50. Voxel-based quantification of L-[1-11C]leucine kinetics for determination of regional rates of cerebral protein synthesis
G. Tomasi1, A. Bertoldo2, S. Bishu3, A. Unterman3, C. Beebe Smith3 and K. Schmidt3
1Department of Diagnostic Radiology, Yale University, New Haven, Connecticut, USA; 2Department of Information Engineering, University of Padova, Padova, Italy; 3Section on Neuroadaptation and Protein Metabolism, National Institute of Mental Health, Bethesda, Maryland, USA
Objective: The L-[1-11C]leucine PET method1,2 for measurement of regional rates of cerebral protein synthesis (rCPS) has been employed to date only at the Region-of-Interest (ROI) level.2,3 We adapted and validated a Basis Function Method (BFM)4 to estimate parameters of the homogeneous tissue kinetic model of L-[1-11C]leucine at the voxel level, and determined rCPS and the fraction (λ) of unlabeled leucine in the tissue precursor pool for protein synthesis derived from arterial plasma.
Methods: At the voxel level BFM performance was assessed both on simulated data, where time-activity curves (TACs) with noise levels typical of voxel data were generated, and on measured data, where BFM estimates were compared to those determined with nonlinear least squares (NLLS). Performance indices were bias% and RMSE% (Root Mean Square Error). BFM parameter estimates, averaged over all voxels in a ROI, were also compared to estimates obtained by fitting the ROI TAC to either a homogeneous or heterogeneous tissue model; both simulated data and data from L-[1-11C]leucine PET studies in six healthy subjects were analyzed.
Results: In simulations of voxel data BFM yielded low-bias estimates of λ and rCPS; when tested on measured data BFM estimates at the voxel level were in good agreement with those determined with NLLS. In simulation of large numbers of voxels comprising a ROI (Figure) fits of the ROI TAC with the homogeneous tissue model gave substantial negative biases for the parameters K1 and k2+k3, and positive biases for rCPS; results on measured data were consistent with the simulations. The heterogeneity model fit of the ROI TAC provided good parameter estimates in simulation, but did not perform well on measured data. rCPS estimated with BFM was slightly negatively biased, but its variability was low.
Simulation of all voxels comprising occipital cortex. Parameters were computed in three ways: averaging BFM estimates over all voxels within the ROI (BFM), and fitting the homogeneous tissue model (HOM) and heterogeneous tissue model (HET) to the simulated ROI TAC. The heterogeneity model does not provide estimates of k2 + k3 and k4 in the mixed tissue.
Conclusion: BFM is a useful and robust method for analyzing L-[1-11C]leucine PET data at the voxel level. It provides reliable estimates with a computational cost substantially lower than that of NLLS. At the ROI level parameters computed by averaging BFM estimates over all voxels within the ROI were less biased on simulated data, and less variable on both simulated and measured data, than parameter estimates obtained by fitting the ROI TAC to either a homogeneous or heterogeneous tissue model.
Supported by: IRP/NIMH/NIH.
62. A statistical analysis of the hypothesis of the standard two-compartment model for dynamic PET-FDG data in normal brain
F. O'sullivan1,2, N. Fitzgerald1, J. O'sullivan1, M. Muzi2, D.A. Mankoff2, A.M. Spence2 and K.A. Krohn2
1Statistics, University College Cork, Cork, Ireland; 2Radiology, University of Washington, Seattle, Washington, USA
Objectives: While most PET-FDG studies focus on late time retention characteristics in tissue, there is often an opportunity to use the dynamic time-course to obtain a more complete kinetic description of factors influencing FDG metabolism. The usual approach models the tissue residue with the established 2-compartment model.1 But non-parametric approaches to residue modeling are used with dynamic contrast MR and CT data2,3 and recently we have developed a corresponding non-parametric approach that is applicable to dynamic PET data. The present study uses this technique to more carefully evaluate local cerebral residues from dynamic PET-FDG data.
Methods: The residue description of PET time-course data presents a density deconvolution estimation problem. Our solution uses the method of regularization with cross-validation. The computational implementation uses adaptive cubic B-spline and quadratic programming. Numerical simulation is used for evaluation of sampling variation and is also critically used as a way to assess the 2-compartment residue model hypothesis with FDG. Time-course datasets corresponding to 10 MR-identified brain regions in 12 normal subjects were analysed using the non-parametric and parametric (2-compartment) residue models. The 2-compartment model hypothesis was assessed by the data-fit, with appropriate adjustment for the increased flexibility of the non-parametric procedure.
Results: Strong (P-values <0.051) statistical evidence against the compartment model is found in 7 out of 8 gray matter regions considered (cerebellum, putamen, thalamus, and occipital, temporal, frontal and parietal lobes). The only exception is the caudate. The deficit in the 2-compartment model, is its inability to correctly represent the early temporal structure of the tracer residence. The impact of deviations from the compartmental residue on the quantitation of summary characteristics of FDG metabolism including flux, flow, transit time and extraction were also evaluated. In all cases statistically significant deviations were found. The magnitude of deviations in flux are, not-surprisingly, small (4%).
Conclusions: Our study does not refute the basic biochemistry of the 2-compartment model for FDG, this has been well validated by in-vitro test-tube experimentation. Rather the difficulty arises from the realization that a typical PET region of interest need not behave as a well-stirred test-tube sample. Our result is most likely explained in terms of tissue heterogeneity. The brain regions considered here, similar to those in many PET studies, are relatively large. Within these regions it is likely that neither the vascular delivery characteristics nor local biochemistry will be homogeneous. More generally, the non-parametric residue technique has broader application to the quantification of PET imaging data for which compartmental model structures may not be sufficiently validated.
165. Quantification of [11C]flumazenil binding using a reference tissue model on an experimental high-resolution versus a routine clinical PET scanner
F.H.P. van Velden1, B.N.M. van Berckel1, R.W. Kloet1, F.L. Buijs1, N. Tolboom1,2, S.P.A. Wolfensberger1,3, G. Luurtsema1, A.A. Lammertsma1 and R. Boellaard1
1Department of Nuclear Medicine & PET Research; 2Department of Neurology; 3Department of Psychiatry, VU University Medical Center, Amsterdam, The Netherlands
Background and aims: The High Resolution Research Tomograph (HRRT, CTI/Siemens) is a dedicated human brain positron emisson tomography (PET) scanner. To date, however, only a limited number of human brain studies have been performed using the HRRT, with most studies having been acquired on whole body PET scanners, such as the ECAT EXACT HR+ (CTI/Siemens). Previously, an excellent interscanner comparison (slope: 1.00 to 1.07, intercept fixed to origin) has been reported for an HRRT-HR+ test-retest study using plasma input models.1 In reference regions, however, tracer concentrations are lower and, therefore, more prone to bias. The purpose of the present study was to quantitatively compare HRRT and HR+ scanners in case of a reference tissue model approach.
