Sage Journals: Discover world-class research

Abstract

Alzheimer’s disease is characterized by regional reductions in cerebral blood flow (CBF). Although the gold standard for measuring CBF is [¹⁵O]H₂O PET, proxies of relative CBF, derived from the early distribution phase of amyloid and tau tracers, have gained attention. The present study assessed precision of [¹⁵O]H₂O derived relative and absolute CBF, and compared precision of these measures with that of (relative) CBF proxies. Dynamic [¹⁵O]H₂O, [¹⁸F]florbetapir and [¹⁸F]flortaucipir PET test-retest (TrT) datasets with eleven, nine and fourteen subjects, respectively, were included. Analyses were performed using an arterial input model and/or a simplified reference tissue model, depending on the data available. Relative CBF values (i.e. K₁/K₁′ and/or R₁) were obtained using cerebellar cortex as reference tissue and TrT repeatability (i.e. precision) was calculated and compared between tracers, parameters and clinical groups. Relative CBF had significantly better TrT repeatability than absolute CBF derived from [¹⁵O]H₂O (r = −0.53), while best TrT repeatability was observed for [¹⁸F]florbetapir and [¹⁸F]flortaucipir R₁ (r = −0.23, r = −0.33). Furthermore, only R₁ showed, better TrT repeatability for cognitively normal individuals. High precision of CBF proxies could be due to a compensatory effect of the extraction fraction, although changes in extraction fraction could also bias these proxies, but not the gold standard.

Keywords

Cerebral blood flow PET precision Alzheimer’s disease healthy controls

Introduction

In Alzheimer’s disease (AD), regional reductions in cerebral blood flow (CBF) have been reported in the temporal, parietal and posterior cingulate regions,^1–3 predominantly in the later phases of the disease.^4,5 As perfusion is tightly coupled to metabolism,⁶ CBF measures may have clinical relevance in terms of determining disease severity as an indirect marker of neurodegeneration or for differential diagnosis.^5,7,8 Moreover, these reductions in perfusion have been linked to cognitive decline in AD.⁹

The gold standard technique for measuring CBF is [¹⁵O]H₂O positron emission tomography (PET). However, the short half-life of oxygen-15 (122 sec) restricts its application to a few specialised PET centres equipped with an on-site cyclotron. During the last decade, several studies have evaluated whether a proxy of CBF can be derived from the early distribution phase of PET tracers that measure amyloid-β (Aβ) or tau burden.^7,9–12 Promising results from studies investigating these proxies of relative CBF (i.e. R₁ or the early frame standardized uptake value ratio) have led to an increased interest in so called dual-phase or dual-time window acquisition protocols for these tracers, given that they allow for extracting two different biomarkers.^12–15 Apart from these PET-based methods, many other techniques exist for measuring CBF, such as arterial spin labelling (ASL) MRI, which has commonly been used in AD research.¹⁶

During the last decade, longitudinal PET studies have become increasingly prevalent in AD research, which emphasizes the importance of understanding the intrinsic variability of CBF measures to determine what magnitude of change signifies an actual change. Repeatability of [¹⁵O]H₂O PET derived absolute CBF measures has previously been characterized^17,18 and more recently, test-retest (TrT) repeatability has been characterized for a global cortical volume of interest (VOI) for [¹⁸F]florbetapir derived R₁ and across various ROIs for [¹¹C]PiB derived R₁.^19,20 The repeatability for these CBF proxies was substantially better than TrT repeatability of [¹⁵O]H₂O PET derived absolute CBF measures.¹⁷ It has been hypothesized that the lower TrT repeatability of the absolute CBF measures is mainly due to large global variations in CBF, which are normalized when calculating a relative measure of CBF. To contribute to the understanding of repeatability of CBF measures, the main purpose of the present study was to compare repeatability of [¹⁵O]H₂O PET derived relative and absolute CBF measures. In addition, repeatability of these [¹⁵O]H₂O PET derived measures was also compared with that of relative CBF proxies (i.e. relative tracer delivery K₁/K₁′ and/or R₁) estimated from [¹⁸F]florbetapir and [¹⁸F]flortaucipir scans across a range of VOIs. A secondary aim was to assess whether repeatability of relative CBF proxies was stable across diagnostic groups for [¹⁸F]florbetapir and [¹⁸F]flortaucipir.^9,21

Materials and methods

Participants & image acquisition

The following three separate datasets were retrospectively included. Before participating in the study, all participants provided written informed consent in accordance with the Declaration of Helsinki. Study protocols were approved by the local Medical Ethics Review Committee of the Amsterdam UMC, VUmc and in case of the [¹⁵O]H₂O study, also by the local Medical Ethics Review Committee of the Amsterdam UMC, AMC.

[¹⁵O]H₂O dataset

Eleven healthy control participants underwent repeated dynamic [¹⁵O]H₂O PET scanning with arterial blood sampling (described below) on the same day.²² Briefly, a low-dose CT was acquired for attenuation and scatter correction purposes and then, a 10 min dynamic PET scan was performed starting at tracer injection. PET scans were acquired on a Philips Gemini TF-64 PET/CT system (Philips Healthcare, Cleveland, TN, USA) and reconstructed into 25 frames of progressively increasing duration using the row action maximum likelihood algorithm (LOR-RAMLA) reconstruction algorithm, with the vendors’ default settings.²³ T1-weighted MRI scans were acquired for anatomical reference 1–7 days prior to the PET session at a Philips 3 T Intera system (Philips Healthcare, Best, the Netherlands).¹⁸

[¹⁸F]florbetapir dataset

Four cognitively normal (CN), Aβ-negative subjects and five AD dementia (Aβ-positive) patients underwent repeated dynamic [¹⁸F]florbetapir PET scans (average interval: 4.5 ± 3.0 weeks) with arterial blood sampling using a Philips Ingenuity TF PET/CT (Philips Medical Systems, Best, the Netherlands). Prior to the start of a 90 minutes dynamic PET scan, a low-dose CT was acquired. PET images were reconstructed into 22 frames using the BLOB (rotationally symmetric volume elements) ordered-subsets time of flight (BLOB-OS-TF) reconstruction algorithm, with the vendors’ default settings.²⁴ T1-weighted MRI scans were acquired on either a Signa HDxt MRI (General Electric, Milwaukee, WI) or an Ingenuity TF PET/MR (Philips Medical Systems, Cleveland, OH) scanner.¹⁹

