ArticlePDF Available

Abstract and Figures

Mapping the neuronal response during cognitive processing is of crucial importance to gain new insights into human brain function. BOLD imaging and ASL are established MRI methods in this endeavor. Recently, the novel approach of functional PET (fPET) was introduced, enabling absolute quantification of glucose metabolism at rest and during task execution in a single measurement. Here, we report test-retest reliability of fPET in direct comparison to BOLD imaging and ASL. Twenty healthy subjects underwent two PET/MRI measurements, providing estimates of glucose metabolism, cerebral blood flow (CBF) and blood oxygenation. A cognitive task was employed with different levels of difficulty requiring visual-motor coordination. Task-specific neuronal activation was robustly detected with all three imaging approaches. The highest reliability was obtained for glucose metabolism at rest. Although this dropped during task performance it was still comparable to that of CBF. In contrast, BOLD imaging yielded high performance only for qualitative spatial overlap of task effects but not for quantitative comparison. Hence, the combined assessment of fPET and ASL offers reliable and simultaneous absolute quantification of glucose metabolism and CBF at rest and task.
Content may be subject to copyright.
Original Article
Reliability of task-specific neuronal
activation assessed with functional
PET, ASL and BOLD imaging
Lucas Rischka
1
, Godber M Godbersen
1
, Verena Pichler
2,3
,
Paul Michenthaler
1
, Sebastian Klug
1
, Manfred Kl
obl
1
,
Vera Ritter
1
, Wolfgang Wadsak
2,4
, Marcus Hacker
2
,
Siegfried Kasper
1
, Rupert Lanzenberger
1
and Andreas Hahn
1
Abstract
Mapping the neuronal response during cognitive processing is of crucial importance to gain new insights into human
brain function. BOLD imaging and ASL are established MRI methods in this endeavor. Recently, the novel approach of
functional PET (fPET) was introduced, enabling absolute quantification of glucose metabolism at rest and during task
execution in a single measurement. Here, we report test-retest reliability of fPET in direct comparison to BOLD imaging
and ASL. Twenty healthy subjects underwent two PET/MRI measurements, providing estimates of glucose metabolism,
cerebral blood flow (CBF) and blood oxygenation. A cognitive task was employed with different levels of difficulty
requiring visual-motor coordination. Task-specific neuronal activation was robustly detected with all three imaging
approaches. The highest reliability was obtained for glucose metabolism at rest. Although this dropped during task
performance it was still comparable to that of CBF. In contrast, BOLD imaging yielded high performance only for
qualitative spatial overlap of task effects but not for quantitative comparison. Hence, the combined assessment of fPET
and ASL offers reliable and simultaneous absolute quantification of glucose metabolism and CBF at rest and task.
Keywords
fMRI, functional PET (fPET), PET/MRI, task-specific activation, test-retest reliability
Received 29 September 2020; Revised 22 April 2021; Accepted 3 May 2021
Introduction
Human brain function has been subject to research for
centuries and is yet not fully understood due to its vast
complexity. Thus, characterizing the neuronal response
during cognitive processing is of pivotal importance
and can be achieved with various approaches.
The most common imaging method to investigate
task-induced changes in the brain is functional magnet-
ic resonance imaging (fMRI) based on the blood
oxygen level-dependent (BOLD) signal. Advantages
are high sensitivity to changes in blood oxygenation,
readily accessible MRI sequences, established tasks and
high temporal and spatial resolution. However, the
BOLD signal is a composite of changes in cerebral
blood flow (CBF), cerebral blood volume (CBV) and
blood oxygenation
1
and consequently, only an indirect,
non-specific proxy for neuronal activation.
Other drawbacks are the instability of the BOLD
signal for longer task durations, spurious effects such
as scanner drifts
2
and heating,
3
field inhomogeneity,
3
heart rate and respiratory influences,
4
or draining
veins.
5
These are non-trivial issues and to a certain
1
Department of Psychiatry and Psychotherapy, Medical University of
Vienna, Vienna, Austria
2
Department of Biomedical Imaging and Image-guided Therapy, Division
of Nuclear Medicine, Medical University of Vienna, Vienna, Austria
3
Department of Pharmaceutical Sciences, Division of Pharmaceutical
Chemistry, University of Vienna, Vienna, Austria
4
Center for Biomarker Research in Medicine (CBmed), Graz, Austria
Corresponding author:
Andreas Hahn, Department of Psychiatry and Psychotherapy, Medical
University of Vienna, Waehringer Guertel 18-20, 1090 Vienna, Austria.
Email: andreas.hahn@meduniwien.ac.at
Journal of Cerebral Blood Flow &
Metabolism
0(0) 1–14
!The Author(s) 2021
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/0271678X211020589
journals.sagepub.com/home/jcbfm
extent reasons that absolute quantification of BOLD
signal changes require substantial effort.
6
Another fMRI-based approach is arterial spin label-
ing (ASL) which enables quantification of CBF. Here,
arterial blood is magnetically labeled to use it as endog-
enous contrast agent. Perfusion-weighted images are
then computed from a labeled and an unlabeled
image, enabling estimation of CBF.
7,8
This absolute
quantification represents a major advantage of ASL,
facilitating a comparison between rest and task-
specific flow changes. However, the requirement to
acquire images in pairs reduces the temporal resolution
and also the spatial resolution is inferior compared to
BOLD imaging.
9
Task-induced changes are also reflected in altered
glucose metabolism. Hence, positron emission tomog-
raphy (PET) using the glucose analogue [
18
F]FDG is a
suitable tool to map neuronal activation. Indeed, task-
specific changes in glucose metabolism were already
assessed before BOLD imaging and ASL were intro-
duced.
10
A crucial drawback of this method is that
measurements have to be conducted in at least two
separate sessions, mostly carried out even on separate
days, one at rest and another one during task execu-
tion. Changes in daily performance or resting activity
between these measurements will influence the results
11
and subjects are exposed to ionizing radiation twice.
Recently, these issues were resolved with the novel
approach of functional PET (fPET). The method ena-
bles the quantification of resting and task-specific glu-
cose metabolism within a single measurement.
12,13
The
protocol was further optimized by the administration
of [
18
F]FDG as bolus plus constant infusion.
14
This
increases the signal-to-noise ratio and the amount of
freely available radiotracer to track even subtle task-
related changes with a temporal resolution of minutes
(instead of hours or days as previously required).
Disadvantages include the necessity for arterial cannu-
lation to enable absolute quantification and the radia-
tion exposure of participants.
In sum, task-specific neuronal activation is reflected
in subtle changes from the resting activity in glucose
metabolism, blood flow and oxygenation. The above-
mentioned technical challenges and individual physio-
logical effects might limit the detection of these
changes. Hence, a valuable imaging approach is not
only sensitive to task-induced changes but also has to
provide high reliability during rest and task conditions.
This is of pivotal importance for scientific and clinical
applications to ensure that subtle effects can be robust-
ly detected despite the variance inherent to repeated
measurements. As such applications commonly aim
to assess complex cognitive functions, it is essential to
know the reliability for correspondingly complex tasks,
since reliability of simpler tasks may not be extrapolat-
ed adequately.
15
While ASL and BOLD imaging underwent optimi-
zation for decades already, fPET is still in its infancy
and the applicability of this novel imaging approach in
longitudinal studies was not yet assessed. Therefore, we
conducted a test-retest study and employed a challeng-
ing task with varying levels of cognitive load, given by
the video game TetrisV
R. The introduction of fully-
integrated PET/MRI scanners enabled the simulta-
neous acquisition of fPET, ASL and BOLD imaging.
