| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Clinical Investigations |
1 Brain Physiology and Metabolism Section, National Institute on Aging, National Institutes of Health, Bethesda, Maryland
2 Postgraduate Specialty School in Nuclear Medicine, University of Pisa Medical School, Pisa, Italy
3 Human Motor Control Section, National Institute on Neurological Diseases and Stroke, National Institutes of Health, Bethesda, Maryland
4 Clinical Epilepsy Section, National Institute on Neurological Diseases and Stroke, National Institutes of Health, Bethesda, Maryland
5 PET Department, Clinical Center, National Institutes of Health, Bethesda, Maryland
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
Key Words: arachidonic acid • blood volume • aging • partial-volume correction • PET
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
We have recently developed a method to quantify the incorporation of 11C-AA into human brain with PET (7). The kinetics of 11C-AA in the brain were described by an irreversible model with 2 parameters, the plasma-to- brain incorporation rate (K*) for 11C-AA and blood volume (Vb). Results in a group of young healthy subjects indicated that, despite a modest brain uptake of 11C-AA, this tracer could be used to quantify K* for AA (7), an index of neural signaling related to PLA2 (8).
Studies have indicated an age-related decline of the activity of the PLA2 AA signaling system. Reduced release of AA in response to muscarinic and serotonergic 5-HT1A and 5-HT2 receptors was reported in cortical areas of senescent rats (3). Thus, human studies with 11C-AA in healthy aging are of interest. In addition, we have shown (7) that the 11C-AA PET model provides physiologically reasonable estimates of Vb. PET studies have reported age-related reductions in Vb (9,10). However, due to age-related changes in brain morphology, the partial-volume effect has been shown to affect PET studies of cerebral blood flow (CBF) in aging (11).
Thus, the aim of this study was to evaluate the effect of aging on 11C-AA brain incorporation, CBV, and CBF in healthy human subjects. These comparisons were made with and without partial-volume correction (PVC).
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
MRI Procedure
Magnetic resonance (MR) images of the head were acquired with a 1.5-T Horizon (General Electric). T1-weighted volumetric spoiled gradient MR images (repetition time = 14 ms, echo time = 5.4 ms, flip angle = 20°) were acquired in a sagittal orientation (0.94 x 0.94 x 1.5 mm voxel size, 256 by 256 by 124 slices) and resliced to the transverse plane for all analyses.
Radiochemistry
11C-AA was synthesized as previously described (7). High-performance liquid chromatography analysis of the final product showed radiochemical purity of 98.6% ± 1.8% (n = 15).
PET Procedure
Each subject was scanned with a General Electric Advance tomograph. Scans were obtained parallel to the orbitomeatal line, while the subject’s head was held in place by a thermoplastic face mask. Scanning was conducted in a quiet, dim room, with the subject’s eyes open and ears unoccluded. A transmission scan was acquired to correct for attenuation. Then, CBF was measured by injecting 370 MBq (10 mCi) 15O-water as an intravenous bolus. A 60-s scan was acquired in 3-dimensional (3D) mode and quantitative CBF images were produced using the measured arterial input function (12). Approximately 15 min later, 902 ± 157 MBq (24.4 ± 4.3 mCi) of 11C-AA were infused intravenously for 3 min (Harvard Infusion Pump). Serial dynamic 3D scans (30 s to 5 min) were acquired over a 1-h period from the start of infusion.
Motion Correction
To correct for head motion during the 11C-AA scan, a 6-parameter rigid transformation was performed using the intramodality version of automated image registration (AIR) (13). Correction for head motion was performed starting from the first 3 min after injection with each image volume registered to the next image in the sequence. Images were thresholded to include pixels with values > 20% of the peak value in each frame. This limited the computation to extracerebral muscles, which have higher activity than the brain at all times. Images were resliced using a transformation matrix computed as the product of the matrices of each frame-to-frame registration. Motion of the ocular muscles before and after realignment was assessed by the sequential display of frames from a single slice at the level that best depicted muscle activity.
Modeling
Arterial blood samples were collected to measure whole blood, plasma, 11C-CO2, and 11C-AA concentrations (7). To separate 11C-AA from the remaining plasma activity, a previously validated (7) rapid extraction approach was used. On a pixel-by-pixel basis, the reconstructed images were analyzed with the following linear equation (7) to produce parametric images of K* and Vb:
 | (Eq. 1) |
In Equation 1, Ci(t), Cb(t), and Cp(t) are the pixel, whole blood, and plasma 11C-AA time–activity curves; Cco2(t) is the predicted brain tissue concentration of 11C-CO2, and 
t is the delay between the brain and blood curves (7). Calculations were applied to the original radioactivity images and to the PVC images.