[¹⁸F]flortaucipir dataset

A group of 14 participants, consisting of six CN participants (all tau-negative)^25,26 and eight patients with mild cognitive impairment due to AD or AD dementia (three tau-negative, five tau-positive),^25,26 underwent repeated dual-time window [¹⁸F]flortaucipir PET scans (average interval 3.0 ± 1.0 weeks) on a Philips Ingenuity TF PET/CT (Philips Medical Systems, Best, the Netherlands). A low-dose CT was acquired prior to the first phase of the [¹⁸F]flortaucipir PET scan (0–60 min post injection, p.i.). After a 20-min break, the low-dose CT was repeated and the second phase of the PET scan was acquired (80 to 130 min p.i.). PET images were reconstructed into 29 frames using BLOB-OS-TF reconstruction algorithm. T1-weighted MRI images were acquired <6 months from the PET scan on a 3.0 T Philips Ingenuity Time-of-Flight PET/MR scanner.²⁵

Arterial sampling procedure

Arterial blood concentrations were measured continuously using an online blood sampler system.²² In addition, manual samples were drawn at different times, i.e. at 5.5, 8 and 10 min for [¹⁵O]H₂O PET scans and at 5, 10, 20, 40, 60, 75 and 90 min for [¹⁸F]florbetapir PET scans. These samples were used to calibrate the online sampler curve and, for [¹⁸F]florbetapir, to estimate plasma-to-whole blood ratios and correct for plasma metabolite fractions. A metabolite corrected plasma input function was obtained after the aforementioned corrections were applied to the continuous online whole blood sampler curve.

Image processing

During the acquisition of the PET scans, movement was checked regularly using laser beams, and head position was corrected if necessary. A visual quality control was used to assess between-frame motion and scans with severe motion were not included in the present study. Between-frame motion of less than 5 mm was corrected using frame to frame ridged co-registration in VINCI software.²⁷ For all scans, T1-weighted MR images were coregistered to their corresponding PET images and the coregistered MR was segmented into grey matter, white matter and cerebrospinal fluid (CSF) using SPM8.²⁸ For [¹⁸F]flortaucipir, the two parts of the PET scan were first combined after coregistering them using VINCI software (Max Plank Institute, Cologne, Germany).²⁷ Next, volumes of interest (VOI) were delineated based on the Hammers atlas²⁹ as implemented in PVE-lab.³⁰ Time-activity curves were obtained for the following grey matter VOIs: medial and lateral anterior temporal lobe, posterior temporal lobe, superior, middle and inferior temporal gyrus, fusiform gyrus, parahippocampal and ambient gyrus, anterior and posterior cingulate gyrus, middle and orbitofrontal gyrus, gyrus rectus, inferior and superior frontal gyrus, pre- and post-central gyrus, superior parietal gyrus and the (infero)lateral remainder of the parietal lobe, and a region consisting of all grey matter voxels of the brain, total grey matter.

Kinetic modelling

The [¹⁵O]H₂O PET dataset had arterial plasma input data available and therefore, the single-tissue compartment model with two rate constants and blood volume fraction parameter was used to estimate K₁, which equals CBF (F) as the extraction fraction (E) is essentially equal to 100% (K₁ = E · F). In addition, to correct for global variations in perfusion, regional K₁ values were normalised by K₁′ (rate of influx of the tracer into the reference tissue) to obtain a measure of relative CBF (K₁/K₁′). For the second dataset consisting of [¹⁸F]florbetapir PET scans and arterial plasma input data, both the reversible two-tissue compartment model with four rate constants and additional blood volume fraction parameter (2T4k_V_b) and the simplified reference tissue model (SRTM)³¹ were used to determine K₁, K₁/K₁′ and R₁, respectively. Finally, for [¹⁸F]flortaucipir PET scans, no plasma input data were available and, therefore, only SRTM was used to derive regional R₁ values.¹⁹ For all three tracers included in this study, the cerebellar grey matter was used as reference tissue.

Statistical tests

All statistical analyses were performed in R (version: 4.0.2; R Foundation for Statistical Computing, Vienna, Austria).³² Results with p < 0.05 were considered statistically significant. The rank-biserial correlation r for non-parametric tests was used as a measure of effect size because it can be used for between and within study designs.^33–35 The rank_biserial function from the effectsizes package in R was used to compute this measure. Potential between-tracer differences in sex were investigated using a chi-square test, while differences in age were assessed using Kruskal-Wallis³⁶ and Mann Whitney U tests.³⁷ TrT repeatability was calculated according to equation (1) and, for purposes of visualizing TrT distributions, also non-absolute values were calculated and shown in Violin plots.

TrT repeatability (%) = \frac{| T - R |}{0.5 \cdot | T + R |} \cdot 100

(1)

Next, Bland-Altman analyses³⁸ were used to assess whether there was a bias between test and retest measures and to visualize measurement variability. Finally, potential differences in TrT repeatability were assessed for different tracers, metrics, diagnostic groups and groups based on the presence of the target pathology in case of the Aβ and tau tracers. Specifically, Mann-Whitney U tests were used for comparing metrics between different tracers and comparisons between diagnostic groups, while Wilcoxon Signed-Rank tests³⁹ were used for within tracer comparisons between metrics.

Results

An overview of the characteristics of the three datasets can be found in Table 1. The only between-tracer difference was in age (χ² = 21.89, p < 0.001), with the [¹⁵O]H₂O dataset consisting of significantly younger participants compared with the other two datasets (both p < 0.001, r = −1.00). Relative CBF values per tracer, diagnostic group and for both test and retest measurements can be found in Supplementary Table 1. Here, it can be observed that for both cognitively normal and cognitively impaired individuals, relative CBF values were comparable between tracers.