Thus, the test-retest reliability between these imaging
modalities can be most directly compared in a single
scan session. We aimed to assess i) the capability of
fPET to track task-specific changes with a high reliabil-
ity between measurements and ii) its performance in
comparison to that of the well-established modalities
of ASL and BOLD imaging. For this purpose, we
investigated common parameters for each modality
(influx constant K
i
and cerebral metabolic rate of glu-
cose (CMRGlu) for fPET, CBF for ASL and parame-
ter estimates (i.e., beta values obtained from the general
linear model) for BOLD imaging) and assessed their
test-retest reliability with frequently used metrics
(intraclass correlation, coefficient of variation, Dice
coefficient).
Material and methods
Participants
For this study, 53 healthy subjects were initially
recruited, data from 40 were used and test-retest reli-
ability was assessed for 20. Among the 13 drop out
subjects, 7 discontinued voluntarily after the first mea-
surement, for 1 subject arterial blood sampling failed,
for 2 subjects ASL could not be acquired during the
first measurement because of temporary technical chal-
lenges with the PET/MRI scanner (only affecting
ASL), for 3 subjects only MRI was carried out due to
arterial puncture and/or radiotracer synthesis failure.
Cross-sectional data of a subsample was already
included in a previous analysis.
16
All subjects under-
went a routine medical investigation at a screening
visit including electrocardiography, blood tests, neuro-
logical and physiological tests, and a urine drug test.
Psychiatric disorders were ruled out with the Structural
Clinical Interview DSM-IV conducted by an experi-
enced psychiatrist. Female participants additionally
underwent a pregnancy test at the screening visit and
before each PET/MRI measurement. Participants had
to fast for at least four hours prior to the scan, includ-
ing no consumption of sweetened beverages and caf-
feine.
17
Exclusion criteria were weight above 100 kg for
reasons of radiation protection, current or previous
2Journal of Cerebral Blood Flow & Metabolism 0(0)
neurological, physiological or psychiatric disorders,
current breastfeeding or pregnancy, left-handedness,
substance abuse, MRI contraindications, participation
in a study including ionizing radiation exposure (past
10 years) and regularly playing TetrisV
Ror similar games
(including mobile phone games) within the last 3 years.
After detailed explanation of the study protocol, all
subjects gave written informed consent. All subjects
were insured and reimbursed for their participation.
The study was approved by the Ethics Committee of
the Medical University of Vienna (ethics number:
1479/2015) and all procedures were carried out in accor-
dance with the Declaration of Helsinki. The study was
registered at ClinicalTrials.gov (ID: NCT03485066).
Study design
Forty healthy subjects (20 male, mean age sd ¼
23.0 3.4 years, all right-handed) underwent a single
PET/MRI scan on a fully-integrated PET/MRI
system (Siemens Biograph mMR, Siemens
Healthineers, Germany). To assess the test-retest reli-
ability of functional imaging parameters, a subgroup of
20 subjects (10 male, 23.1 3.1 years) also underwent a
second measurement (4.2 0.7 weeks apart). The scans
of the remaining 20 subjects (10 male, 23.0 3.7 years)
solely served for an independent region of interest def-
inition. Measurements were initiated with the acquisi-
tion of a structural T1-weighted image followed by an
ASL sequence during rest. Subsequently, [
18
F]FDG
was administered as bolus followed by constant infu-
sion while a complex cognitive paradigm was
employed. After 8 minutes of rest, 4 task conditions
were carried out with varying difficulty (6 min each, 2
easy, 2 hard, pseudo-randomized order). Each task was
followed by a rest condition (5 min) where subjects
were instructed to look at a black crosshair on grey
background and to let their thoughts wander.
Simultaneously with the task blocks, ASL was acquired
during one easy and one hard condition. During the
remaining task blocks, BOLD imaging was acquired
for functional connectivity as described elsewhere.
16
Immediately after fPET and ASL acquisitions, the
same task was carried out in a conventional block
design and BOLD data was acquired (4 easy, 4 hard
and additionally, 4 control blocks, 30 s each, pseudo-
randomized order), again separated by rest blocks (10 s
each). Although the duration and number of task
blocks differed across imaging modalities, we aimed
to employ an optimal acquisition protocol for each
parameter. The entire study design is depicted in
Figure 1(a).
Cognitive task
An adapted version of the video game TetrisV
Rwas car-
ried out, representing a complex visuo-spatial motor
task, which combines mental rotation, spatial planning
and hand-eye coordination, thus, targeting various
higher-order brain regions. The task comprised two
levels of difficulty (easy and hard) to induce different
cognitive loads (Figure 1(b)). The aim of the task was
to complete full horizontal lines by moving and rotat-
ing bricks falling from the top of the screen. Subjects
played only with their right hand on an MR-
compatible button box (index/small finger: move left/
right, middle finger: rotate, ring finger: move down).
The two levels varied regarding the speed at which the
bricks were falling and the initial number of incomplete
lines at the beginning (Figure 1(b)). During BOLD
acquisition, a control level was introduced where
bricks had to be lead through a chimney sufficiently
Figure 1. Study design and cognitive task. (a) Measurements were initiated with a structural T1-weighted image (grey, 8 mins) and
ASL at rest (green, 6 mins). Thereafter, fPETwas acquired (blue, 52 mins) and an adapted version of the video game TetrisV
Rwas played
four times with varying cognitive load (6 mins, 2 easy, 2 hard, pseudo-randomized order) separated by resting periods. Simultaneously,
ASL was acquired during one easy and one hard condition (green). BOLD data for functional connectivity was acquired during the
second easy and hard task blocks, but these were not used in the current study. Immediately after fPET and ASL, BOLD data was
acquired with the same task and an additional control condition (red, 12 task blocks, 30 s each, 10 s rest). (b) The task consisted of the
conditions easy, hard and control, whereas the latter was only carried out during BOLD imaging. Easy and hard conditions differed by
the speed of the falling bricks and the initial number of incomplete lines. In the control condition bricks had to be guided through a
chimney and were automatically removed at the bottom. This figure was adapted from Hahn et al. (2020) under CC-BY license.
16
Rischka et al. 3
wide that no rotation was necessary (Figure 1(b)). For
the control level, bricks were removed at the bottom.
Thus, no completion of lines was possible. In general,
subjects were instructed to gain as many points as pos-
sible and were explained that completion of several
lines at once scores more points. Right before the
start of the measurement, each condition was played
once (30 s each) in the scanner to familiarize the par-
ticipants with the procedure and the controls.
PET and MRI data acquisition
The radioactive glucose analogue [
18
F]FDG was fresh-
ly synthesized every morning using FASTlab FDG cas-
settes with phosphate buffer formulation
18
and a
FASTlab platform (GE Healthcare). The substance
was administered via a cubital vein as bolus for
1 minute followed by constant infusion for 51 minutes
with an infusion pump (Syramed mSP6000, Arcomed,
Switzerland, dosage: 5.1 MBq/kg, bolus speed: 816 ml/
h, infusion speed: 42.8 ml/h, bolus-infusion ratio of
activity: 20:80%), which was placed in an MR-shield.
PET data was acquired in list-mode, enabling the ret-
rospective definition of frame lengths during
reconstruction.
A T1-weighted structural image was acquired with a
magnetization prepared rapid gradient echo
(MPRAGE) sequence prior to radiotracer administra-
tion (TE/TR ¼4.21/2200 ms, voxel size ¼11
1.1 mm, matrix size ¼240 256, slices ¼160, flip
angle ¼9,TI¼900 ms, 7.72 min). The image was
used to rule out severe brain disorders, for attenuation
correction and normalization to MNI space.