Methods: Seven healthy volunteers underwent two 60 mins scans (one on each scanner) immediately following administration of 365±34 MBq [11C]flumazenil. All data were histogrammed in 16 time frames with variable frame lengths (same on both scanners). Prior to each emission scan a transmission scan was acquired. HR+ studies were reconstructed using 2D filtered backprojection with Fourier rebinning followed by 5 mm FWHM Gaussian smoothing. HRRT studies were reconstructed using 3D ordered subsets weighted least squares (7 iterations, 16 subsets).2 Reconstructed HRRT images were also smoothed with a 6 mm FWHM Gaussian kernel to match image resolution with that of the HR+ scanner. Pharmacokinetic parameters were generated from time-activity curves, obtained for fourteen different anatomical regions from the dynamic emission frames. Binding potential (BPND) data were generated using a basis function implementation of the simplified reference tissue model.3
Results: One subject showed patient motion (>5 mm) on the HR+ scans and was excluded from comparisons. Another subject had been administered a relatively low dose in both scans (∼311 MBq), causing unreliable reference region input curves (due to noise) and was therefore excluded. For the remaining subjects, BPND values derived from HRRT and HR+ scans showed excellent correlation (r = 0.95±0.02) with a slope of 1.04±0.15 (intercept fixed to origin), which reduced to 0.90±0.10 after resolution matching. Test-retest variability was 14.9±9.4% and 12.3±10.0% without and with resolution matching, respectively.
Conclusions: The somewhat lower values seen in HRRT studies versus those of the HR+ when using reference tissue rather than plasma input models might be due to low counts in the reference region, causing bias in HRRT reconstructions2 and/or inaccuracies in attenuation and scatter corrections. Higher BPND values derived from HRRT scans prior to resolution matching indicate improved quantification due to a reduction in partial volume effects (consistent with1,4,5).
Acknowledgments: Financial supported provided by the Netherlands Organization for Scientific Research (NWO, VIDI Grant 016.066.309).
169. Robust fitting of PET data to improve estimation: application to a [11C]-WAY-100635 study
F. Zanderigo1, R.T. Ogden1,2, C. Chang3, S. Choy1, A. Wong1 and R.V. Parsey1,2
1Department of Molecular Imaging and Neuropathology, New York State Psychiatric Institute; 2Department of Psychiatry, Columbia University, College of Physicians and Surgeons, New York, New York; 3Department of Mathematical Sciences, New Jersey Institute of Technology, University Heights, Newark, New Jersey, USA
Objectives: In the analysis of PET time activity curves (TAC), model fitting is typical accomplished by minimizing the least squares (LS) criterion, which is known to be optimal for data having a Gaussian distribution, but not robust in the presence of outliers (e.g., subject head motion). In contrast, quantile regression (QR) provides robust estimates not heavily influenced by outliers.1 In practice, when no influential points are present LS is used2 otherwise QR performs better. Choosing a fitting method (QR or LS) to use on each TAC based only on that TAC may be termed the data adaptive (DA) approach. We apply our DA method to clinical PET TAC data. We expect that DA will result in reduced variance of estimating model parameters within groups of subjects thus enhancing the power to detect differences among different groups.
Methods: [11C]-WAY-100635 TACs from 28 female and 21 male controls were modeled using a two-tissue constrained compartment model.3 Both LS and DA were applied to estimate the radioligand distribution volume (VT) in 13 regions of interest (ROI). Mean and standard deviation (s.d.) of VT values estimated by LS (meanLS, s.d.LS) and DA (meanDA, s.d.DA), respectively, were calculated for each ROI within each group. Males and females were separated because females have ∼20% higher binding than males (meanLS across ROIs and subjects: 3.84 versus 3.17). The meanDA/meanLS and s.d.DA/s.d.LS ratios were then calculated in each group and each ROI.
Results: For both the groups, Figure 1 reports the meanDA/meanLS (black line) and s.d.DA/s.d.LS (gray line) ratios obtained in each ROI. Application of DA leads to a decrease in the s.d. of VT estimates within females of 14.1% in the best (RN) and 1.1% in the worst case (ORB), respectively. At the same time, the meanDA/meanLS ratios stay close to unity. In cases in which DA fails to improve the accuracy of estimated VT (AMY, INS), the performance of the two approaches is comparable. Results obtained in males show a decrease in s.d. of 7.2% in the best (PIP) and 2% in the worst case (TEM), respectively. The meanDA/meanLS ratios are closer to unity, and DA fails to improve the accuracy of estimated VT just in one ROI (RN).
Conclusions: By reducing the variance of parameters estimates, the statistical power in group analysis and the sensitivity in occupancy studies are increased (i.e. the number of people to be recruited for clinical trials is reduced). This can be implemented without the need for additional hardware and/or image registration algorithms to monitor and correct subject head motion.
251. Benefit of pixel-by-pixel delay correction for measurement of hemodynamic parameters in patients with cerebrovascular disease using O-15 PET
H. Okazawa1, T. Kudo1, M. Kobayashi1, M. Isozaki2, Y. Arai2, T. Tsujikawa1, Y. Fujibayashi1 and T. Kubota2
1Biomedical Imaging Research Center; 2Department of Neurosurgery, University of Fukui, Eiheiji-cho, Japan
Objectives: In calculation of regional cerebral blood flow (rCBF) using O-15 water PET, arterial input function is usually determined by arterial blood sampling, and delay of the input is corrected for radioactivity in the whole brain or at slice-by-slice levels. However, delay of the tracer is assumed to be different between the hemispheres in patients with unilateral arterial occlusive lesions. To calculate precise quantitative hemodynamic parameters, pixel-by-pixel estimation for delay of the tracer arrival time was performed and the delay map was applied for calculation of parameters.
Methods: Fifty-two patients (mean age = 65±11y) with unilateral major cerebral arterial stenoocclusive disease underwent O-15 gas and water PET scans. All patients had occlusion or stenosis (>70%) in the internal carotid or middle cerebral arteries (MCA). For estimation of input function, arterial blood radioactivity was measured continuously using an automatic counter with a constant flow from the brachial artery. The delay of the input was estimated pixel-by-pixel and delay map of tracer arrival time was created. The 3-weighted-integral method was employed for calculation of rCBF and arterial-to-capillary blood volume (V0) with pixel-by-pixel delay correction. Cerebral blood volume (CBV), oxygen extraction fraction (OEF) and cerebral metabolic rate of oxygen (CMRO2) were also calculated from the O-15 gas PET scans using the bolus inhalation method, and asymmetry-index (AI) of these parameters were compared.