Table 1.

Subject demographics.

	[¹⁵O]H₂O	[¹⁸F]florbetapir		[¹⁸F]flortaucipir
	CN (N = 11)	CN (N = 4)	AD (N = 5)	CN (N = 6)	MCI/AD (N = 8)
Age	22.4 ± 1.7	63.3 ± 2.6	65.0 ± 4.2	64.5 ± 8.8	65.5 ± 9.8
Sex (% female)	45.5%	50%	80%	50%	50%
Injected dose (MBq)
Test	1100	292 ± 35		237 ± 15
Retest	1100	312 ± 11		245 ± 18
Scan interval	21.0 ± 1.0 min	4.5 ± 3.0 weeks		3.0 ± 1.0 weeks

CN: cognitively normal; AD: Alzheimer’s disease dementia; MCI: mild cognitive impairment. Values are depicted as mean ± SD.

Figure 1 shows the distribution of (non-absolute) TrT values for each tracer and parameter. Bland-Altman analyses showed that average differences (mean ± SD) between test and retest measures ranged from −0.1 to −2.0% for relative measures: [¹⁵O]H₂O K₁/K₁′: −2.0 ± 7.8%, [¹⁸F]florbetapir K₁/K₁′: −1.3 ± 5.4%, [¹⁸F]florbetapir R_1: −1.7 ± 4.1%, [¹⁸F]flortaucipir R₁: −0.1 ± 4.2,% while for the absolute measure (i.e. [¹⁵O]H₂O K₁) it was −5.7 ± 14.6% (Figure 2). The highest variability between test and retest measures was observed for [¹⁵O]H₂O K₁. Of note, for [¹⁵O]H₂O K₁, a cluster of data points with larger differences between test and retest measures was observed, corresponding to a single subject (Figure 2).

Figure 1.

Violin plot showing the distribution of (non-absolute) test-retest values (%) for each tracer and parameter. Dotted coloured lines correspond to quartiles and coloured dashed lines to the median.

Figure 2.

Bland-Altman plots illustrating the test-retest repeatability (%) for (a) [¹⁵O]H₂O K₁, (b) [¹⁵O]H₂O K₁/K₁′, (c) [¹⁸F]florbetapir K₁, (d) [¹⁸F]florbetapir K₁/K₁′, (e) [¹⁸F]florbetapir R₁ and (f) [¹⁸F]flortaucipir R₁. Dotted lines correspond to 95% Limits of Agreement and dashed lines to the average % difference.

Test-retest repeatability

[¹⁵O]H₂O

Across all participants, average TrT repeatability for K₁ was best in the orbitofrontal gyrus 7.8 ± 12.3% and worst in fusiform gyrus 17.7 ± 10.2. For K₁/K₁′, TrT repeatability ranged from 3.1 ± 3.0 in total grey matter to 10.5 ± 6.1 in the fusiform gyrus (Table 2). Overall, TrT repeatability significantly improved for K₁/K₁′ compared with K₁ (V = 7510, p < 0.001, r = −0.53).

Table 2.

Average test-retest repeatability (%) of [¹⁵O]H₂O K₁ and K₁/K₁′.

Regions	CBF (K₁)	Relative CBF (K₁/K₁′)
Anterior temporal lobe medial part	14.1 ± 9.5	8.3 ± 5.9
Anterior temporal lobe lateral part	13.7 ± 10.2	6.7 ± 7.0
Parahippocampal and ambient gyri	13.1 ± 8.2	7.0 ± 4.8
Superior temporal gyrus	12.6 ± 14.3	8.0 ± 5.1
Middle and inferior temporal gyri	10.7 ± 10.8	5.2 ± 5.7
Fusiform gyrus	17.7 ± 10.2	10.5 ± 6.1
Insula	11.0 ± 9.7	4.3 ± 4.0
Lateral remainder of occipital lobe	11.9 ± 11.8	8.1 ± 6.3
Gyrus cinguli anterior part	10.5 ± 10.1	5.4 ± 3.3
Gyrus cinguli posterior part	9.6 ± 9.4	4.6 ± 3.8
Middle frontal gyrus	9.4 ± 11.7	5.0 ± 3.1
Posterior temporal lobe	10.9 ± 11.8	4.9 ± 4.3
Inferolateral remainder of parietal lobe	11.2 ± 11.3	3.7 ± 3.4
Precentral gyrus	11.7 ± 8.8	4.9 ± 5.3
Gyrus rectus	13.7 ± 12.5	8.3 ± 5.3
Orbitofrontal gyri	7.8 ± 12.3	4.4 ± 3.6
Inferior frontal gyrus	9.5 ± 13.7	4.0 ± 3.5
Superior frontal gyrus	8.6 ± 10.1	4.5 ± 4.1
Postcentral gyrus	12.0 ± 8.0	5.0 ± 4.0
Superior parietal gyrus	9.9 ± 9.8	4.8 ± 5.1
Lingual gyrus	13.2 ± 11.0	8.7 ± 8.6
Cuneus	13.8 ± 10.7	8.9 ± 8.6
Total grey matter	9.7 ± 10.4	3.1 ± 3.0

All values are depicted as mean ± SD.

