Early Phase PIB-PET as a Surrogate for Global and Regional Cerebral Blood Flow Measures

Abstract

Background and Purpose:

To explore the potential for simplified measures of [¹¹C]PIB uptake to serve as a surrogate for cerebral blood flow (CBF) measures, thereby, providing both pathological and functional information in the same scan.

Methods:

Participants (N=24,16M,8F,57–87 years) underwent quantitative [¹⁵O]water imaging and dynamic [¹¹C]PIB imaging. Time-activity curves (TACs) were created for each participant’s regional [¹¹C]PIB data scaled in standardized uptake values (SUVs). The frame in which maximal uptake occurred was defined for each subject (i.e., “peak”). The concentration (SUV) for each region at the individual’s peak, during the 3.5 to 4 minute time interval and for the initial 6 minute sum were determined. R1 (i.e., relative delivery using cerebellum as reference tissue) from the simplified reference tissue model 2 (SRTM2) was determined for each region. PIB SUVs were compared to the absolute CBF global and regional values (in mL/min/100mL) and the R1 values were compared to the cerebellar-normalized rCBF.

Results:

Significant linear relationships were found for all SUV measures with measures of absolute global and regional CBF that were comparable to the relationship between normalized CBF and R1. The individual SUV_peak exhibited the strongest relationship both regionally and globally. All individuals and all regions had highly significant regression slopes. Age, gender or amyloid burden did not influence the relationship.

Conclusion:

Early PIB uptake has the potential to effectively serve as a surrogate for global and regional cerebral blood flow measures. The simple and readily obtainable individual’s SUV_peak value was the strongest predictor regionally and globally of CBF.

Introduction

Cardiovascular risk factors¹ and cerebrovascular dysfunction² have been associated with an enhanced risk for the development of Alzheimer’s Disease (AD). The existence of common risk factors and the positive effects of the treatment of vascular risk factors on cognitive decline in patients with AD without cerebrovascular disease³ reinforce the view that in the least, the pathologies are linked. Storkebaum, et al.⁴ detail “a vicious circle of reciprocal interactions between β-amyloid accumulation and vascular defects that ultimately can lead to hypoperfusion and metabolic demise of neuronal and vascular cells.” The data-driven model of late-onset Alzheimer’s disease (LOAD), in contrast to the hypothetical progression model, cites the critical role of vascular dysfunction as the “earliest/strongest altered event, followed by Aβ deposition…”⁵ Because of these critical links, assessments of developing Alzheimer’s pathology should ideally address both cerebrovascular function as well as amyloid pathology. The mechanism of action for the amyloid imaging agents involves a wash-in/wash-out of the radiotracer with retention in the amyloid neuritic plaques. The potential of the initial phase of the amyloid study to reflect cerebral blood flow (CBF) has been recognized. Reports have compared CBF to estimates derived from the summed early phase [¹¹C]PIB images (6 minutes,^6,7 1 to 4 minutes⁸), and WARM (washout allometric reference method)^9,10 and the normalized CBF (to cerebellum) to the R1 parameter from the standard reference tissue model.¹¹ These techniques are generally computationally demanding (e.g., R1 or WARM) or produce relative CBF measures only.¹² It was the purpose of this research to explore the potential of readily-derived, early PIB-based parameters and demographic information to model absolute global and regional CBF as determined by the gold-standard, quantitative [¹⁵O]water PET imaging. This research was based on the hypothesis that early phase [¹¹C]PIB uptake, globally and regionally, is significantly related to global and regional measures of CBF, respectively. If a clinically-accessible metric, significantly related to CBF can be established for [¹¹C]PIB, then the stage is set for extension of this methodology to alternative amyloid agents with widespread clinical use potential, i.e., F-18 labeled radiotracers. If that goal is realized, a single agent in a single scan session could be used to define both the functional (CBF) and pathological (amyloid burden) status of a patient at-risk for vascular and/or Alzheimer’s dementia.

Methods

Participants:

Twenty-four participants underwent the imaging procedures. The sample consisted of 8 females and 16 males, ranging in age from 57 to 87 years (mean = 70.0 ± 10.0 yr, males = 68.0 ± 10.5, females = 73.9 ± 8.2) and represented the cognitive spectrum from healthy controls to mild cognitive impairment (MCI) (MMSE 23 – 30). Participants were excluded if they were unable to undergo MR imaging, had a history of or current diagnosis of major psychiatric disease (other than dementia), a history of abuse of psychoactive substances or medications, a history of stroke, a history of head injury with loss of consciousness for greater than 30 minutes, or a history of other neurological disorder or systemic illness that could potentially affect cognition or brain function (outside of a diagnosis MCI) or could affect their safety or comfort while undergoing the imaging studies. The study was approved by the University of Iowa Institutional Review Board and Medical Radiation Protection Committee and all participants signed a written informed consent form. No potential conflicts of interest relevant to this article exist.