Results: Eleven patients showed ipsilateral significant increase in OEF (55.1±4.1%) and all of them showed delay of tracer arrival and significant rCBF decrease (32.4±4.2 mL/100 g/min) in the affected region compared to the contralateral hemisphere. The remaining 41 patients had OEF in normal range or slight increase as a global change and showed no differences in delay between the bilateral hemispheres. V0 showed a slight decrease in the impaired hemisphere of patients with misery perfusion. AI for OEF (OEF-AI) was well correlated with those of delay (delay-AI) and CBF (CBF-AI). Correlation coefficient between OEF-AI and delay-AI (r = 0.70) was better than that between OEF-AI and CBF-AI (r = 0.56). The mean AI values for hemodynamic parameters in all patients were OEF-AI = 1.06±0.09, delay-AI = 1.05±0.06, 1/(CBF-AI) = 1.10±0.11, 1/(CMRO2-AI) = 1.05±0.08, CBV-AI = 1.02±0.16, and V0-AI = 0.89±0.17. Patients with misery perfusion showed significantly higher OEF-AI (1.17±0.12, P<0.05) and longer delay (0.34±0.19 secs, P<0.005) as compared with patients with normal OEF.
Conclusion: Pixel-by-pixel estimation of tracer arrival time provided a delay map for regional differences and precise parametric values for evaluation of hemodynamic status in the patients with cerebrovascular disease. The delay image and delay-AI, correlated well with OEF change, would be able to estimate OEF elevation in impaired circulation in calculation of rCBF and V0 images using O-15 water PET.
284. Unbiased logan graphical analysis using the renormalization method
H. Hontani1, N. Hoshino1, M. Naganawa2, K. Sakaguchi2, M. Sakata3, K. Ishiwata3 and Y. Kimura2
1Department of Computer Science and Engineering, Nagoya Institute of Technology, Nagoya; 2Molecular Imaging Center, National Institute of Radiological Sciences, Chiba; 3Positron Medical Center, Tokyo Metropolitan Institute of Gerontology, Tokyo, Japan
Objective: The Logan Graphical Analysis (LGA) is used for imaging a distribution volume VT. For LGA, we compute a set of {(x(t), y(t))} from the measured time-activity curves in tissue (tTAC) and plasma (pTAC) to find a best-fitting line y(t) = α x(t)+β. (Equation 1).
Here, x(t) and y(t) are defined as a ratio of an integrated pTAC over tTAC and an integrated tTAC over tTAC, respectively. As known,1 linear regression (LR) underestimates VT and its unbiased estimator is expected.
Renormalization Method (RM)2 enables an unbiased maximum likelihood estimation under the existence of inhomogeneous noises both in x and y by successive evaluation of bias. In this study, the applicability of RM to LGA was investigated.
Methods: Let Xt = (x(t),y(t),1)T and U = (u1,u2,u3)T. Then, we can rewrite (Equation 1) as XtTU = 0, where ‖U‖ = 1 and VT = −u1/u2. Let Ct denote the covariance matrix of the noise of Xt. The maximum likelihood estimates of ui minimize JMLE(U) = ΣtWt(U) (XtTU)2, where Wt(U) = 1/(UTCtU). Though, the perturbation theorem tells us that the estimates become biased.
RM removes the bias by iteratively minimizing JREN(U) instead of JMLE: JREN(U) = ΣtWt(U) {(XtTU)2−UTCtU}, where the last term compensates the bias. In RM, the covariance matrix Ct should be given, and it is unknown in advance. Thus, a set of voxel-based noisy TACs were simulated using physiologically plausible kinetic parameters, and the mean of Ct was calculated from the set of simulated TACs.
We applied RM and LM to synthesized tTACs and to real one of [11C]SA4503-PET. For generating the synthesized data, we simulated a set of voxel-based tTACs using a measured pTAC and the rate constant of [11C]SA4503.3
Results: The simulation results are summarized in Figure (A). RM plotted in red was almost identical (y = 0.99x+0.23, r2 = 1.00), and LR plotted in blue showed the underestimation especially in large VT (y = 0.70x+6.14, r2 = 0.94). The estimation of deviation was larger than that of LM. However, RM successfully suppressed the bias.
Results.
The figures (B) and (C) show the results of imaging of VT obtained from the real data by RM and by LR, respectively. For the estimation, t* was set to be 15 mins post-injection. The computational time for RM was 10 mins for 60 thousands voxels. RM gave brighter images than LR, and improved their contrast.
Conclusions: For computing unbiased estimates, we introduced RM. We estimated the average of each Ct based on simulations. Simulation results showed that RM suppresses the bias and has the potential to realize bias-free parametric imaging of VT.
312. Comparison of in vivo kinetics of 18F-fallypride and 11C-FLB-457
N. Vandehey1, J. Moirano1, D. Murali1, A.K. Converse2, J. Engle1, R.J. Nickles1, J. Mukherjee3, M. Schneider4, J. Holden1 and B. Christian1
1Department of Medical Physics; 2Waisman Laboratory for Brain Imaging and Behavior, University of Wisconsin—Madison, Madison, Wisconsin; 3Department of Psychiatry & Human Behavior, University of California Irvine, Irvine, California; 4Harlow Center for Biological Psychology, University of Wisconsin—Madison, Madison, Wisconsin, USA
Objectives:18F-Fallypride (FAL) and 11C-FLB-457 (FLB) are two commonly used PET radioligands for imaging extrastriatal dopamine D2/D3 receptors. Though both tracers provide excellent visualization of receptor binding in regions with low D2/D3 receptor density (e.g. cortex), there are differences in their in vivo kinetics that may affect sensitivity for measuring subtle changes in receptor binding. Focusing on regions of low binding, we made a direct comparison of the kinetics of FAL and FLB in the rhesus monkey.
Methods: Single scan, multiple-injection studies; which includes high, medium, and low specific activity injections, were performed on two male rhesus monkeys with both FAL and FLB using a microPET scanner. Arterial blood samples were taken and measured for both whole blood and parent compound in plasma, which were used as the input functions to the kinetic model. Dynamic ROI data drawn on the cerebellum, occipital cortex, thalamus, and substantia nigra were fit using a hot-cold compartment model in the COMKAT environment to obtain estimates of the transport (K1, k2) and binding (kon, Bmax, koff) parameters.
Results: Measurement of the plasma-tissue transport constants, K1 and k2 and their ratio, reveals that FLB (K1 = 0.49±0.11, k2 = 0.59±0.10, K1/k2 = 2.69±0.51) has a considerably higher free space distribution volume (K1/k2) than FAL (K1 = 0.47±0.12, k2 = 0.23±0.05, K1/k2 = 0.82±0.21) averaged over the regions, due to a reduced tissue to plasma efflux constant (k2). Both tracers have comparable koff values of 0.025 (FLB) and 0.023 (FAL) but FLB has a slightly higher rate of association (kon = 0.087±0.02) compared to that of FAL (kon = 0.070±0.02), yielding a lower equilibrium dissociation constant, KD. Within the first injection period of these scans (up to 90 mins), FAL reaches pseudo-equilibrium in regions of mid-to-low binding, but FLB does not. Input function analysis showed that FAL is retained at a higher concentration in plasma for a longer period of time than FLB, having a 3 × higher concentration after 60 mins.
Conclusions: These findings in the rhesus monkey are consistent with human data for free space distribution volume (K1/k2).1,2 The slightly higher affinity (1/KD) of FLB may be advantageous for measurement of baseline cortical D2/D3 binding, however the higher k2 of FAL may enhance the sensitivity for detecting changes in endogenous dopamine release due to its higher rate of clearance out of the free space.