Under certain assumptions (Discussion), K* can be used to calculate the net flux of AA from plasma into brain (Jin, µmol/d/g) by multiplying by the unlabeled unesterified AA plasma concentration. Jin may be used to assess the rate of metabolic consumption of AA by brain (5), which may be of interest due to the proposed relationship between dietary intake of AA and cognitive function.
To quantify unesterified AA concentration in plasma, total lipids were extracted from 2 plasma samples (100 µL, 0 and 60 min) using a partition system of chloroform, methanol, and water (14). Unesterified heptadecanoic acid (17:0, 50 nmol/100 µL plasma) was added as an internal standard. Unesterified fatty acids, including AA, in total lipid extracts were isolated by thin-layer chromatography (15), converted to fatty acid methyl esters (16), and separated on a gas chromatograph (model 6890N; Agilent Technologies). The unesterified AA concentration (nmol/mL) was calculated by proportional comparison of chromatographic peak area for AA with that of the standard. The 2 measurements were averaged. Data were not available for 1 young subject.
Registration to MR Images
For each subject, K* images derived from the original PET volumes were registered to the CBF volume to correct for motion between the 11C-AA and 15O-water scans using a 6-parameter transformation and the mutual information cost function (17). The CBF and MR volumes were then coregistered using the same algorithm. K* and Vb images were transformed to MR space using the product of the 2 transformation matrices.
Partial-Volume Correction
Two MR-based PVC approaches were used: 1 with 3 segments (3S: gray matter, white matter, and cerebrospinal fluid [CSF]) (18), and 1 using 2 segments (2S: brain tissue and CSF) (19). The 3S-PVC method is based on MR segmentation, which creates binary mask images for gray matter (mGM), white matter (mWM), and CSF (mCSF). A mask has a value of 1 in the pixels of that segment and a value of 0 in all other pixels. Gray matter pixels were corrected for spill-out of activity and for spill-in of activity from white matter. In the 3S-PVC approach, activity in CSF is assumed to be zero and activity in white matter is assumed to be uniform. The corrected values were calculated as follows:
 | (Eq. 2) |
where C3S is the corrected concentration in a gray matter pixel after PVC, C is the original uncorrected concentration, CWM is the estimated white matter value, and sGM and sWM are the pixel values from the smoothed masks for gray and white matter, respectively. The smoothed masks are created by convolving the binary mask with a 6-mm full width at half maximum 3D gaussian kernel.
To obtain an accurate estimate of CWM, pixel values that represent 100% white matter (sWM = 1) should be used. To estimate this value automatically, pixels with sWM values > 0.99 were identified. These are pixels with a radioactivity value that is unaffected by gray matter or CSF activity (i.e., "pure" white matter); this normally applies to white matter pixels that are far from the other segments—for example, pixels in the centrum semiovale. For each frame, the PET activity values of pixels with sWM values > 0.99 were then fitted as a linear function of sWM, and the fitted value at sWM = 1 was used as CWM. This approach required no operator interaction.
In the 2S method (19), no distinction between gray and white matter is made. Brain tissue is corrected only for spillover into CSF and nonbrain. In our approach, a smoothed brain mask, sB was created from the sum of sGM and sWM. The 2S-corrected (C2S) values are:
 | (Eq. 3) |
MR Segmentation
Extracerebral tissue was eliminated with automatic and manual methods. Segmentation of the edited MR volumes was performed with an adaptive fuzzy C-means algorithm (20), which computes a membership probability for each voxel in gray matter, white matter, and CSF. Each voxel was then assigned to the segment with the highest probability. Visual comparison of the segmented and original images revealed a high-quality segmentation in cortex, but poor results in the basal ganglia and thalamus, where a large fraction of gray matter pixels was misidentified as white matter. Other algorithms—namely, FAST (Oxford University, U.K.) (21) and SPM99 (22)—provided comparable or larger misclassifications. Therefore, gray matter in caudate, putamen, and thalamus was identified manually and the mask images were adjusted.
For 3S-PVC, the smoothed masks (sGM and sWM) were then created from the binary masks. Depending on the local geometry, some pixels defined as gray matter in the binary mask had a very small sGM value. This occurs, for example, at the tips of thin gyri. Correction of such pixels with Equation 2, would produce large changes (and large errors) in those pixel values due to the small sGM in the denominator (23). Therefore, under the assumption that PVC values in such pixels are unreliable, gray matter pixels with sGM values < 0.2 were reassigned to the segment (white matter or CSF) with the larger smoothed mask value at that pixel.