Radiopharmaceuticals:

N-Methyl-[¹¹C]-2-(4’-methylaminophenyl-6-hydroxybenzothiazole ([¹¹C]PIB or PIB) was prepared according to the methods and met all specifications detailed in the CMC section of IND 109,386 (Michael Graham, M.D., Ph.D., physician-sponsor). [¹⁵O]water was prepared according to the methods and met all specifications detailed in the CMC section of IND #115,599 (Michael Graham, M.D., Ph.D., physician-sponsor).

Imaging Procedures:

Imaging consisted of structural MRI and two different positron emission tomography (PET) imaging sessions: 1) amyloid imaging with [¹¹C]PIB and 2) quantitative cerebral blood flow imaging with [¹⁵O]water (with arterial blood sampling) under 3 conditions (counting task, list remembering task, rest task (eyes open, ears unplugged). All PET imaging was performed on a Siemens ECAT EXACT HR+ with transmission imaging ([⁶⁸Ge]) for attenuation correction performed prior to the injection of the radiotracers. MR imaging was performed on a 3.0T Siemens TIM Trio MRI Scanner or on a GE 750W 3T scanner. All image analyses were performed using the PMOD suite of tools (PVIEW, PFUSION, PNEURO, PXMOD, PKIN, PALZ, v. 3.7 and 3.8, PMOD Technologies, Ltd, Zurich, Switzerland).

MRI:

Structural MRI was acquired as volumetric sagittal T1 MP-RAGE (TI=900ms, TE=3.2ms, TR=8.5ms, flip angle=8°, FOV=256×256×192mm, matrix=256×256×192, bandwidth= 250Hz/pixel, acceleration=2) on a 3.0T Siemens TIM Trio MRI Scanner or on a GE 750W 3T scanner.

[¹⁵O]Water PET:

Participants underwent quantitative [¹⁵O]water imaging (1665 ± 10% MBq/injection) (45 ± 10% mCi/injection) for the determination of cerebral blood flow (CBF in mL/min/100mL of tissue). The quantitative [¹⁵O]water approach required arterial blood sampling. Technical methodology was implemented as described for work at this institution.^13–15 Dynamic imaging and arterial blood sampling commenced at tracer injection and continued for 100 seconds (5 seconds/frame × 20 frames). Dynamic data were iteratively reconstructed (Gaussian 5.0, 2 iterations/8 subsets, attenuation, scatter, decay and deadtime correction = ON). Parametric images were calculated from the summed activity images (40 seconds post-bolus transit) and the arterial blood curve using the autoradiographic model in the pixel-wise modeling module (PXMOD). CBF measurements were performed under three conditions - a rest condition, during performance of a simple cognitive task (counting) and while remembering a pre-learned word list. The “rest” condition mimicked that employed for studies designed to delineate the default mode network¹⁶ whereas, counting represents an over-learned simple cognitive task with few cognitive demands and list remembering constitutes task-based changes in blood flow. Since “rest” is a relatively uncontrolled condition that may be imperfectly reproduced between the two imaging sessions with respect to cognitive activity and differs in imaging duration and sequence of injections, comparisons were made to all three measures of CBF individually as well as the average of the measurements, i.e., CBF_mean.

[¹⁵O]Water PET Image Processing:

[¹⁵O]Water parametric blood flow images were co-registered to the T1-weighted MRI using the Fusion tool (rigid matching configuration using normalized mutual information routine) of PMOD Biomedical Image Quantification, version 3.9 (PMOD Technologies, Ltd., Zurich, Switzerland). Global CBF (gCBF) was determined for each injection of [¹⁵O]water by volume-weighted averaging of the flow values for all regions defining intracerebral pixels as determined from brain parcellation analysis of the co-registered MRI. For each subject, there were four measures of global CBF – one for each activation condition (count, memory, rest) and average of these three measures (i.e., CBF_{global mean}). Regional CBF (rCBF) was determined for each lobar, cortical and subcortical region defined on the co-registered T1 MRI using the maximum probability atlas¹⁷ (N = 78 regions/subject) of Neuro Tool of PMOD for each activation condition as well as the average of the three conditions (i.e., CBF_{regional mean}).