317. A new spectral analysis method for measuring regional rates of cerebral protein synthesis in L-[1-11C]leucine PET studies
M. Veronese1, A. Bertoldo1, S. Bishu2, A. Unterman2, G. Tomasi3, C.B. Smith2 and K.C. Schmidt2
1Department of Information Engineering, University of Padova, Padova, Italy; 2Section of Neuroadaptation & Protein Metabolism, National Institute of Mental Health, Bethesda, Maryland; 3Department of Diagnostic Radiology, Yale University, New Haven, Connecticut, USA
Objective: Due to the limited spatial resolution of PET, a region of interest (ROI) likely contains a heterogeneous mixture of tissues. To date, however, we have been applying a homogeneous tissue kinetic model (HOM) for determination of regional rates of cerebral protein synthesis (rCPS) from L-[1-11C]leucine PET data in humans.1 Since Spectral Analysis (SA)2 applies to heterogeneous as well as homogeneous tissues, it might be an alternative quantitative method. In this study, we developed and tested a new SA algorithm to estimate rCPS at the ROI level.
Methods: Because previous SA methods2,3 did not consistently produce reliable estimates when applied to L-[1-11C]leucine data, we developed a new SA iterative filtering (SAIF) method. SAIF uses prior information concerning irreversibility of trapping of the tracer as well as components that cannot be distinguished from blood; intermediate components are assumed to reflect tissue reversible compartments. Performance of SAIF was compared to previous SA methods2,3 in simulation studies. Bias and precision (coefficient of variation (CV)) of the estimates were used as performance indices. We also compared three methods of analysis of L-[1-11C]leucine PET data measured in 9 healthy male subjects:
SAIF;
HOM applied at the ROI level with a nonlinear least squares algorithm; and
HOM applied at the voxel level with a basis function method.
Goodness-of-fit was assessed by the weighted residual sum of squares.
Results: In simulation studies SAIF yielded the best fit, lowest bias (<2%), and lowest CV (6%), compared with previous SA methods. In measured data, rCPS estimated with SAIF agreed with the average of rCPS values estimated by using HOM at the voxel level, but differed substantially from ROI-based rCPS estimated by HOM (Table). This is consistent with the presence of a greater degree of tissue heterogeneity in the ROI, compared to individual voxels, that is not accounted for with the homogeneous tissue kinetic model. In most regions CVs with SAIF were somewhat higher than when a fixed kinetic model was used, probably due to the increased number of parameters estimated with SAIF.
rCPS (nmol g−1 min−1)
ROIs
Homogenous model ROI analysis
Homogenous model voxel analysis
SAIF ROI analysis
Whole brain
1.85±0.10
1.61±0.08
1.59±0.13
Frontal cortex
2.16±0.13
1.90±0.08
1.85±0.11
Thalamus
1.72±0.10
1.54±0.09
1.49±0.06
Corona radiata
0.98±0.07
0.84±0.06
0.88±0.08
Values are mean±s.d. for 9 subjects.
Conclusions: SAIF demonstrated low bias and good precision. SAIF accounts for tissue heterogeneity, unlike HOM, and may be a useful and robust method for estimating rCPS at the ROI level.
Supported by IRP/NIMH/NIH.
348. Spatio-temporal mapping of neurotransmitter activation using a linearization of the parametric ntPET model
M.D. Normandin1,2, R.D. Badgaiyan3 and E.D. Morris1,2,4
1Department of Radiology, Indiana University School of Medicine, Indianapolis; 2Weldon School of Biomedical Engineering, Purdue University, West Lafayette, Indiana; 3Division of Nuclear Medicine, Masschusetts General Hospital, Boston, Massachusetts; 4Department of Biomedical Engineering, Purdue School of Engineering and Technology, Indiana University-Purdue University Indianapolis, Indianapolis, Indiana, USA
Objectives: We recently introduced,1 characterized,2 and validated3 parametric ntPET (p-ntPET), a model for estimation of neurotransmitter kinetics from dynamic PET data. Here, we demonstrate voxel-by-voxel analysis with lp-ntPET, a linearization of the original model.
Methods: lp-ntPET extends the LSRRM4 using basis functions to estimate the time course of neurotransmitter activation. The basis function that yields the best model fit is mapped to the optimal neurotransmitter response. Bases were chosen to allow estimation of response onset, peak time, and sharpness. This model includes four estimated parameters—R1, k2 (interpreted in the two-tissue model sense), k2a (apparent efflux rate in the one-tissue model sense), and response magnitude—and is thus called lp-ntPET4.
Mathematical simplifications were considered to promote stable performance with noisy voxel data. Following the approach of SRTM25 and MRTM2,6 analysis was carried out using a global value for the reference region k2. This model (lp-ntPET3) has three estimated parameters. The model was further simplified by fixing baseline BP (= [k2/k2a]−1) at each voxel according to prior analysis of pre-activation data. This model (lp-ntPET2) has two estimated parameters.
Response t maps overlaid on emission image.
A human subject was injected with a bolus of [11C]raclopride and scanned for 45 mins. An emotional processing task hypothesized to elicit striatal dopamine release was initiated at 26 mins. The dynamic data set was analyzed with lp-ntPET and subsequent simplifications as described above.
Results: Results of lp-ntPET4 analyses were variable and showed limited activation in striatum. Regions with high blood flow and low specific binding were prone to false positives. lp-ntPET3 and lp-ntPET2 indicated bilateral striatal activation with stronger responses in left versus right caudate and putamen, consistent with results obtained using LSRRM. The onset of significant responses was generally synchronized with task initiation and peaked 1 to 2 mins thereafter, in agreement with the anticipated response. Insignificant responses had little temporal coherence. Model simplifications reduced noise in the spatial distribution of activation, increased statistical power, and enhanced the temporal consistency of significant responses.
Conclusions: The efficiency of lp-ntPET facilitates parametric image analysis, an intractable task for the computationally demanding p-ntPET. Simplifications of the lp-ntPET model improved the quality of results. Such simplifications are known to enhance precision at the expense of increased bias; simulation studies are under way to investigate this relationship for lp-ntPET.
Acknowledgements: Society of Nuclear Medicine Student Fellowship (MDN) and R21 AA015077 (EDM).
391. k3 is a better parameter than binding potential as a quantitative indicator of beta amyloid by [11C]PIB PET
T. Ohya, K. Fukushi, H. Shimada, K. Sato, H. Shinoto, M. Miyoshi, S. Hirano, N. Tanaka, T. Ota and T. Irie
Molecular Imaging Center, National Institute of Radiological Sciences, Chiba, Japan
Aims: Quantitative measurement of beta amyloid (Aβ) in brain of Alzheimer diseases (AD) is necessary for monitoring the progression of the Aβ deposition and the efficiency of the Aβ immunotherapy. Though Logan-distribution volume ratio (DVR) is currently used for the diagnosis of AD by PET, DVR is not adequate as a quantitative indicator for Aβ, because it is a complex parameter. DVR is based on the assumption that K1/k2 is almost equal within whole brain. Thus, we noted the two parameters, k3 and binding potential (BP = k3/k4), which are proportional indicators to the Aβ deposition, and examined their diagnosis power by clinical study. The stability of k3 and BP as a quantitative indicator and the effect of k4 condition on their stability were examined by simulation study.