[¹⁸F]florbetapir

Across participants, average TrT repeatability K₁ ranged from 14.7 ± 8.0% in the middle and inferior temporal gyri to 19.8 ± 10.0% in the inferolateral remainder of parietal lobe (Supplementary Table 2). TrT repeatability for K₁ was worse in CN (Aβ-negative) subjects compared with AD (Aβ-positive) patients (U = 6419, p = 0.008, r = 0.21). More specifically, for CN subjects, average TrT repeatability for K₁ ranged from 15.5 ± 4.0% in the middle and inferior temporal gyri to 22.8 ± 12.5% in the posterior part of the gyrus cinguli, while in case of AD patients, TrT repeatability for K₁ ranged from 11.2 ± 9.9% in the parahippocampal and ambient gyri to 20.1 ± 11.9% in the inferolateral remainder of parietal lobe. For K₁/K₁′, average TrT repeatability ranged from 1.4 ± 1.4% in total grey matter to 6.3 ± 6.7% in the orbitofrontal gyrus (Table 3). There was no significant difference in TrT repeatability for K₁/K₁′ between diagnostic groups (r = 0.00). For R₁, average TrT repeatability across subjects ranged from 2.1 ± 1.1% in total grey matter to 5.2 ± 2.0% in the fusiform gyrus (Table 4). Overall, TrT repeatability for R₁ was better in CN (Aβ-negative) subjects compared with AD (Aβ-positive) patients (U = 4136, p = 0.007, r = −0.22). More specifically, for CN subjects, average TrT repeatability for R₁ ranged from 1.0 ± 0.9% in the insula to 6.0 ± 4.4% in the posterior part of the gyrus cinguli, while in case of AD patients, TrT repeatability for R₁ ranged from 1.7 ± 0.9% in the superior temporal gyrus to 6.4 ± 4.0% in the orbitofrontal gyri. There was no significant difference in TrT repeatability between K₁/K₁′ and R₁ (r = 0.13), while both measures showed better TrT repeatability than K₁ (U = 20477, p < 0.001, r = −0.90 and U = 21046, p < 0.001, r = −0.96, for K₁/K₁′ and R₁, respectively).

Table 3.

Average test-retest repeatability (%) of [¹⁸F]florbetapir K₁/K₁′.

Region	All (N = 9)	CN (N = 4)	AD (N = 5)
Anterior temporal lobe medial part	4.2 ± 5.4	3.1 ± 2.6	5.0 ± 7.1
Anterior temporal lobe lateral part	4.7 ± 5.3	3.4 ± 1.9	5.7 ± 7.1
Parahippocampal and ambient gyri	3.5 ± 2.5	1.5 ± 1.4	5.1 ± 1.8
Superior temporal gyrus	3.6 ± 1.5	3.8 ± 2.2	3.4 ± 0.7
Middle and inferior temporal gyri	2.9 ± 2.9	4.6 ± 3.4	1.5 ± 1.6
Fusiform gyrus	3.0 ± 2.6	3.7 ± 2.4	2.5 ± 3.0
Insula	4.4 ± 6.3	2.7 ± 1.0	5.7 ± 8.6
Lateral remainder of occipital lobe	4.2 ± 3.9	2.4 ± 2.4	5.6 ± 4.4
Gyrus cinguli anterior part	5.7 ± 5.0	3.3 ± 1.1	7.6 ± 6.3
Gyrus cinguli posterior part	4.7 ± 2.7	5.1 ± 3.0	4.3 ± 2.7
Middle frontal gyrus	3.7 ± 2.9	3.3 ± 3.7	4.0 ± 2.5
Posterior temporal lobe	4.0 ± 4.7	3.3 ± 2.3	4.6 ± 6.3
Inferolateral remainder of parietal lobe	5.8 ± 5.1	5.1 ± 2.3	6.4 ± 6.9
Precentral gyrus	2.8 ± 1.9	3.2 ± 2.4	2.5 ± 1.6
Gyrus rectus	3.7 ± 4.3	2.5 ± 1.6	4.7 ± 5.6
Orbitofrontal gyri	6.3 ± 6.7	3.5 ± 3.0	8.5 ± 8.3
Inferior frontal gyrus	3.3 ± 3.3	3.3 ± 2.9	3.3 ± 4.0
Superior frontal gyrus	5.4 ± 5.1	3.2 ± 3.9	7.2 ± 5.6
Postcentral gyrus	4.4 ± 2.6	4.3 ± 1.8	4.5 ± 3.3
Superior parietal gyrus	4.0 ± 1.5	5.0 ± 0.5	3.2 ± 1.6
Lingual gyrus	3.4 ± 2.9	4.2 ± 1.4	2.7 ± 3.8
Cuneus	3.0 ± 2.4	2.8 ± 1.7	3.0 ± 3.0
Total grey matter	1.4 ± 1.4	2.1 ± 1.7	0.9 ± 1.0

All values are depicted as mean ± SD; CN: cognitively normal; AD: Alzheimer’s disease dementia.

Table 4.

Average test-retest repeatability (%) of [¹⁸F]florbetapir R_1.

Region	All (N = 9)	CN (N = 4)	AD (N = 5)
Anterior temporal lobe medial part	3.9 ± 2.6	3.4 ± 0.9	4.4 ± 3.5
Anterior temporal lobe lateral part	4.9 ± 3.2	3.7 ± 1.9	5.9 ± 3.9
Parahippocampal and ambient gyri	3.9 ± 2.6	1.6 ± 0.9	5.8 ± 2.0
Superior temporal gyrus	2.6 ± 2.1	3.8 ± 2.6	1.7 ± 1.0
Middle and inferior temporal gyri	4.5 ± 4.0	5.8 ± 4.9	3.6 ± 3.3
Fusiform gyrus	5.2 ± 2.0	4.1 ± 0.9	6.1 ± 2.3
Insula	2.5 ± 2.5	1.0 ± 0.9	3.6 ± 3.0
Lateral remainder of occipital lobe	2.6 ± 2.2	2.1 ± 1.6	3.1 ± 2.6
Gyrus cinguli anterior part	3.9 ± 2.8	3.4 ± 2.6	4.3 ± 3.2
Gyrus cinguli posterior part	4.5 ± 3.3	6.0 ± 4.4	3.3 ± 1.8
Middle frontal gyrus	4.2 ± 2.0	2.5 ± 1.1	5.6 ± 1.3
Posterior temporal lobe	3.2 ± 2.5	3.5 ± 3.1	3.0 ± 2.3
Inferolateral remainder of parietal lobe	3.8 ± 1.6	3.7 ± 1.3	3.9 ± 1.9
Precentral gyrus	2.5 ± 1.6	1.5 ± 1.2	3.3 ± 1.4
Gyrus rectus	3.7 ± 3.2	3.8 ± 2.4	3.7 ± 4.0
Orbitofrontal gyri	4.7 ± 4.1	2.5 ± 3.5	6.4 ± 4.0
Inferior frontal gyrus	4.0 ± 3.7	2.8 ± 1.6	4.9 ± 4.8
Superior frontal gyrus	4.4 ± 2.4	2.5 ± 1.9	5.8 ± 1.7
Postcentral gyrus	3.4 ± 1.4	3.0 ± 0.8	3.7 ± 1.8
Superior parietal gyrus	3.0 ± 1.9	3.2 ± 1.3	2.8 ± 2.3
Lingual gyrus	2.3 ± 1.7	2.6 ± 2.1	2.0 ± 1.4
Cuneus	2.6 ± 2.1	1.6 ± 1.3	3.4 ± 2.4
Total grey matter	2.1 ± 1.1	1.9 ± 1.0	2.2 ± 1.3