A 2 D pseudo-continuous arterial spin labeling
(pcASL) sequence was recorded at rest prior to radio-
tracer administration and during task conditions simul-
taneously with PET acquisition (TE/TR ¼12/4060 ms,
post-labeling delay ¼1800 ms, labeling dura-
tion ¼1.5 sec, readout duration per slice ¼35 ms,
voxel size ¼3.44 3.44 5mmþ1 mm gap, matrix
size ¼64 64, slices ¼20, flip angle ¼90, 6 min). The
labeling plane was 9 cm inferior of the center of the
field of view, which was placed on the anterior-
posterior commissure line. No background suppression
or further optimizations were applied.
BOLD data was acquired with an echo-planar
imaging (EPI) sequence following PET acquisition
(TE/TR ¼30/2000 ms, voxel size ¼2.5 2.5 2.5 mm
þ0.825 mm gap, matrix size ¼80 80, slices ¼34, flip
angle ¼90, 8.17 min).
Blood sampling
Prior to each PET/MRI measurement, the individual
fasting blood glucose level was measured (Glu
plasma
,
triplicate measurement). Arterial blood samples were
drawn from a radial artery throughout the radiotracer
administration (time points: 3, 4, 5, 14, 25, 36 and
47 min) and were timed not to interfere with task per-
formance and the MRI acquisition. Blood samples
were processed as previously described.
12
In short,
whole blood activity and plasma activity after centrifu-
gation were measured in a c-counter (Wizard
2
, 3”;
Perkin Elmer, USA). The whole blood curve was line-
arly interpolated and resampled to match the time
points of the reconstructed PET frames. The plasma-
to-whole-blood ratio was averaged across time points.
The whole blood curve was then multiplied with the
mean plasma-to-whole-blood ratio to obtain an arterial
input function for absolute quantification.
PET data preprocessing and quantification of glucose
metabolism
Data was reconstructed with an ordinary Poisson -
ordered subset expectation maximization algorithm
(OP-OSEM, 3 iterations, 21 subsets, matrix size:
344 344, slices: 127, voxel size: 2.09 2.09 2.03 mm)
and binned into 104 frames of 30 s. In addition to stan-
dard corrections such as dead time and decay, attenu-
ation and scatter correction was performed with a
pseudo-CT approach
19
based on the structural MRI
acquired at the first measurement. Preprocessing and
quantification steps were similar to our previous
reports:
14,16,20
SPM12 was used for head movement
correction (quality ¼1, registration to mean image),
spatial normalization to MNI space and spatial
smoothing with an 8 mm Gaussian kernel. The spatial
normalization was estimated with the structural MRI.
The mean PET image was then coregistered to the
structural MRI and both transformations (coregistra-
tion, normalization) were applied to the dynamic PET
data. Images were masked so that only grey matter
voxels were present and a low pass filter with a cutoff
frequency of half the task duration was applied to the
time course of every voxel. A general linear model
(GLM) was employed to separate task-specific and
baseline metabolism including four regressors: baseline,
one for each task condition (easy/hard, linear ramp
function, slope ¼1 kBq/frame) and the first principal
component of the six movement regressors estimated
during the movement correction step. The baseline
term was defined with a multimodal approach (as
well as an independent approach, see statistical analy-
sis). The individual BOLD data was used to identify
voxels in MNI space that exhibit significant task effects
(see below and Figure 2) in the hard vs rest condition
(p <0.05 FWE corrected voxel level). These voxels
were then masked out in the spatially normalized
PET frames. The remaining grey matter voxels were
4Journal of Cerebral Blood Flow & Metabolism 0(0)
considered inactive during the task and were averaged
within each frame, yielding a task-free baseline time
course. This approach has been proven useful in our
previous investigation
14
and yields similar task effects
compared to a BOLD-independent baseline defini-
tion.
16
The Gjedde-Patlak plot was applied to obtain
the influx constant K
i
with linearity set to 15 min after
tracer application resulting in 3 K
i
maps: rest, easy vs
rest and hard vs rest. Finally, CMRGlu was quantified
as CMRGlu ¼K
i
*Glu
plasma
/LC * 100, with LC being
the lumped constant ¼0.89.
ASL data preprocessing and cerebral blood flow
quantification
The 2 D pseudo-continuous arterial spin labeling data
was processed as described previously:
21
Voxels with
intensity below 80% of the mean of raw ASL data
(across all voxels within each frame) were set to 0 to
remove areas with insufficient signal. Movement cor-
rection was carried out with SPM12 (quality ¼1) fol-
lowed by calculation of the equilibrium magnetization
(M
0
) map as the temporal average of all unlabeled
images. The brain was extracted from the M
0
image
with the brain extraction tool implemented in FSL
22
and the resulting mask was applied to the images of
the time series. Data was spatially normalized to MNI
space via the T1-weighted structural image as done for
the PET data (i.e., mean image coregistered to T1, both
transformations applied to all ASL data) and
smoothed with an 8 mm Gaussian kernel. CBF was
calculated as
CBF ¼kDMR1a
2aM0fexp xR1a
ðÞ
exp½ sþw
ðÞ
R1ag (1)
where kis the blood-tissue water partition coefficient
(¼0.9 ml/g), DM the difference between pairs of labeled
and unlabeled images, R1a the longitudinal relaxation
rate of blood (¼0.67 sec
1
), athe tagging efficiency
(¼0.8), xthe post-labeling delay time adapted for
slice timing (¼1800 ms at slice 1) and sthe labeling
pulse duration (¼1.5 sec). CBF was averaged across
the time series.
Since the resulting maps acquired during task per-
formance represent the sum of baseline and task
effects, pure task-specific CBF was calculated by sub-
tracting CBF at rest from CBF obtained during the
easy and hard condition, respectively.
BOLD-derived task changes
Data preprocessing was carried out in SPM12 as
described previously:
14
BOLD data was slice timing
corrected to the middle slice, realigned to the mean
image (quality ¼1), spatially normalized to MNI
space and smoothed (8 mm Gaussian kernel). First
level analysis was performed as block design with one
regressor for each task condition (control, easy, hard)
in the GLM. Additionally, regressors for movement,
white matter and cerebrospinal fluid (CSF) were
included. The following contrasts of interest were esti-
mated from the GLM’s beta values: control vs rest,
easy vs rest, hard vs rest, easy vs control, hard vs
control.
Region of interest definition
In order to obtain an unbiased comparison between the
different imaging parameters, we defined functional
regions of interest (ROIs) based on all three imaging
modalities similar to our previous investigation
16
as
Figure 2. Task-specific changes and functional ROIs. Task effects were obtained with functional PET (fPET), arterial spin labeling
(ASL) and blood oxygenation (BOLD). The presented maps depict group t-maps of the contrasts hard vs rest for fPET and ASL (a and
b) and hard vs control for BOLD (c), all p
FWE
<0.05 corrected cluster level, height threshold of p <0.001 uncorrected voxel level. For
a robust analysis of the test-retest reliability, ROIs were determined with a conjunction analysis across all three modalities (inter-
section, d), revealing overlapping task changes in the frontal eye field (blue), intraparietal sulcus (pink), occipital cortex (green) and
supplementary motor area (SMA, red). We focused on large, robust clusters (>500 voxels) therefore, analysis of the SMA was
omitted. The slices z ¼4 and z ¼56 are presented in MNI space.
Rischka et al. 5
described below. To avoid potential bias, the ROI def-
inition and the test-retest evaluation was carried out
with separate study cohorts (see study design).
A group-wise one-sample t-test was performed for
each modality (fPET: K
i
, hard vs rest; ASL: CBF,
hard vs rest; BOLD: GLM beta values, hard vs control,
all p
FWE
<0.05 corrected cluster level, height threshold
of p <0.001 uncorrected voxel level). The BOLD con-
trast hard vs control was chosen for a similar extent of
task effects in the three modalities. Finally, a conjunc-
tion analysis (i.e., intersection) across the three FWE-
corrected and binarized t-maps was computed to
obtain task-specific ROIs. The left and corresponding
right side of each ROI were merged. We focused on the
largest clusters (>500 voxels) to provide a robust def-
inition of task-relevant changes (Figure 2). These
included the occipital cortex (OCC), intraparietal
sulcus (IPS) and the frontal eye field (FEF) (see
results), which were also identified in our previous
work with a partly overlapping sample.