Region-of-Interest (ROI) Measurements
ROI measurements were limited to voxels that were defined as gray matter. First, for each subject, large 3D ROIs were manually drawn on MR images. Due to the much finer sampling of MR data compared with PET data, ROIs were placed on every fourth slice for cortical ROIs and every other slice for subcortical structures. The ROIs were then transferred to the binary gray matter mask image and all pixels not assigned to gray matter were eliminated from the ROIs. In this way, only gray matter pixels contributed to the ROI values. The ROIs were applied to the coregistered PET images and values for the uncorrected, 2S, and 3S data were obtained from CBF, K*, and Vb images. Global gray matter values were calculated as the average of the ROI values.
Statistical Analysis
Analysis was performed using unpaired and paired t tests, as appropriate. Right and left measurements of paired ROIs were averaged. Statistical significance was set at (uncorrected) P < 0.05. Absolute and normalized (to global gray matter) values were computed.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
Before image registration, head motion was detected in 3 of 8 young and 5 of 7 old subjects by visual inspection. After realignment, head motion was still detected in 1 of 8 young and 2 of 7 old subjects; in the latter 2 subjects, residual movement was reduced. Since residual motion was small and the ROIs were large, the effect on regional PET values was most likely quite small.
Figure 1 depicts the method used to estimate white matter concentration (CWM) for 3S-PVC. For each frame, the activity values of voxels with smoothed white matter mask sWM between 0.99 to 1.0 were fitted to a straight line and the extrapolated value for sWM = 1 was taken as CWM. The volume of white matter voxels with sWM from 0.99 to 1.0 was 14 ± 7 mL. For K* and Vb, the white matter value was determined by applying the model equation (Eq. 1) to the estimated CWM values. For CBF, the extrapolation method was applied directly to the parametric image. There were no significant group differences in white matter values for any tracer. In the whole group, white matter values equaled 2.2 ± 0.5 µL/min/mL (K*), 0.024 ± 0.004 mL/mL (Vb), and 17 ± 2 mL/min/100 g (CBF).
|
|
|
The net utilization (Jin) of AA was calculated using K* and the plasma concentration of AA. The plasma values were similar in young and old subjects and equaled 3.8 ± 1.7 nmol/mL, consistent with previous results (24). There were no significant differences between young and old subjects in Jin for any ROI, with or without PVC. Global gray matter Jin values for the whole group were 0.025 ± 0.013, 0.030 ± 0.016, and 0.039 ± 0.024 µmol/d/g for uncorrected, 2S-PVC, and 3S-PVC, respectively. Note the high intersubject variation in Jin, presumably caused by the high physiologic variance of the plasma unesterified AA concentration (24).
Absolute and normalized Vb values estimated from the 11C-AA studies before and after PVC are summarized in Table 2. No significant group difference was detected in absolute values. After normalization, a lower Vb value (P < 0.05) was found in the frontal lobe of old subjects; this difference was not significant after PVC. There were no other regional group differences before or after PVC.
|
|
|
For each subject, K* values for global gray matter were plotted against corresponding CBF values (Fig. 3). In agreement with previous results in young subjects (7), there was no significant relationship between K* and CBF for uncorrected data (K* = 2.93 + 0.039 CBF; P = 0.35). Application of PVC did not introduce significant K*–CBF correlations. There were also no significant correlations when young and old groups were analyzed separately.
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
Motion Correction
Sequential AIR registration of dynamic frames was adopted to correct for head movement during the 11C-AA acquisition. This algorithm assumes that registered images have a relatively uniform ratio between corresponding pixel values (13). Initial attempts to motion correct 11C-AA frames using a threshold that included brain tissue failed, possibly due to spillover of the high nonbrain activity into the brain. Therefore, registration based on noncerebral pixels was used. Motion of the ocular muscles, as detected from the sequential display of activity frames, was considered a sufficient, although not ideal, index of head movement. A threshold equal to 20% of peak frame activity provided the best performance, removing or reducing head motion in 14 of 15 subjects.
PVC Methods
The 3S-PVC method (18) corrects for spill-out of activity from gray matter as well as spill-in of white matter activity into gray matter pixels. Other ROI-based PVC methods also correct for partial-volume averaging between adjacent gray matter regions (25–27). Such methods are the preferred choice when gray matter activity is heterogeneous—for example, in receptor studies (28). Alternatively, the creation of PVC images permits the use of various ROI sampling strategies independent of the PVC process.
Estimation of White Matter Activity
The 3S-PVC method requires estimation of a global white matter value (CWM). We developed a new method that uses pixels with smoothed white matter mask values (sWM) close to 1. Previously, white matter was sampled from the centrum semiovale (18). Our approach has the advantage of being automatic and properly accounting for white matter size. In the young group, white matter K* (2.14 ± 0.40 µL/min/mL) was smaller than that previously calculated in the same population with an ROI approach (2.57 ± 0.54 µL/min/mL) (7). Thus, the ROI approach may overestimate the true white matter activity.