All values are depicted as mean ± SD. CN: cognitively normal; AD: Alzheimer’s disease dementia.

[¹⁸F]flortaucipir

Across participants, average TrT repeatability for R₁ ranged from 1.8 ± 2.0% in total grey matter to 4.3 ± 3.3% in the medial part of the anterior temporal lobe (Table 5). TrT repeatability was better in case of CN participants compared with AD patients (U = 10504, p = 0.008, r = −0.17). In CN participants, average TrT repeatability for R₁ ranged from 0.6 ± 0.5% in the posterior temporal lobe to 4.8 ± 3.3% in the gyrus rectus. In AD patients, it ranged from 1.4 ± 1.0% in the precentral gyrus to 8.5 ± 8.3% in the orbitofrontal gyri. There were no differences in TrT repeatability between tau-negative and tau-positive groups (r = −0.04).

Table 5.

Average test-retest repeatability (%) of [¹⁸F]flortaucipir R₁.

Regions	All (N = 14)	CN (N = 6)	AD (N = 8)
Anterior temporal lobe medial part	4.3 ± 3.3	2.9 ± 3.6	5.4 ± 2.6
Anterior temporal lobe lateral part	2.5 ± 2.2	1.7 ± 1.3	3.1 ± 2.6
Parahippocampal and ambient gyri	3.9 ± 3.0	3.0 ± 2.4	4.5 ± 3.2
Superior temporal gyrus	3.3 ± 2.2	2.8 ± 2.6	3.7 ± 1.7
Middle and inferior temporal gyri	2.5 ± 2.2	2.2 ± 1.9	2.7 ± 2.3
Fusiform gyrus	2.5 ± 2.2	1.8 ± 2.0	3.0 ± 2.3
Insula	2.5 ± 2.4	1.0 ± 0.8	3.7 ± 2.5
Lateral remainder of occipital lobe	2.9 ± 2.9	1.2 ± 0.7	4.1 ± 3.3
Gyrus cinguli anterior part	2.8 ± 2.1	2.6 ± 1.3	2.9 ± 2.5
Gyrus cinguli posterior part	2.6 ± 1.8	1.5 ± 1.4	3.5 ± 1.6
Middle frontal gyrus	3.0 ± 1.8	3.5 ± 2.2	2.7 ± 1.2
Posterior temporal lobe	1.8 ± 2.0	0.6 ± 0.5	2.7 ± 2.2
Inferolateral remainder of parietal lobe	3.2 ± 3.3	1.9 ± 1.5	4.1 ± 3.8
Precentral gyrus	1.9 ± 1.1	2.6 ± 0.9	1.4 ± 1.0
Gyrus rectus	4.1 ± 3.0	4.8 ± 3.3	3.5 ± 2.6
Orbitofrontal gyri	3.2 ± 2.6	3.3 ± 2.9	3.2 ± 2.5
Inferior frontal gyrus	2.0 ± 2.0	2.1 ± 2.2	2.0 ± 1.8
Superior frontal gyrus	3.3 ± 2.4	3.7 ± 2.8	3.0 ± 2.0
Postcentral gyrus	2.3 ± 1.3	2.0 ± 1.0	2.6 ± 1.4
Superior parietal gyrus	3.8 ± 3.8	2.1 ± 1.4	5.1 ± 4.4
Lingual gyrus	2.7 ± 2.3	2.7 ± 1.9	2.7 ± 2.6
Cuneus	3.0 ± 2.3	2.1 ± 1.8	3.6 ± 2.5
Total grey matter	1.8 ± 1.3	1.3 ± 1.1	2.2 ± 1.3

All values are depicted as mean ± SD, CN: cognitively normal, AD: Alzheimer’s disease dementia.

Tracer comparisons

Between tracer comparisons showed that [¹⁵O]H₂O derived K₁ had better TrT repeatability than [¹⁸F]florbetapir derived K₁ (U = 35851, p < 0.001, r = −0.37). Nonetheless, [¹⁸F]florbetapir derived K₁/K₁′ and R₁ (U = 20323, p < 0.001, r = −0.22 and U = 20064, p < 0.001, r = −0.23, respectively), and [¹⁸F]flortaucipir derived R₁ (U = 27096, p < 0.001, r = −0.33) had better TrT repeatability than [¹⁵O]H₂O K₁/K₁′. Furthermore, [¹⁸F]flortaucipir derived R₁ showed better TrT repeatability than [¹⁸F]florbetapir derived R₁ (U = 28608, p = 0.006, r = −0.14).

Discussion

The present study assessed precision of absolute and relative CBF measures through retrospective analysis of a dynamic [¹⁵O]H₂O PET dataset. Furthermore, precision of [¹⁵O]H₂O derived absolute/relative CBF was compared with that of (relative) CBF proxies (i.e. K₁, K1/K₁′ and R₁) derived from dynamic [¹⁸F]florbetapir and [¹⁸F]flortaucipir PET scans.