16
The
mean value of each ROI was extracted for further
analyses.
Statistical analysis
Statistical analyses were based on commonly reported
parameters to enable comparison with previous litera-
ture, namely K
i
and CMRGlu for fPET, CBF for ASL
and parameter estimates for BOLD imaging. Similarly,
frequently used metrics of test-retest reliability were
used to compare imaging parameters between the two
measurements. Data were visually inspected for normal
distribution (Figure 4 and 5).
Quantitative comparisons were assessed using the
group-wise median within-subject coefficient of varia-
tion (CoV [%] ¼SD/mean*100) and the intraclass cor-
relation (ICC
3,1
, equation (2)) for each modality, each
ROI and each condition:
ICC3;1¼MSBS MSE
MSBS þk1
ðÞ
MSE (2)
where MSBS is the mean square between subjects and
MSE the mean square error. MSBS and MSE were
calculated from an n k matrix with n ¼20 observa-
tions and k ¼2 measurements.
Spatial similarity between the task effects of the first
and second measurement was assessed with the
Sørensen-Dice similarity coefficient (DICE) at group
level. Group level was chosen because no individual
statistical maps could be computed for ASL with the
current study design. Second-level analyses were per-
formed with a one-sample t-test (n ¼20) for each
measurement and each task condition for all three
modalities. T-maps were thresholded at p <0.05 FWE
corrected cluster level. DICE was calculated as
DICE ¼2jX\Yj
jXjþjYj(3)
where Xare all active voxels in the first measurement of
one modality and one condition and Yare all active
voxels in the corresponding second measurement.
Additional analyses were conducted to rule out spu-
rious findings and corresponding statistics were cor-
rected for multiple testing with the Bonferroni-Holm
procedure (multiple ROIs and conditions). The use of
BOLD data to identify task-specific voxels in the fPET
analysis may introduce dependencies between the two
modalities. Thus, we also computed task changes in
glucose metabolism independent from the BOLD
data by modeling the fPET baseline term with a
third-order polynomial
12
and then re-calculated reli-
ability parameters.
Next, we investigated potential effects of task per-
formance on test-retest reliability. Individual task per-
formance was computed as points per minute for each
condition following the original NintendoV
Rscoring
system for TetrisV
R. Differences in performance between
the two measurements were assessed by paired t-tests
and associations with imaging parameters were calcu-
lated by Pearson’s correlation.
Finally, the effects of head motion were assessed.
Average framewise displacement was computed from
the realignment parameters for each subject and imag-
ing modality.
16
Differences in framewise displacement
between the two measurements were assessed by paired
t-tests and associations with coefficients of variation
were calculated by Pearson’s correlation.
Results
The following paragraphs represent the key findings.
A comprehensive list of all results can be found in
Table 1. Figure 3 shows voxel-wise data from a
representative subject (i.e., with average coefficient of
variation). Furthermore, we provide scatter and Bland-
Altman plots to compare imaging parameters between
the two measurements at rest (Figure 4) and during
task performance (Figure 5). Reliability of CMRGlu
is presented in Supplementary Table 1.
Regions of interest
The conjunction analysis revealed task-induced
changes across all three imaging modalities in the
6Journal of Cerebral Blood Flow & Metabolism 0(0)
Table 1. Reliability of K
i
, CBF and BOLD changes.
CoV (IQR) [%] ICC
3,1
DICE
FEF IPS OCC FEF IPS OCC Whole brain
K
i
Rest 5.4 (4.2) 4.6 (2.6) 5.0 (4.0) 0.88 0.89 0.87 -
Easy vs Rest 26.1 (43.1) 27.9 (43.1) 29.9 (52.2) 0.50 0.52 0.33 0.53
Hard vs Rest 21.0 (34.8) 17.8 (16.2) 13.6 (21.3) 0.65 0.76 0.65 0.61
CBF
Rest 10.1 (10.4) 6.4 (10.5) 5.3 (7.8) 0.39 0.69 0.76 -
Easy vs Rest 38.7 (62.0) 32.4 (28.3) 24.9 (39.3) 0.51 0.43 0.67 0.68
Hard vs Rest 15.0 (31.8) 25.1 (43.2) 17.5 (24.5) 0.67 0.48 0.72 0.73
BOLD
Easy vs Rest 31.7 (31.2) 20.5 (37.3) 28.9 (49.8) 0.40 0.45 0.23 0.81
Hard vs Rest 16.4 (19.5) 21.1 (24.2) 24.5 (20.4) 0.41 0.13 0.09 0.78
Control vs Rest 22.3 (40.9) 38.9 (49.1) 24.5 (51.4) 0.72 0.44 0.13 0.78
Easy vs Control 61.4 (80.0) 43.4 (54.8) 40.2 (141.9) –0.06 0.46 0.55 0.35
Hard vs Control 34.0 (63.7) 33.8 (104.4) 49.4 (51.7) 0.12 –0.08 0.12 0.70
Commonly used metrics for test-retest reliability were estimated, namely median within-subject coefficient of variation (CoV), and intraclass cor-
relation (ICC) for each ROI and each condition. Additionally, whole-brain DICE coefficient was calculated. At resting state, K
i
exhibited the highest
reliability. During task performance, reliability dropped but was comparable between K
i
and CBF. With BOLD imaging the best performance was
achieved with the DICE coefficient, i.e. the qualitative spatial overlap of task effects between measurements.
FEF: frontal eye field, IPS: intraparietal sulcus, OCC: occipital cortex.
Figure 3. Visual comparison between measurements 1 and 2 (M1 and M2) for a representative (i.e. average test-retest reliability)
subject. (a) At rest, fPET and ASL demonstrate highly similar patterns of K
i
and CBF between the two measurements. (b) Task-specific
changes in K
i
, CBF and BOLD signal show clear activations for the regions of interest frontal eye field (FEF), intraparietal sulcus (IPS)
and occipital cortex (OCC) in both measurements. Similar slices as in Figure 2 are presented (z ¼8 and z ¼56, MNI space).
Rischka et al. 7
frontal eye field (FEF), intraparietal sulcus (IPS) and
the occipital cortex (OCC), depicted in Figure 2.
Additionally, the precentral gyrus and supplementary
motor area were identified but analyses were omitted
because of the limited size of the clusters.
Coefficient of variation
The smallest CoV were found for K
i
at rest with a
median of 4.6–5.4% for the different ROIs. Slightly
higher CoV were obtained for CBF at rest, namely
5.3–10.1%. For the two task conditions (easy/hard vs
rest, respectively) the CoV of K
i
increased, ranging
between 13.6 and 29.9%. Similarly, higher CoV was
observed for CBF during task performance (between
15.0% and 38.7%). The CoV of BOLD changes com-
pared to rest fluctuated between 16.4% and 38.9%.
Higher CoV were obtained for the contrast of task
conditions vs control (33.8–61.4%). CoV of CMRGlu
was similar to K
i
(rest: 7.0–7.9%, task: 17.4–29.3%).
Intraclass correlation
Similar to CoV, excellent reliability was obtained for K
i
at rest (ICC ¼0.87–0.89 for the different ROIs).