To assess the sensitivity to the estimated white matter value, in 1 young subject, PVC values were calculated with CWM set to 0 and compared with the original 3S values. In this case, K* and CBF were overestimated by 17% ± 3% and 11% ± 3%, respectively. Thus, a 10% error in the CWM estimate would introduce only a 1%–2% error; this result is consistent with previous simulations (18,23).
Extracranial Activity and PVC
Ideal 11C-AA PVC would require a fourth segment to account for spillover of muscle activity into cerebral pixels. Although most brain pixels are far from extracranial muscles, a worst-case scenario was evaluated by defining an orbitofrontal region and a nearby ocular muscle segment in a typical subject on a late 11C-AA frame. K* obtained without correction for muscle spillover was 
16% higher than the corrected value. This suggests that correction for extra-brain activity may be necessary for the orbitofrontal cortex. Since the exact physiologic mechanisms underlying 11C-AA-derived extracranial radioactivity are not known, additional evaluation is necessary to define the magnitude and intersubject variability of this muscle partial-volume effect.
Aging Effect
No significant group differences were detected before or after PVC in K* or Vb in cortical regions. In contrast, lower CBF values were found in the frontal lobe of the older population; this difference was no longer significant after PVC. Because of the small number of subjects, normalization to global gray matter was performed to increase statistical power. Significantly lower normalized K*, Vb, and CBF values were detected in the frontal cortex of older subjects; these differences also lost statistical significance after PVC. Reductions in frontal cortex function in normal aging have been observed in previous PET studies, both before (9–11,29) and after (11,30) PVC. In our 3S-PVC data, a few significant group differences in normalized cortical K* and CBF were found. We attribute the increases in normalized occipital 3S-PVC values to normalization artifact produced primarily by the (nonsignificant) reduction in global gray matter values.
Cerebellum
Lower CBF values were detected in the cerebellum in the elderly group with uncorrected and 2S-PVC data; differences were close to significance after 3S-PVC. These results differ from previous studies, in which cerebellar physiology was not affected by age (9–11,31). This difference may be due to our ROI strategy, using segmented MR even for the uncorrected data. If there are group differences in segmentation accuracy, artificial differences will be introduced into uncorrected and, potentially, PVC-corrected data.
Influx of AA
It has been postulated that the net brain utilization of AA, Jin (5), can be determined by multiplying K* by the unlabeled AA plasma concentration. However, there are limitations to this postulate. The 11C-AA model only estimates the irreversible uptake rate into membrane lipids. It does not directly measure the PLA2-induced release of AA from phospholipids, nor does it include the production of pro-inflammatory metabolites from released AA. However, since AA cannot be synthesized in the mammalian brain, Jin may be an index of AA metabolism (5). Results in a rat model of neuroinflammation are consistent with this hypothesis (32).
|
CONCLUSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
For correspondence or reprints contact: Richard E. Carson, PhD, PET Department, National Institutes of Health, Building 10, Room 1C-401, 10 Center Dr. MSC 1180, Bethesda, MD 20892-1180.
E-mail: richard-e-carson{at}nih.gov
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
Related articles in JNM:
This article has been cited by other articles:
|
|
 |
|
  G. Esposito, G. Giovacchini, J.-S. Liow, A. K. Bhattacharjee, D. Greenstein, M. Schapiro, M. Hallett, P. Herscovitch, W. C. Eckelman, R. E. Carson, et al. Imaging Neuroinflammation in Alzheimer's Disease with Radiolabeled Arachidonic Acid and PET J. Nucl. Med., September 1, 2008; 49(9): 1414 - 1421. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. Giovacchini, M. T. Toczek, R. Bonwetsch, A. Bagic, L. Lang, C. Fraser, P. Reeves-Tyer, P. Herscovitch, W. C. Eckelman, R. E. Carson, et al. 5-HT1A Receptors Are Reduced in Temporal Lobe Epilepsy After Partial-Volume Correction J. Nucl. Med., July 1, 2005; 46(7): 1128 - 1135. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| JOURNAL OF NUCLEAR MEDICINE TECHNOLOGY | THE JOURNAL OF NUCLEAR MEDICINE |
), after 2S-PVC (
), and after 3S-PVC (
). Each point represents global gray matter value for 1 subject (n = 15). No significant relation between K* and CBF was detected. Regression equations were as follows: Uncorrected, K* = 2.93 + 0.039 CBF (r = 0.26; P > 0.05); 2S, K* = 1.65 + 0.075 CBF (r = 0.45; P > 0.05); 3S, K* = 3.98 + 0.047 CBF (r = 0.30; P > 0.05).