[¹¹C]PIB PET:

[¹¹C]PIB (555 ± 10% MBq) (15 ± 10% mCi) was administered as an IV bolus at rest, imaged dynamically for 90 minutes (4 × 15 sec, 8 × 30 sec, 9 × 60 sec, 2 × 180 sec, 14 × 300 sec), and the images were iteratively reconstructed (Gaussian 3.0, 6 iterations/16 subsets, zoom = 2.6, attenuation, scatter, decay and deadtime correction = ON). A summed image was created with the data from 50 to 70 minutes post-administration. Both the dynamic frames and the summed image were converted to standardized uptake value images.

[¹¹C]PIB PET Image Processing:

Dynamic imaging commenced at injection and continued over a 90 minute period. The early phase of the dynamic sequence (i.e., first 6 minutes, see below) was used to characterize cerebral blood flow and the latter phase for amyloid burden. The dynamic images were co-registered to the anatomical MRI for tissue and region definition. Time-activity curves (TACs) were created for each participant’s regional [¹¹C]PIB data (PNEURO PMOD) scaled in standardized uptake values (SUVs) (N=24 × 78 regions/participant=1872 regions). The time frame in which the maximal uptake occurred for the mean of all regions was defined for each participant (i.e., “peak”). See Figure 1. The concentrations (SUV units) were determined for each region and for a volume-weighted global value for each of the following time intervals:

• SUV_peak: Individual’s peak (30 second interval)

• SUV_3.5–4min: 3.5 to 4 minute time interval (latest individual peak = Frame 10)

• SUV_6min: Initial 6 minute sum.

The rationale for each of these SUV-based measures is that the peak represents the end of the wash-in phase for that particular participant, whereas the 3.5 to 4 minute time interval represents the latest time for any of the participants peak values and therefore, would constitute a time window of general applicability. The sum over the first 6 minutes represents a time window employed by Forsberg, et al.⁶ and Gietl⁷ when reporting on the utility of PIB as a potential dual pathology tracer. The global and regional PIB SUVs were compared to the absolute CBF global and regional values (rCBF) for each condition and to the mean global and regional CBF values, respectively. The regional R1 values (i.e., relative delivery derived from DVR analyses described below) were compared to the cerebellar-normalized rCBF (i.e., specific region’s absolute CBF divided by the absolute CBF in the bilateral cerebellum) because R1 values are inherently relative to the cerebellum.

Figure 1.

Time-activity curves for [11C]PIB uptake (scaled in SUV units) versus time (seconds) for subjects A through E. Values next to arrows are global cerebral blood flows (CBF) determined with [15O]water. Red markers are the individual peaks. Vertical line marks the midpoint of the 3.5 – 4 minute interval. Cortical Retention Ratio (CRR) for PIB uptake in SUVR determined from 50 to 70 minutes averaged over regions that accumulate amyloid are listed in the legend to illustrate the lack of relationship between the early peak values and the latter accumulation of PIB as demonstrated by CRR.

The entire 90 minute sequence and the time interval between 50 and 70 minutes were used for quantifying the amyloid burden using methods previously described^18–20. The regional distribution volume (DV) normalized by the nonspecific retention in the reference region (cerebellum), referred to as DVR, was determined for each region from the simplified reference tissue model 2 (SRTM2) using the cerebellum as the reference tissue as implemented by the kinetic modeling tool (PKIN PMOD) was the primary measure used to assess amyloid burden. The standardized uptake volume ratio (SUVR normalized to cerebellar gray matter) for the interval from 50 to 70 minutes was also calculated as an alternate measure of the amyloid burden. The cortical retention ratios (CRR = volume-weighted mean of AD-related tissues) were calculated for each individual based on the regional DVR and SUVR (standardized uptake value ratio) values by calculating the mean gray matter values averaged across the lateral and medial frontal, anterior and posterior cingulate, lateral parietal and lateral temporal regions as described by Landau, et al.²¹

Statistical Analyses:

All analyses were designed to compare PIB SUV measures with global and regional CBF measures. Analyses consisted of simple linear regression models with CBF as the outcome measure and PIB SUV as the predictor variable. The intercept was set equal to zero to force the model through the origin. Linear mixed models were used to account for the within subject correlation via a random intercept to include multiple regional values in the same analysis. These models were performed with and without inclusion of potentially relevant demographic information (e.g., age, sex). The models were compared based on the R² value and the p-value associated with the slope coefficient.

Results

Comparison between early [¹¹C]PIB and [¹⁵O]Water CBF:

The peak SUV occurred at 105 to 225 seconds post-initiation of the bolus injection. Regression models were used to evaluate the relationship between early PIB SUV measures (specifically, the individual participant’s peak (SUV_peak), the fixed early timeframe (SUV_3.5–4min) and the initial 6 minute summed image (SUV_6min)) versus global and regional CBF measures. All of the early PIB SUV measures exhibited positive slopes, significantly different from zero, when regressed to the global and regional CBF measures determined from each of the tasks (counting, memory and rest) as well as the mean of all CBF measures. With the exception of the 6 minute summed image SUV (SUV_6min) for which the highest R² value was observed for the comparison with CBF_{global count}, the highest R² values were observed for the comparisons with the mean CBF values for both comparisons to the participant’s global measures as well as the individual anatomical regional CBF. Therefore, Table 1 presents results for the analyses with the regional and global mean CBF comparisons.