Contrarily, the ICC of CBF was lower during rest com-
pared to that of K
i
(0.39–0.76). Again, the reliability
decreased during task performance, with higher ICC
during the hard vs rest condition for K
i
(0.65–0.76)
and CBF (0.48–0.72). During the easy vs rest condition,
ICC was lower for K
i
(0.33–0.52), whereas ICC for
CBF was similar to hard (0.43–0.67). For BOLD
signal changes relative to rest, the ICC varied substan-
tially: easy (0.23–0.45), hard (0.09–0.41) and control
(0.13–0.72). Comparisons of task conditions vs control
even reached negative values (0.06–0.55). For
CMRGlu, ICC was slightly lower compared to K
i
ranging from 0.68 to 0.74 at rest and 0.31 to 0.61
during task performance.
Sørensen-Dice coefficient
Different from the other metrics, whole-brain similarity
of the task changes between the first and second mea-
surement was highest for BOLD changes for easy vs
rest (DICE ¼0.81) and hard vs rest (0.78). When con-
trasting against the control condition, DICE slightly
decreased to 0.70 for hard but dropped to 0.35 for
easy. For the other imaging modalities, DICE coeffi-
cients were lower than BOLD (K
i
: easy/hard: 0.53/0.61,
CMRGlu: easy/hard: 0.52/0.61 and CBF: 0.68/0.73).
Control analyses
We have carried out a number of additional analyses to
further support our findings. First, we calculated task-
Figure 4. Test-retest variability at rest. Scatter and Bland-Altman plots of glucose metabolism (K
i
) and cerebral blood flow (CBF)
comparing measurements 1 and 2 (M1 and M2). Both imaging modalities demonstrate good overall agreement between the two
measurements with few outliers and no proportional bias. No significant differences were observed between the measurements (all
p>0.5 corrected), indicating no relevant systematic bias. See Table 1 for complementary results. Colors for ROIs match those in
Figure 2. FEF: frontal eye field, IPS: intraparietal sulcus, OCC: occipital cortex.
8Journal of Cerebral Blood Flow & Metabolism 0(0)
specific changes in glucose metabolism independent of
the BOLD data. In line with our previous results this
showed similar activation patterns
14,16
and test-retest
reliability (Supplementary table 2). We further investi-
gated the potential effect of changes in task perfor-
mance. Even though the subjects did not train
between the two measurements there was a slight but
significant increase in performance for the easy (points
per minute 427 vs 710) and the hard task levels (1176 vs
1621, both p <0.01 corrected). However, there were no
significant changes in any of the corresponding imag-
ing parameters (all p >0.5 corrected) or any significant
correlations with changes in task performance (all
p>0.5 corrected). Finally, head motion was not differ-
ent between the two measurements for any of the imag-
ing modalities (all p >0.1 corrected). Furthermore, no
association between individual framewise displacement
and coefficients of variation was observed (all p >0.5
corrected), indicating that motion did not drive the
test-retest reliability.
Discussion
In this study, the test-retest reliability of functional
imaging parameters reflecting neuronal activation
during a complex visuo-spatial task was assessed. The
multimodal data acquisition approach facilitated a
direct comparison of fPET test-retest performance to
the well-established methods of ASL and BOLD imag-
ing. We observed excellent reliability of glucose metab-
olism and CBF at rest. The reliability decreased during
task performance for both K
i
and CBF but was lowest
for BOLD signal changes.
Measurements at resting state reflect resting activity,
which plays a key role in research and clinical applica-
tions, for example, to investigate altered metabolism or
blood flow in patients. A high test-retest reliability of
glucose metabolism at rest was already demonstrated
decades ago with 2-[1-
11
C]deoxyglucose measurements
within the same day
23
but also with [
18
F]FDG and
intervals of one to twelve weeks between measure-
ments.
24
Similar reliability was achieved in this study
with excellent ICC and small CoV. In a comprehensive
evaluation of CBF, test-retest reliability was compared
within session and 1 hour and 1 week apart.
25
As
expected, the CoV increased with the time between
measurements, which could be based on technical chal-
lenges and varying resting activity. Although our meas-
urements were separated by 4 weeks, we could achieve
similar CoV. With the caveat of longer measurement
time, K
i
exhibited a higher reliability than CBF in
terms of CoV and ICC, suggesting fPET as a robust
quantitative tool to map resting activity. Other than K
i
and CBF, estimation of BOLD changes during rest is
theoretically possible by calibrating the signal
6
but is
usually not performed due to technical challenges.
Consequently, a comparison to K
i
and CBF was omit-
ted in this work.
In addition to imaging parameters at rest, a compre-
hensive assessment of brain function requires analysis
of changes during stimulation and performance of
Figure 5. Task-specific test-retest variability. Scatter and Bland-
Altman plots of glucose metabolism (K
i
), cerebral blood flow
(CBF) and blood oxygen level dependent signal (BOLD, param-
eter estimates) comparing measurements 1 and 2 (M1 and M2).
As for the rest condition, the Bland-Altman plots indicate good
agreement without proportional bias or major outliers and
comparable limits of agreement between the easy (a) and the
hard task (b). No significant differences were observed between
the measurements (all p >0.5 corrected), indicating no relevant
systematic bias. See Table 1 for complementary results. Colors
for ROIs match those in Figure 2. FEF: frontal eye field, IPS:
intraparietal sulcus, OCC: occipital cortex.
Rischka et al. 9
specific tasks. Although a drop in reliability for glucose
metabolism and CBF in both task conditions was
observed, the rather novel fPET technique showed sim-
ilar task-specific reliability compared to the established
ASL approach, however, coming at the cost of longer
measurement time.
CMRGlu had slightly lower reliability than K
i
.By
definition, these inconsistencies can be ascribed to
changes in the LC or plasma glucose level between
measurements. Since LC is assumed to be stable in
healthy brains,
26
the differences are likely to occur
due to alterations in the plasma glucose levels. Thus,
acquiring a temporal profile of the latter might
improve the reliability. Regarding CBF, even lower
ICC values were previously reported for target regions,
although the measurement interval was only 1–2 days
27
compared to 4 weeks in our study. Also robust para-
digms such as finger tapping exhibited lower reliability
within a measurement interval of 1 to 4 weeks.
28
The
reliability of fMRI has been subject to long-lasting dis-
cussions as results vary widely. Recently, a meta-
analysis on task-specific BOLD changes, including
more than 50 studies, revealed a low test-retest reliabil-
ity (mean ICC ¼0.40) across various task durations,
test-retest intervals and task types.
15
A similar average
ICC was achieved with our task when compared to rest
(ICC ¼0.33), which was even lower for contrasts of
task vs control (ICC ¼0.17). For comparison, the aver-
age ICC of K
i
and CBF was markedly higher with
values of 0.52 and 0.58. Interestingly, the DICE coef-
ficient (i.e., a metric that quantifies the spatial extent of
overlapping activation, independent of its amplitude)
was highest for BOLD changes compared to CBF
and K
i
in almost all conditions. Together, ICC and
DICE indicate a high spatial overlap of task effects
for BOLD, but with fluctuating effect sizes. This sug-
gests to use BOLD more selectively, such as in research
questions were simple identification of active regions is
of greater importance than reliable quantitative
parameters.
In summary, the test-retest reliability during task
performance varied across the presented imaging
approaches. A possible explanation is that each modal-
ity reflects different factors that are coupled to neuro-
nal activation to varying degrees. While a close
relationship was observed between metabolic processes
and CBF, a mismatch occurs upon task performance
between oxygen supply and utilization.
29,30
During
task-specific neuronal activation, the glutamatergic
release is increased which triggers several neurovascu-
lar signal pathways, including vasodilating agents such
as nitric oxide or prostaglandins.
31
These factors yield a
higher CBF and blood oxygen level in capillaries and
arterioles. The net decrease in deoxygenated hemoglo-
bin is the underlying mechanism of the BOLD effect.
32
Hence, BOLD signal changes are mediated by gluta-
mate release and influenced by the abovementioned
factors. A more direct measure of metabolism is
given by the radioactive glucose analogue [
18
F]FDG.