Methods: Ninety minutes [11C]PIB PET studies were performed in 22 volunteers with or without AD. The metabolite-corrected input function was measured in all cases. The PET data were analyzed by two-tissue compartment model with four kinetic parameters (K1∼k4). The t-test with two parameters (k3, BP) between NC (N = 10) and AD (N = 12) groups was performed on Fine-SRT ROI (FUJIFILM RI Pharma Co.) (50 regions) and voxel ROIs, and their diagnosis power were compared by the number of the positive ROIs, where k3 or BP significantly increased (t-test P<0.05). In the Monte Carlo simulation study, tissue-time activity curves of NC and AD with statistical noise were made as follows. The three parameters were according to the typical NC or AD conditions from the clinical data: k3 = 0.018 min−1 (NC), 0.040 (AD), K1 = 0.18 mL/g/min, k2 = 0.18 min−1 and input function were fixed. The simulations were performed under various k4 conditions, and more than 2000 parameter results were obtained in each k4 condition. The coefficient of variation (CV) of k3 and BP were calculated from them.
Results: In the Fine-SRT ROIs study, the k3 showed higher diagnosis power than BP to discriminate between NC and AD: 48 positive ROIs by k3 and 40 ROIs by BP. Similar result was obtained with the parametric image for voxel-ROIs. From the simulation study, CVs of k3 and BP were approximately the same under AD condition (k3 = 0.040, k4≈0.025), whereas CV of k3 was sufficiently lower than that of BP under NC condition (k3 = 0.018, k4≈0.020) (Figure 1). These results supported the clinical result. The CV values of k3 and BP were much dependent on the k4 conditions, that is, BP became unstable when the k4 became smaller.
Conclusions: The k3 may be a better quantitative indicator for Aβ monitoring than BP in the case of [11C]PIB PET, though the selection of indicator for k3 or BP is dependent on the k4 value of the utilized PET tracer.
398. Simplified quantification of adenosine A2A receptor with [11C]TMSX PET and graphical analyses
M. Naganawa1,2, M. Mishina2,3, M. Sakata2, K. Oda2, K. Ishii2, K. Ishiwata2 and Y. Kimura1,2
1Molecular Imaging Center, National Institute of Radiological Sciences, Chiba; 2Positron Medical Center, Tokyo Metropolitan Institute of Gerontology, Tokyo; 3Neurological Institute, Nippon Medical School Chiba-Hokusoh Hospital, Chiba; Japan
Objectives: [11C]TMSX is one of the few available radioligands to quantify adenosine A2A receptor (A2AR).1 Although A2ARs are rich in the basal ganglia and poor in other regions, [11C]TMSX binds to not only classical A2ARs but also atypical sites. Therefore, it is difficult to find a reference region. In previous studies with arterial input function, we selected the centrum semiovale (CSO) as a reference region.2,3 The purpose of this study was to investigate the availability of graphical analyses for human [11C]TMSX studies with region of interest (ROI) analysis.
Methods: Seven healthy subjects were included in this study. The average dose was 570±86 MBq and the specific activity was 39±21 GBq/μmol. Dynamic PET scans were acquired for 1 h with a SET-2400W (Shimadzu, Japan). A total of 25 ROIs was manually delineated on the summed PET image: CSO (reference), anterior and posterior putamen, caudate, midbrain, posterior cingulate, cerebral cortical areas, and cerebellum. The regional time-activity curves (TACs) were analyzed using one- and two-tissue compartment models (1T, 2T), Logan graphical analysis (LGA) and multilinear analysis (MA1) with a metabolite-corrected input function. TACs were also analyzed using reference Logan graphical analysis (LGAR). The clearance rate constant of the reference region (k2′) for LGAR was fixed to be 0.12 min−1, the mean value from the 1T model. For all graphical methods, t* was set to be 20 mins post-injection.
Results: Metabolism of [11C]TMSX was very slow with parent fraction of 0.98±0.01 at 3 mins and 0.92±0.05 at 60 mins. The ROIs were well described by 2T model with a constraint of VND from the 1T estimation of CSO. The values of VT ranged from 0.58±0.16 (CSO) to 1.25±0.29 mL/cm3 (anterior putamen), with K1 values of 0.068±0.014 to 0.30±0.07 mL/min/cm3. The estimated VT of graphical analyses and 2T matched well (VT(LGA) = 0.99 VT(2T)+0.01 (r2 = 0.99), VT(MA1) = 1.01 VT(2T)+0.01 (r2 = 0.99)). The estimated VT in CSO was 5% higher in graphical analyses than in 1T model. As a result, the estimated BPND of LGA and MA1 showed smaller values compared with compartment model analysis (BPND(2T) = 0.83BPND(MA1)+0.06 (r2 = 0.92)). The LGAR provided similar results (BPND(LGAR) = 0.99BPND(MA1)−0.01 (r2 = 0.99)).
Conclusions: All graphical analyses provided similar estimates of VT and BPND for ROI curves. This study suggests that LGAR is available for quantification of [11C]TMSX ROI analysis. The difference of BPND values in graphical analyses and compartment analysis is due to the estimates of VND value. The suitability of CSO as a reference region should be further investigated and the noise-reduction method is required for LGAR imaging.
Acknowledgments: This work was funded by Grants-in-Aid for Scientific Research (B) No. 16390348, (C) No. 17590901 and No. 20591033 from JSPS.
547. k3 imaging of [11C]PIB PET using the three-parameter estimation in the short scan-time (TPESS) method
H. Shimada1,2, K. Fukushi3, K. Sato1, H. Shinotoh1, M. Miyoshi1, S. Hirano1, N. Tanaka1, T. Ota1, H. Ito1, S. Kuwabara2, T. Irie3 and T. Suhara1
1Department of Molecular Neuroimaging, Molecular Imaging Center, National Institute of Radiological Sciences; 2Department of Neurology, Chiba University; 3Molecular Probe Group, Molecular Imaging Center, National Institute of Radiological Sciences, Chiba, Japan
Background and aims: Quantitative analysis of amyloid imaging agents may be necessary for assessment of the longitudinal changes in amyloid deposition and effects of anti-amyloid therapy on brain amyloid deposition in patients with Alzheimer's disease (AD). The 2-tissue compartment 4-parameter (2T-4k) model analysis of reversible ligands, however, needs long scanning period such as 90 to 120 mins. It is often difficult to perform PET scans without head movements in such long scanning period and measure the plasma input function accurately with carbon-11 labeled ligands, resulting in unstable and inaccurate estimates of parameters. In [11C] PIB PET, our simulation study showed that the brain time activity curve is almost independent on k4 during the first 30 mins, and thus k3, the bimolecule association rate (konBmax, min-1), could be estimated in the first 30 mins by three-parameters model analysis excluding k4 parameter. Deliberately ignoring k4 stabilizes k3 estimates in the first 30 mins. Here we call this method as the three-parameter estimation in the short scan-time (TPESS) method. In this study, we compared the TPESS method with the conventional 2T-4k model analysis by making k3 images in both methods.