Table 1.

Results of Modeling of Cerebral Blood Flow (CBF) by Condition versus PIB-derived Measures

Simple Linear Models
Global: (Regression with intercept = 0)*	Full model R2
CBFglobal mean = 10.71 SUVpeak	R2 = 0.45 (intercept p = 0.16)
CBFglobal mean = 11.01 SUV3.5–4min	R2 = 0.40 (intercept p = 0.21)
CBFglobal mean = 12.42 SUV6min	R2 = 0.24 (intercept p = 0.03)
Reproducibility: CBFglobal rest = 0.98 CBFglobal count	R2 = 0.67 (intercept p = 0.04)
Reproducibility: CBFglobal rest=0.995 CBFglobal memory	R2 = 0.73 (intercept p = 0.006)
Reproducibility: CBFglobal count=1.01 CBFglobal memory	R2 = 0.83 (intercept p = 0.05)
Regional: (Regression with intercept = 0)*	Full model R2
CBFregional mean = 10.53 SUVpeak	R2 = 0.52
CBFregional mean = 10.80 SUV3.5–4	R2 = 0.50
CBFregional mean = 12.17 SUV6min	R2 = 0.41
CBFregional/CBFcerebellar = 0.77 R1 + 0.15	R2 = 0.64
Reproducibility: CBFregional rest=0.97 CBFregional count	R2 = 0.74
Reproducibility: CBFregional rest=0.995 CBFregional memory	R2 = 0.76
Reproducibility: CBFregional count = 1.02 CBFregional memory	R2 = 0.85
Linear Mixed Regression Models
CBFregional mean = 8.57 SUVpeak + 6.68	R2 = 0.60 (p < 0.0001)
CBFregional rest = 8.52 SUVpeak + 6.79	R2 = 0.58 (p < 0.0001)

^*All coefficients significant at the p<0.0001 level. Condition = rest, count, memory condition or mean of conditions. PIB-derived measures = Standardized Uptake Values (SUV) by Timeframe (Peak, 3.5 – 4 minutes Frame, 6 minute sum

Regression analyses were performed for each subject (N = 24) and region separately (N=78) to relate CBF_mean to SUV_peak. Figure 2 illustrates an example image along with plots of the individual’s regional CBF values vs SUV_peak values and the individual’s reproducibility between two different tasks (rest vs counting). Although noisier than the [¹⁵O]water images, the early PIB images obviously display a similar blood flow pattern across the entire brain without anatomical discrepencies.

Figure 2.

Example of [15O]water cerebral blood flow image (mL/min/100 mL) (Figure 2A, upper row) and [11C]PIB individual peak image (30 second time interval at 180 – 210 seconds post-injection in SUV units) (Figure 2A, lower row). The PIB peak image is noisier than the [15O]water CBF image but has similar information content regarding the distribution of cerebral blood flow as illustrated by the plot of regional [15O]water CBF versus PIB SUV peak (Figure 2B). For comparison purposes, the plot of [15O]water CBF from the rest versus the counting task is presented (Figure 2C).

Figure 3 plots regional CBF_mean vs PIB SUV_peak for all individuals along with the individual’s fitted regression lines with Figure 4 presenting the corresponding reproducibility of two separate CBF measures using [¹⁵O]water PET for consecutive but not identical cognitive tasks (rest versus counting). The enhanced variability illustrated in Figure 3 compared to Figure 4 represents the uncertainty associated with using the PIB-based measure for CBF compared to the gold-standard [¹⁵O]water PET.

Figure 3.

Plot of mean regional cerebral blood flow (mL/min/100 mL) from quantitative [15O]water imaging versus PIB peak SUV (as defined in the text). The bold line represents the fitted line for all data (assuming zero intercept) along with the individual fitted lines for all subjects. All individual slope values were within ±2 standard deviations of the slope calculated from all data.

Figure 4.

Plot of mean regional cerebral blood flow (CBF) (mL/min/100 mL) from quantitative [15O]water imaging determined during the rest condition versus CBF determined during the counting condition. The bold line represents the fitted line for all data (assuming zero intercept). This plot illustrates the reproducibility of CBF measures acquired under minimally different conditions.

