Sasakawa Y, Monden T, Yamamoto Y, Kawai N, Iida H, Nishiyama Y. Imaging of the appearance time of cerebral blood using [15O]H2O PET for the computation of correct CBF. EJNMMI Res. 2013; 3(1): 41.

Lammertsma AA. Quantification of cerebral blood flow. Neuromethods 2012; 71: 99-109. doi: 10.1007/7657_2012_43.

Leenders KL, Perani D, Lammertsma AA, Heather JD, Buckingham P, Healy MJR, Gibbs JM, Wise RJS, Hatazawa J, Herold S, Beaney RP, Brooks DJ, Spinks T, Rhodes C, Frackowiak RSJ, Jones T. Cerebral blood flow, blood volume and oxygen utilization: normal values and effect of age. Brain 1990; 113: 27-47. PMID: 2302536.

Perlmutter JS, Powers WJ, Herscovitch P, Fox PT, Raichle ME. Regional asymmetries of cerebral blood flow, blood volume, oxygen utilization and extraction in normal subjects. J Cereb Blood Flow Metab. 1987; 7: 64-67. doi: 10.1038/jcbfm.1987.9.

Raichle ME, Grubb RL Jr, Gado MH, Eichling JO, Ter-Pogossian MM. Correlation between regional cerebral blood flow and oxidative metabolism. Arch Neurol. 1976a; 33: 523-526. doi: 10.1001/archneur.1976.00500080001001.

Raichle ME, Eichling JO, Straatmann MG, Welch MJ, Larson KB, Ter-Pogossian MM. Blood-brain barrier permeability of 11C-labeled alcohols and 15O-labeled water. Am J Physiol. 1976b; 230(2): 543-552. doi: 10.1152/ajplegacy.1976.230.2.543.

Raichle ME. Quantitative in vivo autoradiography with positron emission tomography. Brain Res Rev. 1979; 1: 47-68. PMID: 385113.

Raichle ME, Martin WRW, Herscovitch P, Mintun MA, Markham J. Brain blood flow measured with intravenous H215O. II. Implementation and validation. J Nucl Med. 1983; 24: 790-798. PMID: 6604140.

Ter-Pogossian MM, Eichling JO, Davis DO, Welch MJ, Metzger JM. The determination of regional cerebral blood flow by means of water labeled with radioactive oxygen 15. Radiology 1969; 93(1): 31-40. doi: 10.1148/93.1.31.

Walker MD, Feldmann M, Matthews JC, Anton-Rodriguez JM, Wang S, Koepp MJ, Asselin M-C. Optimization of methods for quantification of rCBF using high-resolution [15O]H2 PET images. Phys Med Biol. 2012; 57: 2251-2271. doi: 10.1088/0031-9155/57/8/2251.

Watabe H, Itoh M, Cunningham V, Lammertsma AA, Bloomfield P, Mejia M, Fujiwara T, Jones AKP, Jones T, Nakamura T. Noninvasive quantification of rCBF using positron emission tomography. J Cereb Blood Flow Metab. 1996; 16: 311-319. doi: 10.1097/00004647-199603000-00017.



Tags: , , , , , ,


Updated at: 2019-03-30
Created at: 2014-05-07
Written by: Vesa Oikonen