As expected, for [¹⁵O]H₂O PET, TrT repeatability of relative CBF was higher than that of absolute CBF and the same was found for [¹⁸F]florbetapir. The finding that K₁/K₁′ is less variable than K₁, might be because it contains an intrinsic correction for global fluctuations in CBF,¹⁷ and/or because any measurement errors in the arterial input function is cancelled out.

Furthermore, better TrT repeatability was observed for [¹⁵O]H₂O derived K₁ compared with [¹⁸F]florbetapir derived K₁. A factor contributing to this finding could be the difference in acquisition time between test and retest [¹⁸F]florbetapir scans.⁴⁰ For [¹⁸F]florbetapir and [¹⁸F]flortaucipir, TrT repeatability of relative CBF (K₁/K₁′ and/or R₁) was significantly better than that of [¹⁵O]H₂O PET with a higher effect size for [¹⁸F]flortaucipir compared with [¹⁸F]florbetapir. Apart from differences in age and group composition, which could have affected TrT repeatability and will therefore be discussed in more detail later, this finding may be explained by several methodological factors. First, for tracers other than [¹⁵O]H₂O, K₁ is not only determined by flow, but also by the extraction fraction (K₁ = E · F). Thus, fluctuations in flow might be compensated by changes in extraction fraction. Although higher TrT repeatability (i.e. higher precision) is desirable, it is important to note that this higher repeatability of these relative CBF proxies may, at least in part, be due to dissociation from CBF itself (i.e. by a compensatory change in extraction fraction), which of course is not the case for [¹⁵O]H₂O PET. Another plausible explanation for the better TrT repeatability observed with [¹⁸F]florbetapir and [¹⁸F]flortaucipir, is the fact that oxygen-15 has a half-life of only 2 min, which means that the total number of counts in the [¹⁵O]H₂O scans was much lower than in the [¹⁸F]florbetapir and [¹⁸F]flortaucipir scans. This results in higher noise levels in the [¹⁵O]H₂O scans, and hence poorer measurement repeatability. The improved TrT repeatability observed for the relative CBF proxies does not necessarily indicate that this measure would be preferred over the gold standard, [¹⁵O]H₂O PET. As mentioned above, the relative CBF proxies R₁ or K₁/K₁′ might be biased through changes in extraction fraction, making them less accurate measures than K₁ or K₁/K₁′ derived from [¹⁵O]H₂O PET. Furthermore, an important disadvantage of these relative CBF proxies is that, by definition, they can only be used to measure relative changes in flow. Therefore, they cannot be used in research studies that aim to investigate (or are affected by) global changes in CBF. Finally, when designing a research study, it should also be taken into account that R₁ might be less sensitive than the gold standard for measuring changes over time, as reported previously.¹¹

In the diagnostic group comparisons, better TrT repeatability was observed for CN participants compared with the cognitively impaired groups for R₁ derived from [¹⁸F]florbetapir and [¹⁸F]flortaucipir scans, which might be related to the small datasets used for this comparison and R₁’s relatively small SD. Furthermore, it should be noted that there was a clear difference in age and group composition for the [¹⁵O]H₂O PET dataset compared with the [¹⁸F]florbetapir and [¹⁸F]flortaucipir datasets, possibly impacting the results. Furthermore, to date, it remains unclear whether TrT repeatability is comparable between young ([¹⁵O]H₂O dataset) and elderly ([¹⁸F]florbetapir and [¹⁸F]flortaucipir datasets) CN participants, which could also have had an effect on the present results. Evidently, in an ideal scenario, repeatability of all tracers would have been compared within the same participant/patient sample. However, considering ethical regulations regarding radiation exposure of healthy individuals, such a design is not feasible, at least not for the sole reason of comparing K₁ or R₁ repeatability.

Regional differences in TrT repeatability were also observed for all tracers, although no clear pattern could be established. These regional differences may be related to various methodological factors such as size of the region, signal strength, effects of signal spill-in/spill-out from/to adjacent regions, but also biological factors such as atrophy or ageing, considering that these effects are not uniform across the brain.^41,42 Nonetheless, across tracers, best TrT repeatability of relative CBF was observed for the largest region, which comprised total grey matter. The magnitude of the observed TrT repeatability of R₁ estimated from [¹⁸F]florbetapir and [¹⁸F]flortaucipir scans was very comparable to what has previously been reported for [¹¹C]PiB, despite the older scanner (Siemens ECAT EXACT HR+) that was used in that study. More specifically, across a nearly identical set of regions and in a group of CN participants, and MCI and AD dementia patients, [¹¹C]PiB R₁ TrT repeatability ranged from 1.5 to 5.8%, while in the present study R₁ TrT repeatability was 2.1–5.2% for [¹⁸F]florbetapir and 1.8–4.3% for [¹⁸F]flortaucipir. This suggests that despite distinct tracer kinetics and differences in group composition, R₁ remains a stable metric.

Furthermore, it is important to note, that there was a difference in timing between scans in the [¹⁵O]H₂O dataset, where all participants received their repeat scan within one day, compared with the [¹⁸F]florbetapir and [¹⁸F]flortaucipir datasets, where participants received their repeat scan within a few weeks. Although in a prospective study, timing between repeat scans ideally would have been similar between studies, no changes in CBF are to be expected within a few weeks,^43,44 thus no considerable effects on the results were expected. Finally, for [¹⁸F]flortaucipir, no arterial blood data had been collected, which prevented a direct comparison of repeatability of the K₁/K₁′ parameter between all three tracers. However, from a theoretical perspective, no substantial differences were expected in TrT between K₁/K₁′ and R₁, which was also confirmed by comparisons performed using the [¹⁸F]florbetapir dataset.

Conclusion

Relative CBF showed higher precision than absolute CBF derived from [¹⁵O]H₂O PET scans. Furthermore, relative CBF proxies derived from commonly used amyloid-β and tau tracers appeared to have even higher precision, possibly due to a compensatory effect of extraction fraction. It is important to keep in mind that changes in extraction fraction may bias these proxies, but not the gold standard, [¹⁵O]H₂O PET.