Every individual participant (N = 24) exhibited statistically significant correlational relationships between CBF_mean and PIB SUV_peak at the p< 0.0001 level (r = 0.73 to 0.90) (see Table 2 and Figure 5) that would remain significant even with a Bonferroni correction (using a 0.05/24 = 0.002 level of significance). Every region (N = 78) exhibited statistically significant positive correlations (r = 0.44 to 0.82) between CBF_mean and PIB SUV_peak (see Table 3). Of these regions, 35/78 would remain significant even with a Bonferroni correction (using a 0.05/78 = 0.0006 of significance). Figure 6 illustrates the consistency across the all brain anatomical regions between the mean [¹⁵O]water and PIB SUV_peak estimates of CBF. All PIB-based rCBF estimates were within the ±1 SD of the rCBF measures derived from [¹⁵O]water. Age and gender were not significant parameters in describing the relationship between the early PIB measures and CBF (data not shown). As was the case for the first five participants illustrated in Figure 1, the cortical retention ratio (CRR) was not related to any of the early PIB measures (data not shown).

Table 2.

Correlations between Regional Mean Cerebral Blood Flow and PIB-based Measures by Subject

Subject	PIB SUVpeak	PIB SUV3.5–4min	PIB SUV6min	R1
A	0.852	0.844	0.876	0.881
B	0.796	0.742	0.815	0.821
C	0.761	0.713	0.784	0.788
D	0.860	0.830	0.862	0.873
E	0.726	0.774	0.776	0.787
F	0.810	0.781	0.832	0.831
G	0.805	0.801	0.816	0.822
H	0.784	0.810	0.829	0.833
I	0.773	0.773	0.794	0.601
J	0.811	0.815	0.829	0.834
K	0.904	0.889	0.912	0.912
L	0.799	0.773	0.779	0.779
M	0.853	0.839	0.853	0.852
N	0.780	0.776	0.811	0.813
O	0.764	0.740	0.786	0.770
P	0.815	0.796	0.834	0.839
Q	0.777	0.770	0.803	0.806
R	0.785	0.806	0.825	0.840
S	0.857	0.841	0.878	0.862
T	0.824	0.822	0.854	0.858
U	0.823	0.840	0.868	0.871
V	0.777	0.745	0.797	0.779
W	0.734	0.734	0.765	0.755
X	0.729	0.772	0.789	0.790
MEAN	0.800	0.783	0.824	0.817
SD	0.044	0.073	0.038	0.060
CV%	5.53%	9.34%	4.61%	7.40%

PIB peak = standardized uptake value (SUV) at the individual’s peak, PIB SUV_3.5–4min = SUV during the frame acquired from 3.5 – 4 minutes post-administration, PIB SUV_6min = SUV calculated from the sum of the first 6 minutes post-administration, R1 = relative delivery using cerebellum as reference tissue calculated from the simplified reference tissue model 2 (SRTM2), SD = standard deviation, CV% = coefficient of variation expressed as a percent of the mean.

Figure 5.

Individual participant plots of CBFmean (mL/min/100 mL) versus PIB SUVpeak (g/mL) for all regions (N = 78). Red lines are the fitted lines for the equation at the base of the plot. The correlations ranged from 0.73 to 0.90 for the relationships with R2 values ranging from 0.53 to 0.82. All individual relationships are significant at the p < 0.0001 level.

Table 3.

Correlations between Mean Cerebral Blood Flow (CBF_mean) and PIB-based Measures by Region-of-Interest and Cerebellar-Normalized CBF_mean versus R1