Methods: Participants were 12 patients with AD (mean age: 70.0±9.6 years; male/female: 5/7; MMSE: 20.2±4.2) and age-matched 11 healthy controls (mean age: 65.2±12.4 years; male/female: 3/8; MMSE: 28.0±2.2). A dose (about 370 Mq) of [11C]PIB was intravenously injected and sequential PET scans were performed for 90 mins with arterial blood sampling. Kinetics of [11C]PIB in 90 mins were analyzed using the 2T-4k compartment model analysis and the TPESS method, in which the scan duration for analysis was shortened to 30 mins. K3 values, an index of amyloid deposition, were compared between AD group and healthy control group in both methods, and SPM5 analysis was performed.
Results: There was the extensive increase in k3 values in AD compared with normal controls both in the 2T-4k model analysis and the TPESS method (P<0.01, FDR corrected). The increase in k3 was more extensive and significant in the TPESS method. Mean cortical k3 values in control and AD groups were 0.0238±0.0077 and 0.0414±0.0273 in the 2T-4k model analysis, and 0.0125±0.0034 and 0.0263±0.0089 in the TPESS method respectively.
Conclusions: The TPESS method yields similar SPM analysis results as in 2T-4k compartment model analysis. This TPESS method is timesaving and thus applicable to clinical AD research.
550. Evaluation of sensitivity of kinetic macro-parameters to changes in [18F]fluorodopamine storage and metabolism in the striatum
K. Matsubara1,2, H. Watabe2, Y. Ikoma1,2, T. Hayashi2, K. Minato1 and H. Iida2
1Lab. of Technology of Radiological Science, Nara Institute of Science and Technology, Ikoma; 2Department of Investigative Radiology, National Cardiovascular Center, Suita, Japan
Objectives: Dopa utilization and dopamine retention in the striatum can be investigated by [18F]F-dopa PET study. [18F]fluorodopamine synthesized from [18F]F-dopa is stored in the intracellular vesicles or metabolized to [18F]FDOPAC and [18F]FHVA, which can diffuse out of brain tissue. In other hand, several approaches has been proposed to analyze [18F]F-dopa PET data to obtain kinetic macro-parameters such as K, Vd, and kloss which represent the net kinetics of [18F]F-dopa as above.1,2
It was reported that reduction for dopamine storage and acceleration for dopamine metabolism to DOPAC and HVA occur in brain of patients with Parkinson's disease.3,4 According to these findings, we hypothesized that the parameter sensitive to the storage reduction and metabolic acceleration might be a good index for diagnosing Parkinson's disease. For the evaluation of sensitivity for estimated macro-parameter to change for dopamine storage and metabolism, we simulated the tissue time-activity curves (TACs) under variable change in dopamine storage or metabolism using a model describing the detail pathway of dopamine kinetics in tissue involving storage to the vesicle and metabolism to DOPAC and to HVA (Figure). We estimated macro-parameters from these TACs by conventional analytical methods and evaluated sensitivity of these parameters.
Model with the detail pathway.
Methods: First, we generated a standard TAC as normal condition, which mimics actual striatal TAC obtained by 120 mins [18F]F-dopa PET scan in normal monkey (macaca fascicularis, body weight:3.8 kg). Input function for F-dopa and OMFD obtained from blood sampling and metabolite analysis were used. Second, we simulated TACs with increasing or decreasing rate constants for dopamine storage (k7) or metabolism to DOPAC (k9dopac). Third, simulated TACs were analyzed by conventional graphical analysis, Patlak and Logan analysis, and multilinear method proposed by Kumakura et al.2 Finally, for evaluation of sensitivity, difference between macro-parameters for a standard and altered TAC, which is defined as %change, were calculated.
Results: In case to decrease k7 or increase k9dopac, %change of kloss estimated by the multilinear method was the largest among the macro-parameters (64.4% for kloss and 21.6 to 37.2% for other parameters in case that k7 decrease by 80% to standard condition).
Conclusion: Our results suggest that kloss estimated by the multilinear analysis has better performance to detect changes in dopamine storage and metabolism sensitively than other macro-parameters. These findings agree with those in previous clinical study in Parkinson's disease.2,5
664. New strategy for k3 quantification of reversible PET tracers
K. Sato1,2, K. Fukushi1, N. Tanaka1,3, H. Shimada1, H. Shinotoh1, S. Hirano1, T. Ota1, M. Miyoshi1, T. Ohya1 and T. Irie1
1Molecular Imaging Center, National Institute of Radiological Sciences; 2Department of Psychiatry, Teikyo University Chiba Medical Center, Chiba; 3Department of Neurosurgery, Tokyo Woman's Medical College Daini Hospital, Tokyo, Japan
Objectives: In recent years, reversible ligands are actively performed in the neuroreceptor PET studies. Full compartment analysis with nonlinear least squares (NLS) algorithm based on tissue 2-compartment reversible model is generally impractical, because the computation is time consuming and parameter estimation is uncertain due to excessive number of parameter. In this study, pre-equilibrium phase, where tissue time-radioactivity curve (TAC) is almost independent of k4, was appropriately extracted from total dynamics of reversible tracer. Time-radioactivity data was then analyzed by NLS analysis with three-parameter to precisely determine k3 value, which is of greatest physiological interest.
Methods: The properties of this method were explored by Monte-Carlo simulation. To reflect the real situation, parameters and input function referred the experimental results of [11C]Pittsburgh Compound-B (PIB), the ligand for beta-amyloid imaging, in human study. Here, partial diiferential derivative of TAC with respect to each parameter was defined as the sensitivity of parameter. Difference (delta) was practically used in place of partial differentiation, and squares residual error was remarked, since the NLS algorithm includes the minimization of sum of squares error. Figure shows the sensitivity curve representing squares residual of TAC, (delta TAC)2 with respect to delta k for typical TAC of [11C]PIB. Our method requires time reduction in data analysis for the contribution of k4 in TAC to be ignored. However, analysis time must be long enough for precise k3 estimation. Considering this trade-off based on sensitivity curve, the appropriate pre-equilibrium phase was determined. The time length necessary for this situation is about 30 mins.
Results: In the simulation, k3 values were precisely calculated by early phase analysis with CV of 5 to 10%. Significant minus-bias of around −30% (−25% to −35%) that related to disregard of k4 was shown in k3 estimates. Nevertheless, the underestimates were comparable in degree, across 1 order of magnitude of k3 values, from 0.010 to 0.100 min−1. Additional examination of k3 change in the clinical follow-up study showed relatively small bias in the ratio of k3 change (0 to −20%) compared to those in k3 value itself. Once the method was developed, it was pilot tested on the study with [11C]PIB.