Supplemental Material

sj-pdf-1-jcb-10.1177_0271678X221135270 - Supplemental material for Precision estimates of relative and absolute cerebral blood flow in Alzheimer’s disease and cognitively normal individuals

Supplemental material, sj-pdf-1-jcb-10.1177_0271678X221135270 for Precision estimates of relative and absolute cerebral blood flow in Alzheimer’s disease and cognitively normal individuals by Fiona Heeman, Denise Visser, Maqsood Yaqub, Sander Verfaillie, Tessa Timmers, Yolande AL Pijnenburg, Wiesje M van der Flier, Bart NM van Berckel, Ronald Boellaard, Adriaan A Lammertsma, Sandeep SV Golla in Journal of Cerebral Blood Flow & Metabolism

Footnotes

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The acquisitions mentioned in the Acknowledgements section are supported by grants from ZonMW and from the Nuts-Ohra Foundation (Amsterdam, The Netherlands): 0903-044 and 1002-03.

Acknowledgements

The authors would like to thank the staff of the department of Radiology and Nuclear Medicine of the Amsterdam UMC, location VUmc for skilful acquisition of the scans. Acquisition of [¹⁸F]flortaucipir and [¹⁸F]florbetapir PET scans were made possible by Avid Radiopharmeuticals Inc., a wholly owned subsidiary of Eli Lilly and Company (NYSE: LLY).

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Authors’ contributions

FH, DV, AAL and SSVG contributed to the concept and design of the study. FH, DV, MY, SV, TT, RB, YALP contributed to the data acquisition and analysis. FH, DV, SSVG and AAL interpreted the data and drafted the manuscript. FH, DV, MY, SV, TT, YALP, WMF, BNMB, RB, AAL and SSVG read, critically reviewed, and approved the final manuscript.

ORCID iDs

Fiona Heeman

Denise Visser

Wiesje M. van der Flier

Adriaan A Lammertsma

Supplemental material

Supplemental material for this article is available online.

References

Jagust

Eberling

Reed

, et al. Clinical studies of cerebral blood flow in Alzheimer’s disease. Ann N Y Acad Sci 1997; 826: 254–262.

Binnewijzend

MAA

Benedictus

Kuijer

JPA

, et al. Cerebral perfusion in the predementia stages of alzheimer’s disease. Eur Radiol 2016; 26: 506–514.

Postiglione

Lassen

Holman

BL.

Cerebral blood flow in patients with dementia of Alzheimer’s type. Aging (Milano) 1993; 5: 19–26.

Binnewijzend

MAA

Kuijer

JPA

Benedictus

, et al. Cerebral blood flow measured with 3D pseudocontinuous arterial spin-labeling MR imaging in Alzheimer disease and mild cognitive impairment: a marker for disease severity. Radiology 2013; 267: 221–230.

Ottoy

Verhaeghe

Niemantsverdriet

, et al. 18F-FDG PET, the early phases and the delivery rate of 18F-AV45 PET as proxies of cerebral blood flow in Alzheimer’s disease: validation against 15O-H₂O PET. Alzheimer’s & Dementia 2019; 15: 1172–1182.

Schroeter

Stein

Maslowski

, et al. Neural correlates of Alzheimer’s disease and mild cognitive impairment: a systematic and quantitative Meta-analysis involving 1351 patients. Neuroimage 2009; 47: 1196–1206.

Tiepolt

Hesse

Patt

Luthardt

, et al. Early [¹⁸F]Florbetaben and [¹¹C]PiB PET images are a surrogate biomarker of neuronal injury in Alzheimer’s disease. Eur J Nucl Med Mol Imaging 2016; 43: 1700–1709. Early

Son

Kang

P-W

, et al. Early-Phase 18F-Florbetaben PET as an alternative modality for 18F-FDG PET. Clin Nucl Med 2020; 45: e8–e14.

Visser

Wolters

Verfaillie

SCJ

, et al. Tau pathology and relative cerebral blood flow are independently associated with cognition in Alzheimer’s disease. Eur J Nucl Med Mol Imaging 2020; 47: 3165–3175.

10.

Chen

Rosario

Mowrey

, et al. Relative 11C-PiB delivery as a proxy of relative CBF: Quantitative evaluation using single-session 15O-Water and 11C-PiB PET. J Nucl Med 2015; 56: 1199–1205.

11.

Bilgel

Beason-Held

, et al. Longitudinal evaluation of surrogates of regional cerebral blood flow computed from dynamic amyloid PET imaging. J Cereb Blood Flow Metab 2020; 40: 288–297.

12.

Yoon

H-J

Sahn Kim

Hyang Jeong

, et al. Dual-phase 18 F-Florbetaben PET provides cerebral perfusion proxy along with beta-amyloid burden in Alzheimer’s disease. NeuroImage: Clin 2021; 31: 102773.

13.

Bullich

Barthel

Koglin

, et al. Validation of Non-Invasive tracer kinetic analysis of 18 F-Florbetaben PET using a dual Time-Window acquisition protocol. J Nucl Med 2017; 59: 1104–1110.

14.

Heeman

Yaqub

Lopes Alves

, et al. Optimized dual-time-window protocols for quantitative [18F]flutemetamol and [18F]florbetaben PET studies. EJNMMI Res 2019; 9: 1–14.

15.

Tuncel

Visser

Yaqub

, et al. Effect of shortening the scan duration on quantitative accuracy of [18F]flortaucipir studies. Mol Imaging Biol 2021; 23: 604–613.

16.

Zhang

Gordon

Goldberg

TE.

Cerebral blood flow measured by arterial spin labeling MRI at resting state in normal aging and Alzheimer’s disease. Neurosci Biobehav Rev 2017; 72: 168–175.

17.

Bremmer

van Berckel

BNM

Persoon

, et al. Day-to-day test-retest variability of CBF, CMRO2, and OEF measurements using dynamic 15O PET studies. Mol Imaging Biol 2011; 13: 759–768.

18.