Astrocytes metabolize glucose to supply neurons with
energy in form of lactate.
33
Also, in neurons them-
selves, glucose is transformed into ATP for action
potentials and synaptic transmission.
34
The radiotracer
is irreversibly bound in cells and thus, task-specific
increase in energy consumption is proportional to
[
18
F]FDG uptake.
35
Of note, also glucose metabolism
influences the vasodilation by certain neurovascular
signal pathways, including lactate.
36
This complex neu-
rophysiological interplay between the presented imag-
ing approaches suggests to consider them as
complementary and not as substitution of one another.
Another, more technical, reason for differences in
test-retest variability might arise from different mea-
surement durations between the three modalities.
Acquisition times for ASL and BOLD imaging were
shorter than for fPET. It is known that the temporal
signal-to-noise ratio (tSNR) increases with the square
root of the measurement time, which may also result in
higher reliability. Hence, extending the measurement
time of BOLD imaging and ASL to that of fPET
(52 min) would theoretically increase the tSNR by
approximately 150% and 70%, respectively (see
Supplementary figure S1). Of note, this does not
imply an increase in test-retest reliability by the same
amount, because external influences such as physiolog-
ical effects are not considered. Thus, it is likely that the
test-retest reliability of task-based fMRI reaches a pla-
teau similar to that of resting-state fMRI.
37
Currently, BOLD imaging is often the first choice to
map task changes on the whole-brain level, although
several pitfalls are known. As mentioned, the BOLD
effect is rather unspecific due to its signal complexity
which is often compensated for by employing a control
condition. However, finding a suitable control condi-
tion is challenging as it bears the risk to remove poten-
tial active regions after contrasting the conditions and
has the drawback of introducing additional variance as
also shown in this work.
Based on our results and aforementioned difficulties
with BOLD imaging, we propose to consider fPET and
ASL as a robust combined tool to tackle novel research
questions using fully-integrated PET/MRI systems.
This enables the simultaneous acquisition of two com-
plementary aspects of neuronal activation. Different to
BOLD imaging, fPET and ASL directly reflect glucose
metabolism and CBF, respectively, making them less
dependent on control conditions to achieve higher spe-
cificity. The advantage of absolute quantification fur-
ther enables longitudinal comparisons during task
performance and at rest. However, the acquisition
10 Journal of Cerebral Blood Flow & Metabolism 0(0)
requires an adaption of already established paradigms.
While fPET requires longer tasks, which have to be
stable for several minutes, BOLD imaging employs
shorter paradigms with several repetitions.
Furthermore, BOLD changes can resolve event-
related task effects due to the higher temporal resolu-
tion. Differently, fPET and ASL are currently limited
to block designs, although real-time and event-related
designs were proposed for ASL within a specific tech-
nical setup.
38,39
Given the current improvement of PET
scanners with higher sensitivity,
40
temporal resolutions
of up to 100 milliseconds and advanced reconstruction
algorithms,
41
event-related designs also seem to be
highly feasible with fPET in the near future.
In cases where the simultaneous acquisition of fPET
and BOLD is favorable, we suggest short task blocks
14
or a hierarchical design.
42
For the latter, a fast on/off
design is embedded in longer blocks, enabling a simul-
taneous acquisition of fPET and BOLD signal with
respect to the paradigm requirements. Utilizing a
dual-echo ASL sequence
43
would further allow the
acquisition of all three modalities at the same time
with such a design. Still, this would imply tradeoffs
mostly at the cost of the BOLD signal as ASL employs
longer TR. Although a hierarchical design enables a
direct comparison between the modalities, the short
rest periods within a block might influence the fPET
signal.
The assessment of task-specific glucose metabolism
may also prove feasible in clinical routine. The current
European Association of Nuclear Medicine (EANM)
procedure guidelines for [
18
F]FDG PET brain imaging
suggest the acquisition of a 15-30 min static image. This
may also include stimulation paradigms outside the
scanner to identify brain areas involved in specific
task performance e.g., before surgery.
17
However, this
omits metabolic dynamics and therefore potentially
misses important functional alterations connected to
a disease. With fPET, resting and task-specific glucose
metabolism can be robustly determined within a single
30 min measurement, enabling an optimized therapy
planning (e.g., in tumor patients), but also potentially
allowing for an enhanced monitoring of progression in
neurodegenerative diseases.
Although high test-retest reliability of fPET was
demonstrated, we have to note a few limitations.
fPET currently requires arterial cannulation for abso-
lute quantification of glucose metabolism but there is
great effort to substitute the AIF with an image-derived
input function which will broaden the applicability of
the approach.
28,44
Another drawback is the radiation
exposure of participants (here approx. 6.4 mSv for a
subject with 75 kg body weight). However, the develop-
ment of PET systems with high sensitivity
40,41
will
allow further reduction of the radiation burden up to
an order of magnitude.
45
Reliability obtained with a
complex visuo-spatial motor task with different task
loads might be more susceptible to individual daily per-
formance than simple tasks such as checkerboard stim-
ulation or finger tapping. However, mapping the
neuronal response to complex behavioral processes
requires performance of similarly challenging tasks,
which makes knowledge about the corresponding reli-
ability an essential aspect. Another limitation is the
slight improvement in task performance between the
two measurements. This may be ascribed to an
improved control to align the bricks, which may have
resulted in altered task-specific activation in the second
measurement. To diminish this effect, the measure-
ments were separated by 4 weeks and the ICC
3,1
was
used describing consistency rather than absolute agree-
ment. Furthermore, no changes in imaging parameters
or correlations with task performance were observed.
This indicates that the reported test-retest reliability
was not affected by the change in task performance.
Lastly, task-specific test-retest metrics were directly
compared between modalities, although data acquisi-
tion varied across modalities in duration and the
number of task blocks. Strictly speaking, only K
i
and
CBF were acquired simultaneously and BOLD changes
immediately afterwards within the same session. These
differences emerge from the different requirements of
imaging modalities. fPET and ASL perform best with
relatively long task blocks, whereas the BOLD signal is
unstable for long continuous task durations. To meet
these requirements, we employed a study design that is
technically most feasible but also allows to optimize the
task-specific acquisition independently for each imag-
ing modality. Our results therefore provide test-retest
data similar to commonly employed designs and it is
likely that longer measurement times would increase
the test-retest reliability of BOLD imaging and ASL
because of a higher tSNR. However, this has to be
evaluated systematically with the acquisition of more
task blocks and longer measurement durations to iden-
tify a plateau of reliability.
In conclusion, matching task-specific neuronal acti-
vations were robustly detected with fPET, ASL and
BOLD imaging in both task conditions. Despite its
recent introduction, similar or higher reliability was
achieved for fPET in comparison to the optimized
and well-established modalities of ASL and BOLD
imaging, albeit with a longer measurement time.
Indeed, BOLD changes exhibited a substantially
lower reliability, especially when using a control condi-
tion. Nevertheless, BOLD effects showed the highest
qualitative overlap between the measurements, suggest-
ing to use BOLD imaging preferably for the spatial
assessment of task changes. On the other hand, the
combination of fPET and ASL offers a robust tool,
Rischka et al. 11
enabling the simultaneous absolute quantification of
glucose metabolism and blood flow during rest and
task performance. Considering that fPET will benefit
from numerous further advancements such as progres-
sive modeling strategies, more sensitive scanners and
increased temporal resolution, an even higher reliabil-
ity of this technique is to be expected in the near future.
This enables the application of fPET not only in
research, but also in clinical routine.
Data availability statement
Raw data will not be publicly available due to reasons of data
protection. Processed data can be obtained from the corre-
sponding author with a data-sharing agreement.