Region	PIB SUVpeak	PIB SUV3.5–4min	PIB SUV6min	R1		R1 vs NormCBF
Amygdala_l	0.652	0.626	0.453	0.345		0.469
Amygdala_r	0.531	0.628	0.440	0.360		0.537
Ant_TL_inf_lat_l	0.550	0.556	0.410	0.342		0.751
Ant_TL_inf_lat_r	0.507	0.506	0.384	0.105		0.445
Ant_TL_med_l	0.605	0.562	0.422	0.381		0.565
Ant_TL_med_r	0.525	0.487	0.271	0.024		0.105
Brainstem	0.599	0.630	0.476	0.504		0.188
CaudateNucl_l	0.766	0.749	0.736	0.617		0.739
CaudateNucl_r	0.808	0.793	0.780	0.661		0.706
Cerebellum_l	0.532	0.516	0.374	0.309		0.220
Cerebellum_r	0.526	0.523	0.357	0.309		−0.137
Corp_Callosum	0.824	0.709	0.788	0.785		0.711
FL_inf_fr_G_l	0.702	0.668	0.603	0.535		0.763
FL_inf_fr_G_r	0.580	0.526	0.421	0.386		0.705
FL_mid_fr_G_l	0.681	0.586	0.554	0.411		0.756
FL_mid_fr_G_r	0.597	0.540	0.457	0.297		0.685
FL_OFC_AOG_l	0.734	0.657	0.649	0.554		0.512
FL_OFC_AOG_r	0.655	0.661	0.569	0.570		0.682
FL_OFC_LOG_l	0.622	0.615	0.570	0.427		0.551
FL_OFC_LOG_r	0.491	0.495	0.399	0.262		0.548
FL_OFC_MOG_l	0.666	0.609	0.557	0.460		0.397
FL_OFC_MOG_r	0.588	0.535	0.436	0.317		0.576
FL_OFC_POG_l	0.703	0.622	0.576	0.592		0.606
FL_OFC_POG_r	0.657	0.667	0.486	0.564		0.625
FL_precen_G_l	0.631	0.616	0.541	0.464		0.750
FL_precen_G_r	0.578	0.563	0.489	0.334		0.651
FL_strai_G_l	0.609	0.684	0.524	0.459		0.559
FL_strai_G_r	0.661	0.637	0.530	0.516		0.704
FL_sup_fr_G_l	0.688	0.658	0.578	0.486		0.743
FL_sup_fr_G_r	0.672	0.642	0.551	0.430		0.689
G_cing_ant_l	0.759	0.741	0.638	0.774		0.807
G_cing_ant_r	0.785	0.762	0.645	0.809		0.880
G_cing_post_l	0.734	0.732	0.663	0.644		0.731
G_cing_post_r	0.804	0.763	0.647	0.692		0.813
G_fus_l	0.670	0.685	0.489	0.628		0.644
G_fus_r	0.666	0.635	0.470	0.583		0.662
G_paraH_amb_l	0.707	0.723	0.548	0.653		0.479
G_paraH_amb_r	0.706	0.604	0.503	0.523		0.584
G_sup_temp_ant_l	0.694	0.616	0.454	0.355		0.662
G_sup_temp_ant_r	0.441	0.420	0.338	0.082		0.227
G_sup_temp_post_l	0.620	0.581	0.514	0.279		0.792
G_sup_temp_post_r	0.529	0.425	0.397	0.120		0.577
G_tem_midin_l	0.557	0.563	0.397	0.263		0.706
G_tem_midin_r	0.482	0.455	0.330	0.228		0.512
Hippocampus_l	0.658	0.622	0.497	0.483		0.538
Hippocampus_r	0.607	0.598	0.508	0.399		0.581
Insula_l	0.710	0.706	0.575	0.602		0.720
Insula_r	0.651	0.609	0.484	0.515		0.751
NuclAccumb_l	0.632	0.684	0.539	0.436		0.646
NuclAccumb_r	0.620	0.551	0.542	0.556		0.625
OL_cuneus_l	0.584	0.602	0.526	0.328		0.647
OL_cuneus_r	0.632	0.589	0.626	0.398		0.454
OL_ling_G_l	0.718	0.646	0.611	0.635		0.254
OL_ling_G_r	0.675	0.644	0.606	0.566		0.167
OL_rest_lat_l	0.581	0.573	0.440	0.409		0.774
OL_rest_lat_r	0.454	0.428	0.354	0.263		0.606
Pallidum_l	0.703	0.701	0.569	0.648		0.732
Pallidum_r	0.710	0.701	0.534	0.647		0.737
PL_postce_G_l	0.607	0.571	0.496	0.387		0.685
PL_postce_G_r	0.543	0.507	0.428	0.259		0.622
PL_rest_l	0.575	0.553	0.444	0.280		0.850
PL_rest_r	0.520	0.482	0.391	0.204		0.712
PL_sup_pa_G_l	0.688	0.637	0.567	0.427		0.767
PL_sup_pa_G_r	0.656	0.634	0.534	0.302		0.704
Post_TL_l	0.665	0.599	0.491	0.319		0.757
Post_TL_r	0.586	0.528	0.416	0.313		0.613
Presubgen_antCing_l	0.713	0.598	0.612	0.660		0.788
Presubgen_antCing_r	0.652	0.662	0.506	0.515		0.657
Putamen_l	0.705	0.479	0.568	0.615		0.743
Putamen_r	0.645	0.649	0.497	0.476		0.726
S_nigra_l	0.592	0.245	0.591	0.583		0.360
S_nigra_r	0.553	0.654	0.529	0.557		0.471
Subcall_area_l	0.745	0.767	0.643	0.545		0.829
Subcall_area_r	0.799	0.732	0.755	0.591		0.735
Subgen_antCing_l	0.723	0.668	0.625	0.556		0.627
Subgen_antCing_r	0.711	0.738	0.615	0.543		0.748
Thalamus_l	0.711	0.752	0.616	0.587		0.764
Thalamus_r	0.772	0.721	0.677	0.634		0.830
MEAN	0.643	0.613	0.523	0.457		0.617
SD	0.086	0.097	0.105	0.167		0.187
CV%	13.4%	15.8%	20.1%	36.4%		30.4%