Heijtel

DFR

Mutsaerts

HJMM

Bakker

, et al. Accuracy and precision of pseudo-continuous arterial spin labeling perfusion during baseline and hypercapnia: a head-to-head comparison with ¹⁵O H₂O positron emission tomography. Neuroimage 2014; 92: 182–192.

19.

Golla

Verfaillie

Boellaard

, et al. Quantification of [18F]florbetapir: a test–retest tracer kinetic modelling study. J Cereb Blood Flow Metab 2018; 39: 2172–2180.

20.

Heeman

Hendriks

Lopes Alves

, et al. Test-Retest variability of relative tracer delivery rate as measured by [11C]PiB. Mol Imaging Biol 2021; 23: 335–339.

21.

Visser

Tuncel

Ossenkoppele

, et al. Longitudinal tau PET using [¹⁸ F]flortaucipir: comparison of (semi)quantitative parameters. Radiol Imaging 2021; doi:10.1101/2021.08.27.21262740.

22.

Boellaard

van Lingen

van Balen

SCM

, et al. Characteristics of a new fully programmable blood sampling device for monitoring blood radioactivity during PET. Eur J Nucl Med 2001; 28: 81–89.

23.

Surti

Kuhn

Werner

, et al. Performance of Philips Gemini TF PET/CT scanner with special consideration for its time-of-flight imaging capabilities. J Nucl Med 2007; 48: 471–480.

24.

Kolthammer

K-H

Grover

, et al. Performance evaluation of the ingenuity TF PET/CT scanner with a focus on high count-rate conditions. Phys Med Biol 2014; 59: 3843–3859.

25.

Timmers

Ossenkoppele

Visser

, et al. Test-retest repeatability of [18F]flortaucipir PET in Alzheimer’s disease and cognitively normal individuals. J Cereb Blood Flow Metab 2020; 40: 2464–2474.

26.

Leuzy

Pascoal

Strandberg

, et al. A multicenter comparison of [18F]flortaucipir, [18F]RO948, and [18F]MK6240 tau PET tracers to detect a common target ROI for differential diagnosis. Eur J Nucl Med Mol Imaging 2021; 48: 2295–2305.

27.

Vollmar

Cizek

Sué

, et al. VINCI-volume imaging in neurological research, co-registration and ROIs included. Forschung Und Wissenschaftliches Rechnen 2004: 115–131.

28.

Ashburner

Friston

KJ.

Unified segmentation. NeuroImage 2005; 26: 839–851.

29.

Hammers

Allom

Koepp

, et al. Three-dimensional maximum probability atlas of the human brain, with particular reference to the temporal lobe. Hum Brain Mapp 2003; 19: 224–247.

30.

Rask

Dyrby

Comerci

, et al. Software for correction of functional images for partial volume errors. NeuroImage 2004; 22.

31.

Lammertsma

Hume

SP.

Simplified reference tissue model for PET receptor studies. Neuroimage 1996; 4: 153–158.

32.

R, Core Team R. A Language and environment for statistical computing, www.R-project.org/ (2020, accessed 21 April 2022).

33.

Cureton

EE.

Rank-biserial correlation. Psychometrika 1956; 21: 287–290.

34.

Glass

GV.

A ranking variable analogue of biserial correlation: implications for short-cut item analysis. J Educ Meas 1965; 2: 91–95.

35.

Kerby

DS.

The simple difference formula: an approach to teaching nonparametric correlation. Comprehen Psychol 2014; 3: 11–1T.

36.

Kruskal

Wallis

WA.

Use of ranks in one-criterion variance analysis. J Am Stat Assoc 1952; 47: 583–621.

37.

Mann

Whitney

DR.

On a test of whether one of two random variables is stochastically larger than the other. Ann Math Statist 1947; 18: 50–60.

38.

Martin Bland

Altman

DG.

Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 327: 307–310.

39.

Breslow

A generalized Kruskal-Wallis test for comparing K samples subject to unequal patterns of censorship. Biometrika 1970; 57: 579–594.

40.

Elvsåshagen

Mutsaerts

Zak

, et al. Cerebral blood flow changes after a day of wake, sleep, and sleep deprivation. Neuroimage 2019; 186: 497–509.

41.

Scahill

Schott

Stevens

, et al. Mapping the evolution of regional atrophy in Alzheimer’s disease: unbiased analysis of fluid-registered serial MRI. Proc Natl Acad Sci USA 2002; 99: 4703–4707.

42.

Peters

Ageing and the brain. Postgrad Med J 2006; 82: 84–88.

43.

Frackowiak

RSJ.

Measurement and imaging of cerebral function in ageing and dementia [Chapter 5]. In: Progress in brain research. Amsterdam: Elsevier, 1986, pp.69–85.

44.

Pantano

Baron

Lebrun-Grandié

, et al. Regional cerebral blood flow and oxygen consumption in human aging. Stroke 1984; 15: 635–641.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.29 MB

Precision estimates of relative and absolute cerebral blood flow in Alzheimer’s disease and cognitively normal individuals

Abstract

Keywords

Introduction

Materials and methods

Participants & image acquisition

[15O]H2O dataset

[18F]florbetapir dataset

[18F]flortaucipir dataset

Arterial sampling procedure

Image processing

Kinetic modelling

Statistical tests

Results

Test-retest repeatability

[15O]H2O

[18F]florbetapir

[18F]flortaucipir

Tracer comparisons

Discussion

Conclusion

Supplemental Material

sj-pdf-1-jcb-10.1177_0271678X221135270 - Supplemental material for Precision estimates of relative and absolute cerebral blood flow in Alzheimer’s disease and cognitively normal individuals

Footnotes

Funding

Acknowledgements

Declaration of conflicting interests

Authors’ contributions

ORCID iDs

Supplemental material

References

Supplementary Material

[¹⁵O]H₂O dataset

[¹⁸F]florbetapir dataset

[¹⁸F]flortaucipir dataset

[¹⁵O]H₂O

[¹⁸F]florbetapir

[¹⁸F]flortaucipir