Funding
The author(s) disclosed receipt of the following financial sup-
port for the research, authorship, and/or publication of this
article: This research was funded in whole, or in part, by the
Austrian Science Fund (FWF) KLI 610, PI: Andreas Hahn.
For the purpose of open access, the author has applied a CC
BY public copyright license to any Author Accepted
Manuscript version arising from this submission.
L. Rischka and M. Kl
obl are recipients of a DOC
Fellowship of the Austrian Academy of Sciences at the
Department of Psychiatry and Psychotherapy, Medical
University of Vienna. S. Klug is supported by the MDPhD
Excellence Program of the Medical University of Vienna. The
scientific project was performed with the support of the
Medical Imaging Cluster of the Medical University of
Vienna. The sponsors were not involved in any part of
the study.
Acknowledgements
The authors are especially grateful to Danny JJ Wang from
the Stevens Neuroimaging and Informatics Institute,
University of South Carolina and the Regents of the
University of California, for providing the pCASL sequence.
We also want to thank J. Raitanen, J. V
olkle and A.
Pomberger for radioligand synthesis and K. Papageorgiou,
A. Basaran, L. Silberbauer and the diploma students of the
Neuroimaging Labs (NIL) for medical support. The authors
are further grateful to K. Einenkel and E. Sittenberger for
subject recruitment and A. Jelicic for task implementation.
Declaration of conflicting interests
The author(s) declared the following potential conflicts of
interest with respect to the research, authorship, and/or pub-
lication of this article: M. Hacker received consulting fees
and/or honoraria from Bayer Healthcare BMS, Eli Lilly,
EZAG, GE Healthcare, Ipsen, ITM, Janssen, Roche, and
Siemens Healthineers. S. Kasper received grants/research
support, consulting fees and/or honoraria within the last
three years from Angelini, AOP Orphan Pharmaceuticals
AG, Celgene GmbH, Eli Lilly, Janssen-Cilag Pharma
GmbH, KRKA-Pharma, Lundbeck A/S, Mundipharma,
Neuraxpharm, Pfizer, Sage, Sanofi, Schwabe, Servier, Shire,
Sumitomo Dainippon Pharma Co. Ltd., Sun Pharmaceutical
Industries Ltd. and Takeda. R. Lanzenberger received travel
grants and/or conference speaker honoraria within the last
three years from Bruker BioSpin MR, Heel, and support
from Siemens Healthcare regarding clinical research using
PET/MRI. He is a shareholder of BM Health GmbH since
2019. W. Wadsak declares to having received speaker hono-
raria from the GE Healthcare and research grants from Ipsen
Pharma, Eckert-Ziegler AG, Scintomics, and ITG; and work-
ing as a part time employee of CBmed Ltd. (Center for
Biomarker Research in Medicine, Graz, Austria). All other
authors report no conflict of interest in relation to this study.
Authors’ contributions
LR, VP, WW, MH, SKa, RL and AH designed the study.
GMG, PM, VR recruited the subjects. VP and WW produced
the radiotracer. LR, AH, GMG, PM and SKl acquired the
data. LR and AH analyzed the data and were supported by
MK. MH, SKa, RL and AH supervised the study. LR and
AH wrote the manuscript. All authors reviewed and
approved the final version of the manuscript.
Supplementary material
Supplemental material for this article is available online.
ORCID iDs
Lucas Rischka https://orcid.org/0000-0002-6766-857X
Sebastian Klug https://orcid.org/0000-0001-8714-6608
Manfred Kl
obl https://orcid.org/0000-0003-2107-8803
Andreas Hahn https://orcid.org/0000-0001-9727-7580
References
1. Logothetis NK and Pfeuffer J. On the nature of the
BOLD fMRI contrast mechanism. Magn Reson Imaging
2004; 22: 1517–1531.
2. Smith AM, Lewis BK, Ruttimann UE, et al.
Investigation of low frequency drift in fMRI signal.
Neuroimage 1999; 9: 526–533.
3. Bollmann S, Kasper L, Vannesjo SJ, et al. Analysis and
correction of field fluctuations in fMRI data using field
monitoring. Neuroimage 2017; 154: 92–105.
4. Glover GH, Li TQ and Ress D. Image-based method for
retrospective correction of physiological motion effects in
fMRI: RETROICOR. Magn Reson Med 2000; 44:
162–167.
5. Turner R. How much cortex can a vein drain?
Downstream dilution of activation-related cerebral
blood oxygenation changes. Neuroimage 2002; 16:
1062–1067.
6. Pike GB. Quantitative functional MRI: concepts, issues
and future challenges. Neuroimage 2012; 62: 1234–1240.
7. Williams DS, Detre JA, Leigh JS, et al. Magnetic reso-
nance imaging of perfusion using spin inversion of arte-
rial water. Proc Natl Acad Sci U S A 1992; 89: 212–216.
8. Haller S, Zaharchuk G, Thomas DL, et al. Arterial spin
labeling perfusion of the brain: emerging clinical applica-
tions. Radiology 2016; 281: 337–356.
12 Journal of Cerebral Blood Flow & Metabolism 0(0)
9. Borogovac A and Asllani I. Arterial spin labeling (ASL)
fMRI: advantages, theoretical constrains and experimen-
tal challenges in neurosciences. Int J Biomed Imaging
2012; 2012: 1–13.
10. Phelps ME, Kuhl DE and Mazziota JC. Metabolic map-
ping of the brain’s response to visual stimulation: studies
in humans. Science 1981; 211: 1445–1448.
11. Schaefer SM, Abercrombie HC, Lindgren KA, et al.
Six-month test-retest reliability of MRI-defined PET
measures of regional cerebral glucose metabolic rate in
selected subcortical structures. Hum Brain Mapp 2000;
10: 1–9.
12. Hahn A, Gryglewski G, Nics L, et al. Quantification of
task-specific glucose metabolism with constant infusion
of 18F-FDG. J Nucl Med 2016; 57: 1933–1940.
13. Villien M, Wey H-Y, Mandeville JB, et al. Dynamic func-
tional imaging of brain glucose utilization using fPET-
FDG. Neuroimage 2014; 100: 192–199.
14. Rischka L, Gryglewski G, Pfaff S, et al. Reduced task
durations in functional PET imaging with [18F]FDG
approaching that of functional MRI. Neuroimage 2018;
181: 323–330.
15. Elliott ML, Knodt AR, Ireland D, et al. What is the test-
retest reliability of common task-functional MRI meas-
ures? New empirical evidence and a meta-analysis.
Psychol Sci 2020; 31: 792–806.
16. Hahn A, Breakspear M, Rischka L, et al. Reconfiguration
of functional brain networks and metabolic cost converge
during task performance. Elife 2020; 9: 1–18.
17. Varrone A, Asenbaum S, Vander Borght T, et al. EANM
procedure guidelines for PET brain imaging using [18F]
FDG, version 2. Eur J Nucl Med Mol Imaging 2009; 36:
2103–2110.
18. Long JZ, Jacobson MS and Hung JC. Comparison of
FASTlab 18F-FDG production using phosphate and cit-
rate buffer cassettes. J Nucl Med Technol 2013; 41: 32–34.
19. Burgos N, Cardoso MJ, Thielemans K, et al. Attenuation
correction synthesis for hybrid PET-MR scanners: appli-
cation to brain studies. IEEE Trans Med Imaging 2014;
33: 2332–2341.
20. Hahn A, Gryglewski G, Nics L, et al. Task-relevant brain
networks identified with simultaneous PET/MR imaging
of metabolism and connectivity. Brain Struct Funct 2018;
223: 1369–1378.
21. Wang J, Zhang Y, Wolf RL, et al. Amplitude-modulated
continuous arterial spin-labeling 3.0-T perfusion MR
imaging with a single coil: feasibility study. Radiology
2005; 235: 218–228.