PIB peak = standardized uptake value (SUV) at the individual’s peak, PIB SUV_3.5–4min = SUV during the frame acquired from 3.5 – 4 minutes post-administration, PIB SUV_6min = SUV calculated from the sum of the first 6 minutes post-administration, R1 = relative delivery using cerebellum as reference tissue calculated from the simplified reference tissue model 2 (SRTM2), R1 vs NormCBF = R1 compared to the cerebral blood flow (CBF) normalized by cerebellar CBF, SD = standard deviation, CV% = coefficient of variation expressed as a percent of the mean.

Region definitions: l = left, r = right; Amygdala = amygdala; Ant_TL_inf_lat = anterior temporal lobe, lateral part; Ant_TL_med = anterior temporal lobe, medial part; Brainstem = brainstem; CaudateNucl = caudate nucleus; Cerebellum = cerebellum; Corp_Callosum = corpus callosum; FL_inf_fr_G = inferior frontal gyrus; FL_mid_fr_G = middle frontal gyrus; FL_OFC_AOG = anterior orbital gyrus; FL_OFC_LOG = lateral orbital gyrus; FL_OFC_MOG = medial orbital gyrus; FL_OFC_POG = posterior orbital gyrus; FL_precen_G = precentral gyrus; FL_strai_G = straight gyrus; FL_sup_fr_G = superior frontal gyrus; G_cing_ant = cingulate gyrus, anterior part; G_cing_post = cingulate gyrus, posterior part; G_fus = fusiform gyrus; G_paraH_amb =parahippocampal and ambient gyri; G_sup_temp_ant = superior temporal gyrus, anterior part; G_sup_temp_post = superior temporal gyrus, posterior part; G_tem_midin = middle and inferior temporal gyrus; Hippocampus = hippocampus; Insula = insula; NuclAccumb = nucleus accumbens; OL_cuneus = cuneus; OL_ling_G = lingual gyrus; OL_rest_lat = lateral remainder of occipital lobe; Pallidum = pallidum; PL_postce_G = postcentral gyrus; PL_rest = lateral remainder of occipital lobe; PL_sup_pa_G = superior parietal gyrus; Post_TL = posterior temporal lobe; Presubgen_antCing = pre-subgenual frontal cortex; Putamen = putamen; S_nigra = substantia nigra; Subcall_area = subcallosal area; Subgen_antCing = subgenual frontal cortex; Thalamus = thalamus.

Figure 6.

Mean regional cerebral blood (rCBF) mL/min/100 mL) (N = 24) measured by quantitative [15O]water (blue profile) ± 1 standard deviation and by PIB SUVpeak using linear mixed modeling (CBFmean = 8.57 × SUVpeak + 6.68) (red profile). Note: All PIB-based rCBF estimates are within ± 1 standard deviation and exhibit a similar profile across all brain regions

Comparison between [¹¹C]PIB R1 and [¹⁵O]Water CBF:

Statistically significant relationships, based on simple linear regressions and correlations, were observed between the CBF_mean values derived from [¹⁵O]water and R1 (relative tracer delivery derived from the SRTM2 analysis of the dynamic PIB data) for the composite data set and each individual participant, however, significant correlations were not observed for all anatomical regions, even when comparing to the cerebellar-normalized CBF values (see Table 3, farthest right-hand column). Furthermore, the correlations between CBF values and R1 were lower than those observed between CBF and PIB SUV_peak (e.g., R² = 0.39 versus R² = 0.52, respectively), but, as expected, was higher when compared to the cerebellar-normalized CBF (R² = 0.64).

Discussion

The importance of compromises in cerebral blood flow, not only in cerebrovascular disease, but in the early Alzheimer’s pathological cascade is being recognized.^{1,5, 22, 23} Quantitative [¹⁵O]water is the recognized gold-standard for the assessment of cerebral blood flow in humans since the ground-breaking work of Herscovitch²⁴ and Raichle²⁵ in the 1980s. However, few centers have the geographic configuration or the technical expertise to implement [¹⁵O]water-based quantitative CBF measurements. Alternatives, for the evaluation of relative CBF such as the SPECT agents, Tc^99mECD or Tc^99mHMPAO or BOLD MRI and absolute CBF such as ASL MRI, have been evaluated. The current work presents another promising technique, early phase PIB, with the potential to determine both relative CBF (i.e., distribution throughout the brain) and absolute CBF (i.e., in terms of mL/min/100mL tissue). Early-phase PIB, particularly the SUV_peak, has the advantages of being computationally simple, anatomically independent and informationally coupled with the determination of amyloid burden (i.e., two pieces of AD-relevant information in one scan). Predictions of global and regional CBF can be based on the following equations:

$${\text{CBF}}_{\text{global mean}}=10.71{\text{SUV}}_{\text{global peak}}$$ eq. 1

$${\text{CBF}}_{\text{regional mean}}=8.57{\text{SUV}}_{\text{regional peak}}+6.68$$ eq. 2

where “global” represents values determined from averaging over the entire brain and “regional” represents values derived from individual anatomical regions. Equation 1 is derived from a simple linear regression of global mean values, whereas, equation 2 is derived from a linear mixed regression model using all of the individual’s region-based values. The two models are equivalent at 33.4 mL/min/100 mL or an SUV = 3.12. The use of SUV_peak does require some image-analysis work to determine the time at which the individual’s peak value occurred. An even simpler and only marginally less reliable is the use of an SUV determined over a fixed time interval (in this case, 3.5 – 4 minutes). Both of these measures were superior to the use of the 6 minute average SUV. However, the small sample size that these conclusions are based on is a limitation of the study.

The data reported in this paper were not collected solely for addressing the relationship between CBF and the early PIB parameters. Information was available on CBF under three different activation conditions – a simple memory task (counting), a more difficult memory task (remembering a pre-learned word list) and rest. These conditions were acquired in that order – counting, memory and rest – therefore, the rest condition was the third [¹⁵O]water injection and was acquired when the individual had been on the scanner bed for approximately 45 minutes. The rest condition appears to be the appropriate comparison condition if what the participant was supposed to be “doing” at the time of imaging was the only consideration. However, the magnitude and distribution of CBF measures vary with factors beyond the task being performed. Although the participant was told to “rest” (eyes open, ears unplugged) for the PIB injection, it was the first injection of that particular session, a situation more analogous to the counting task. Secondly, “rest” for the [¹⁵O]water injections was 100 seconds in duration whereas “rest” for the PIB condition lasted for 6 minutes, a time frame during which the subject was likely to engage in cognitive activities. Since the brain is never in an “off” or even standard framework, “rest” is a variable condition from nearly asleep to full-blown cognitive engagement, depending on the participant and how comfortable and bored the individual may be. Therefore, because the conditions could not be matched exactly, comparisons were made between the early PIB parameters and each of the [¹⁵O]water conditions that were available as well as to the average (mean) of all three conditions. Figure 4 illustrates the reproducibility of two quantitative [¹⁵O]water CBF determinations using different activation conditions and represents the potential underlying uncertainty of CBF determinations made on the same day using the same technique but minimally different activation conditions. The comparison of Figure 3 to Figure 4 illustrates the additional uncertainty of CBF determinations using the simpler, early-phase PIB technique as well as differences incurred by making measurements on separate days.

Although promising, the major limitation of the early-phase PIB technique as a surrogate for CBF is comparable to the limitation of PIB as an amyloid imaging agent – specifically, the [¹¹C] label on the radiopharmaceutical. Because of the relatively short half-life of [¹¹C] (20 minutes), PIB requires an on-site cyclotron and radiochemistry production facilities and, regulatorily, will always be an FDA IND (investigational new drug application) rather than an NDA or ANDA (new drug application = “approved”) radiopharmaceutical. So, no matter how promising this technique may be, widespread adoption is unlikely. However, the mechanistic similarity between PIB and the other commercially-available amyloid imaging agents ([¹⁸F]florbetapir, [¹⁸F]florbetaben, [¹⁸F]flutemetamol), presents the possibility of translating this technique to one or more of these agents or to possibly one of the tau imaging agents. In this vein, the potential of early-phase [¹⁸F]florbetapir and [¹⁸F]florbetaben to parallel glucose metabolism from FDG imaging or CBF from ASL MRI have been/are being explored as spin-off studies in the Alzheimer’s Disease Neuroimaging Initiative (ADNI) portfolio.

Early PIB uptake has the potential to effectively serve as a surrogate for global and regional cerebral blood flow (CBF) measures. The strong correlation between PIB uptake and CBF may allow for a single imaging session to capture important clinical information on blood flow in addition to severity of amyloid load. Although, all early PIB SUV metrics explored were significantly related to the absolute cerebral blood flow for all subjects and for all regions across the brain, the simple and readily obtainable individual- specific SUV_peak value was the strongest predictor regionally and globally. Reliable global and regional cerebral blood flow measures along with an assessment of amyloid burden can be obtained with a single administration of [¹¹C]PIB.