Conclusions: In the pre-equilibrium phase analysis, the advantages of precise k3 measurement and significantly shortened scan duration time were provided by giving up the calculation of k4 that is rather of little interest and is essentially difficult to quantify. The new method includes the optimization of time range for k3 estimation, referring to the sensitivity curve. Though this method is more suitable for the reversible tracer with relatively small k4, many reversible tracers are applicable to this method.
891. Modeling blood/tissue transfers at the voxel scale for blood flow quantification with PET
I. Billanou1, S. Lorthois1, M. Quintard1 and P. Celsis2
1Institut de Mécanique des Fluides de Toulouse, UMR CNRS/INPT/UPS 5502; 2Inserm U825 ‘Cerebral Imaging and Neurological Handicap’, CHU Purpan, Toulouse, France
Objectives: The current practice for analyzing dynamic PET scans is based on compartmental models, which relate the kinetics of the measured activity and the function under study.1 In particular, cerebral blood flow modifications during activation studies are determined by considering two well-mixed compartments: an arterial compartment (AC) and a tissue compartment (TC). The Renkin-Crone theoretical description of the exchange between these two compartments, which introduces kinetic rate constants related to the flow rate, considers one single capillary and its surrounding tissue, modeled as two coaxial cylinders (Krogh cylinder). The obtained rate constants, which implicitly describe exchanges at local scale, are nevertheless used in the compartmental model to describe exchanges between AC and TC at the voxel scale, i.e. macroscopic scale. Moreover, the compartmental approach is based on extremely simplifying assumptions. For example, axial diffusion (concentration gradients in the longitudinal direction) is neglected. The limits of the compartmental approach have been recognized and may explain some illogical results obtained by PET.2 In this context, our goal is to propose an improved kinetic model for diffusible tracers (H2O15 PET) valid at the voxel scale.
Methods: Based on the Krogh cylinder geometry, we use the volume averaging method3 to obtain the coefficients describing the blood/tissue exchange at the voxel scale, accounting for axial diffusion. In other words, the purpose is to replace, in the usual compartmental model, the tissue compartment (which also includes microciculation) by a fictitious homogeneous compartment (Figure 1A). The validity of this approach is tested by comparisons with complete solutions of the transport equations at capillary scale using numerical simulations.
Results:Figure 1B displays the concentration field of H2O15 in a 10 μm diameter capillary and its surrounding tissue at the beginning and end of injection, and 0.034 secs after the end of injection (clearance) for a Gaussian bolus of H2O15 injected at the capillary inlet. The results highlight the presence of non-negligible longitudinal gradients. When averaged over a cross section, the longitudinal evolution of these concentration fields are correctly described by the volume averaged model.
Conclusions: The blood/tissue transfers at the voxel scale are properly described by this model. Its relevance for PET in comparison with the classical approach will be discussed.
Acknowledgments: This work is supported by the Skin Research Centre of Pierre Fabre Dermocosmetique and by a doctoral fellowship from ‘Region Midi Pyrénées’.
931. Simplified quantification of adenosine A1 receptor with [11C]MPDX and graphical analysis
Y. Kimura1,2, M. Naganawa1,2, M. Mishina2,3, M. Sakata1,2, K. Oda2, K. Ishii2 and K. Ishiwata2
1Molecular Imaging Center, National Institute of Radiological Sciences, Chiba; 2Positron Medical Center, Tokyo Metropolitan Institute of Gerontology, Tokyo; 3Neurological Institute, Nippon Medical School Chiba-Hokusoh Hospital, Chiba, Japan
Objectives: [11C]MPDX is a clinically available radioligand for quantification of the adenosine A1 receptor (A1R),1 and its kinetic analysis was reported.2 The purpose of this study was to investigate an applicability of simplified graphical algorithms for human studies based on ROI analysis.
Methods: Six normal subjects were included in this study. The dose was 610±126 MBq, and the specific activity was 53±37 MBq/nmol. Dynamic PET scans were performed for 1 h using SET-2400W (Shimadzu, Kyoto, Japan) in two-dimensional mode with arterial blood sampling. A total of 24 ROIs was placed manually on the summed images: cerebellum (reference), pons, midbrain, caudate, putamen, thalamus, posterior cingulate, and frontal, temporal, occipital, and parietal lobes. The ROI-averaged tissue time activity curves (tTACs) were analyzed using five algorithms: one- and two-tissue compartment models (1T and 2T), Logan graphical analysis (LGA), and a reference LGA with or without a clearance rate of a reference region (k2) (LGAR-k2 and LGAR-Nok2).3 For LGAR-k2, the k2 in a reference region was derived from the cerebellum using 1T. All graphical analyses were applied after 20-mins post-injection.
Results: The metabolism of [11C]MPDX was slow; the parent fraction was 0.74±0.07 at 60 mins after injection. The tTACs were well described using 2T with a constraint of VND that was estimated from the reference region using 1T. The values of VT and K1 ranged from 0.52±0.11 (cerebellum) to 0.72±0.13 mL/cm3 (posterior putamen) and from 0.10±0.03 (cerebellum) to 0.13±0.04 mL/min/cm3 (posterior putamen), respectively. The k2 in reference regions were (0.45±0.11 min−1). The kinetics in reference regions was described using 1T model, and 2T fitting could not give us reasonable VND, VND(1T) = 0.98 × VND(LGA)0.06 (r2 = 0.96). In the graphical analyses, BPND using the simplified algorithms of LGAR-k2 and LGAR-Nok2 matched well with those using LGA: BPND(LGAR-k2) = 0.99 × BPND(LGA)+0.01 (r2 = 0.95) and BPND(LGAR-Nok2) = 0.96 × BPND(LGA)0.01 (r2 = 0.94).
Conclusions: Both the simplified algorithms based on Logan graphical analysis without an arterial input function provided corresponding BPND with those of LGA using an input function in ROI-based analysis. These results implied that LGAR-Nok2 was applicable for the quantification of A1Rs using [11C]MPDX.
Acknowledgement: This work was supported by Grants-in-Aid for Scientific Research (B) No. 20390333, (B) 20390334 and (B) No. 16390348.
943. A consistent graphical analysis method to quantify non-equilibrium tracer kinetics in ligand-receptor dynamic PET studies
Y. Zhou, W. Ye, J.R. Brasic, M. Guevara and D.F. Wong
Department of Radiology, Johns Hopkins University, Baltimore, Maryland, USA
A new graphical plot (nGP) was recently proposed to quantify tracer kinetics that attain equilibrium relative to plasma input as t≥t* in radioligand receptor dynamic PET studies.1
Objectives: The objectives of the study are:
to extend the nGP for quantification of non-equilibrium tracer kinetics relative to plasma input and
to evaluate the extended nGP method by human dynamic PET studies.
Methods: A bi-graphical plot (BiGP), consisting of the Patlak plot and the nGP, was first derived to estimate distribution volume (DV). [11C]CFT (19), [11C]diprenorphine ([11C]DNP) (26), and [11C]MDL100,907 ([11C]MDL) (10) human dynamic PET studies with metabolite-corrected plasma input were studied for dopamine transporter, opiate receptor, and serotonin receptor 5HT2A imaging, respectively. Eleven regions of interest (ROIs) were defined on coregistered MRIs and applied to dynamic PET for ROI time activity curves (TACs). The ratio of ROI TAC to plasma input was calculated to evaluate whether the ROI TAC would attain relative equilibrium states. For comparison, the BiGP, the Logan plot (LP), and a two-tissue compartment model (2TCM) with nonlinear regression were applied to ROI TACs for DV estimation. The BiGP with a spatial constraint algorithm was developed to generate DV images, and the ROI estimates from DV images were compared to the estimates from ROI TACs. The DV images generated by the BiGP and the LP were compared.
Results: The ratio of ROI TAC to plasma input increased linearly as t≥40 mins in all ROIs except cerebellum in [11C]CFT and occiptal cortex in [11C]DNP. The plots from BiGP and the LP attained straight lines for t≥40 mins. The DVs obtained by the BiGP were linearly correlated (R2 = 0.99) to those estimated by the 2TCM, and the LP from ROI TACs with very comparable magnitudes (Figure 1). For the ROI estimates from DV images using the BiGP, DV(images) = 1.00DV(ROI TACs)−0.22, R2 = 0.97, for all ROIs of the 3 tracer studies. In contrast to the DVs from ROI TACs, the ROI estimates from the DV images generated by the LP were underestimated ((13±2)%, (51±8)%) in (cerebellum, striatum), ((22±5)%, (68±8)%) in (occipital, thalamus), and ((39±12)%, (68±9)%) in (cerebellum, lateral temporal cortex) for the [11C]CFT, the [11C]DNP, and the [11C]MDL, respectively. The computational time for generating DV images was reduced by 71% on average by the BiGP in contrast to the LP.
DV images and Mean (s.d.) of ROI DVs.
Conclusions: The BiGP is a consistent and computationally efficient graphical method for estimating DV from non-equilibrium tracer kinetics in both ROI and pixel levels.
This work was supported in part by NIH grants MH078175, DA000412, and AA012839.
1026. Measurement of [18F]MPPF binding in rat hippocampus and cortex by PET
K. Tennessen, E. Ahlers, J. Moirano, D. Murali and A. Converse
Waisman Brain Imaging Lab, University of Wisconsin—Madison, Madison, Wisconsin, USA
Objectives: [18F]MPPF is a radiotracer used in positron emission tomography (PET) to image the location of serotonin-1A (5- HT1A) receptors. As part of an attempt to develop a noninvasive assay to measure in vivo changes in synaptic serotonin levels we imaged [18F]MPPF binding in rats.
Methods: Rats were anesthetized with isofluorane, catherterized for venous injection, and positioned four at a time in a microPET P4 small animal scanner. Following a transmission scan, [18F]MPPF was administered and emission scans were performed. Dynamic images were reconstructed with attenuation and scatter correction and aligned to template regions of interest (ROI) and time activity curves (TAC) were determined. Hipppocampus and cortex to cerebellum distribution volume ratios (DVR) were calculated using the Logan graphical reference tissue method.
Results: At the regions of highest binding, the observed DVRs for the period 20 to 40 mins post injection of [18F]MPPF were hippocampus: 2.04±0.159 and anterior cortex: 1.95±0.124 (mean±s.d., n = 8).
Conclusions: Binding of the 5-HT1A receptor ligand [18F]MPPF has been measured by PET in rat hippocampus. Future work will focus on attempts to measure pharmacologically induced serotonin release.
Measurement of [18F]MPPF binding.
References
1.
Smith. JCBFM2005;25:629–40.
2.
Schmidt. JCBFM2005;25:617–28.
3.
Bishu. JCBFM2008;28:1502–13.
4.
Gunn. Neuroimage1997;6:279–87.
5.
PhelpsMEHuangSCHoffmanEJSelinCSokoloffLKuhlDE. Tomographic measurement of local cerebral glucose metabolic rate in humans with [F-18]2-Fluoro-2-deoxy-D-glucose: Validation of method. Ann Neurol1979;6:371–88.
6.
OstergaardLCheslerDAWeisskoffRMSorensenAGRosenBR. Modeling cerebral blood flow and flow heterogeneity from Magnetic Resonance residue data. Magn Reson Med1999;19:690–9.
7.
MeierPZierlerKL. On the theory of the indicator-dilution method for measurement of blood flow and volume. J Appl Physiol1954;6:731–44.
8.
Van Velden. IEEE MIC Abstract2008; M06–1.
9.
Van Velden. J Nucl Med2009;50:72–80.
10.
Gunn. NeuroImage1997;6:279–87.
11.
Van Velden. IEEE NSS Conf Rec2006;5:3097–9.
12.
Leroy. J Nucl Med2007;48:538–46.
13.
Koenker and Bassett. Econometrica1978;46:33–50.
14.
Cook. Technometrics1977;19:15–8.
15.
Parsey. J Cereb Blood Flow Metab2000;20:1111–33.
16.
OhtaH. J Cereb Blood Flow Metab1996;16:765–80.
17.
OkazawaH. J Nucl Med2003;44:1875–83.
18.
OkazawaH. Eur J Nucl Med Mol Imaging2007;34:121–9.
19.
Slifstein. J Nucl Med2000;41:12.
20.
Kanatani. IEEE PAMI1994;16, 3:320–6.
21.
Sakata. NeuroImage2007;35:1–8.
22.
AsselinM-CMontgomeryAJGrasbyPMHumeSP. Quantification of PET studies with the very high-affinity dopamine D2/D3 receptor ligand [11C]FLB 457: Re-evaluation of the validity of using a cerebellar reference region. Journal of Cerebral Blood Flow & Metabolism2007;27:378–92.
23.
MukherjeeJChristianBTDuniganKAShiBNarayananTKSatterMMantilJ. Brain imaging of 18F-fallypride in normal volunteers: Blood analysis, distribution, test-retest studies, and preliminary assessment of sensitivity to aging effects on dopamine D2/D3 receptors. Synapse2002;46:170–88.
24.
BishuS. JCBFM2008;28:1502–13.
25.
CunninghamVJJonesT. JCBFM1993;13:15–23.
26.
Turkheimer. JCBFM1994;14:406–22.
27.
Morris. Molecular Imaging2005;4:473–89.
28.
Normandin and Morris. Neuroimage2008;39:1162–79.
29.
Morris. Molecular Imaging and Biology2008;10:67–73.
30.
Alpert. Neuroimage2003;19:1049–60.
31.
Wu and Carson. Journal of Cerebral Blood Flow and Metabolism2002;22:1440–52.
32.
Ichise. Journal of Cerebral Blood Flow and Metabolism2003;23:1096–112.