22. Smith SM. Fast robust automated brain extraction. Hum
Brain Mapp 2002; 17: 143–155.
23. Bartlett EJ, Brodie JD, Wolf AP, et al. Reproducibility of
cerebral glucose metabolic measurements in resting human
subjects. J Cereb Blood Flow Metab 1988; 8: 502–512.
24. Maquet P, Dive D, Salmon E, et al. Reproducibility of
cerebral glucose utilization measured by PET and the
[18F]-2-fluoro-2-deoxy-d-glucose method in resting, healthy
human subjects. Eur J Nucl Med 1990; 16: 267–273.
25. Chen Y, Wang DJJ and Detre JA. Test-retest reliability
of arterial spin labeling with common labeling strategies.
J Magn Reson Imaging 2011; 33: 940–949.
26. Barrio JR, Huang S-C, Satyamurthy N, et al. Does 2-
FDG PET accurately reflect quantitative in vivo glucose
utilization? J Nucl Med 2020; 61: 931–937.
27. Yang FN, Xu S, Spaeth A, et al. Test-retest reliability of
cerebral blood flow for assessing brain function at rest
and during a vigilance task. Neuroimage 2019; 193:
157–166.
28. Sari H, Erlandsson K, Law I, et al. Estimation of an
image derived input function with MR-defined carotid
arteries in FDG-PET human studies using a novel partial
volume correction method. J Cereb Blood Flow Metab
2017; 37: 1398–1409.
29. Gusnard DA, Raichle ME and Raichle ME. Searching
for a baseline: functional imaging and the resting human
brain. Nat Rev Neurosci 2001; 2: 685–694.
30. Fox PT, Raichle ME, Mintun MA, et al. Nonoxidative
glucose consumption during focal physiologic neural
activity. Science 1988; 241: 462–464.
31. Attwell D, Buchan AM, Charpak S, et al. Glial and neu-
ronal control of brain blood flow. Nature 2010; 468:
232–243.
32. Attwell D and Iadecola C. The neural basis of functional
brain imaging signals. Trends Neurosci 2002; 25: 621–625.
33. Pellerin L and Magistretti PJ. Glutamate uptake into
astrocytes stimulates aerobic glycolysis: a mechanism
coupling neuronal activity to glucose utilization. Proc
Natl Acad Sci U S A 1994; 91: 10625–10629.
34. Harris JJ, Jolivet R and Attwell D. Synaptic energy use
and supply. Neuron 2012; 75: 762–777.
35. Zimmer ER, Parent MJ, Souza DG, et al. [18F]FDG
PET signal is driven by astroglial glutamate transport.
Nat Neurosci 2017; 20: 393–395.
36. Iadecola C. The neurovascular unit coming of age: a
journey through neurovascular coupling in health and
disease. Neuron 2017; 96: 17–42.
37. Birn RM, Molloy EK, Patriat R, et al. The effect of scan
length on the reliability of resting-state fMRI connectiv-
ity estimates. Neuroimage 2013; 83: 550–558.
38. Hernandez-Garcia L, Jahanian H, Greenwald MK, et al.
Real-time functional MRI using pseudo-continuous arte-
rial spin labeling. Magn Reson Med 2011; 65: 1570–1577.
39. Hernandez-Garcia L, Lee GR, Vazquez AL, et al. Fast,
pseudo-continuous arterial spin labeling for functional
imaging using a two-coil system. Magn Reson Med
2004; 51: 577–585.
40. van Sluis J, de Jong J, Schaar J, et al. Performance char-
acteristics of the digital biograph vision PET/CT system.
J Nucl Med 2019; 60: 1031–1036.
41. Zhang X, Cherry SR, Xie Z, et al. Subsecond total-
body imaging using ultrasensitive positron emission
tomography. Proc Natl Acad Sci U S A 2020; 117:
2265–2267.
42. Jamadar SD, Ward PG, Li S, et al. Simultaneous task-
based BOLD-fMRI and [18-F] FDG functional PET for
measurement of neuronal metabolism in the human
visual cortex. Neuroimage 2019; 189: 258–266.
43. Ghariq E, Chappell MA, Schmid S, et al. Effects of back-
ground suppression on the sensitivity of dual-echo arte-
rial spin labeling MRI for BOLD and CBF signal
changes. Neuroimage 2014; 103: 316–322.
Rischka et al. 13
44. Sundar LK, Muzik O, Rischka L, et al. Towards quan-
titative [18F]FDG-PET/MRI of the brain: automated
MR-driven calculation of an image-derived input func-
tion for the non-invasive determination of cerebral glu-
cose metabolic rates. J Cereb Blood Flow Metab 2019; 39:
1516–1530.
45. Cherry SR, Jones T, Karp JS, et al. Total-Body PET:
maximizing sensitivity to create new opportunities for
clinical research and patient care. J Nucl Med 2018;
59: 3–12.
14 Journal of Cerebral Blood Flow & Metabolism 0(0)
... This technique has demonstrated robust increases in glucose metabolism across various cognitive tasks [8][9][10][11], with high test-retest reliability [12]. Furthermore, the approach has been successfully extended to image dynamics of neurotransmitter synthesis during task performance, such as the dopamine [13] and serotonin system [14] with 6-[ 18 F]FDOPA and [ 11 C]AMT, respectively. ...
... The current work consolidates a decade of experience with this technique [5,[7][8][9][10][11][12][13][14][15][16][17][18][19][20][21] small-animal PET insert for 7T MRI ClinScan) and species (humans, non-human primates, rodents). The toolbox provides scripting for efficient batch processing and includes a userfriendly graphical user interface (GUI). ...
... ; https://doi.org/10.1101/2024.11.13.623377 doi: bioRxiv preprint anatomical regions, etc.). Additionally, a third order polynomial function can be used to model the baseline [5], yielding similar stimulation-induced effects [8,16] and test-retest reliability [12], but less optimal fitting [8]. ...
Preprint
Full-text available
Purpose Functional PET (fPET) enables the identification of stimulation-specific changes of various physiological processes (e.g., glucose metabolism, neurotransmitter synthesis) as well as computation of individual molecular connectivity and group-level molecular covariance. However, currently no consistent analysis approach is available for these techniques. We present a versatile, freely available toolbox designed for the analysis of fPET data, thereby filling a gap in the assessment of neuroimaging data. Methods The fPET toolbox supports analyses for a variety of radiotracers, scanners, experimental protocols, cognitive tasks and species. It includes general linear model (GLM)-based assessment of task-specific effects, percent signal change and absolute quantification, as well as independent component analysis (ICA) for data-driven analyses. Furthermore, it allows computation of molecular connectivity via temporal correlations of PET signals between regions and molecular covariance as between-subject covariance using static images. Results Toolbox performance was validated by analysis protocols established in previous work. Stimulation-induced changes in [ ¹⁸ F]FDG metabolic demands and neurotransmitter dynamics obtained with 6-[ ¹⁸ F]FDOPA and [ ¹¹ C]AMT were robustly detected across different cognitive tasks. Molecular connectivity analysis demonstrated metabolic interactions between different networks, whereas group-level covariance analysis highlighted interhemispheric relationships. These results underscore the flexibility of fPET in capturing dynamic molecular processes. Conclusions The toolbox offers a comprehensive, unified and user-friendly platform for analyzing fPET data across a variety of experimental settings. It provides a reproducible analysis approach, which in turn facilitates sharing of analyses pipelines and comparison across centers to advance the study of brain metabolism and neurotransmitter dynamics in health and disease.
... Generally, with oxygen and glucose being the primary fuels, understanding their dynamics is crucial for further insights into neuroenergetics. Therefore, in the present study, we integrated recent advances in functional neuroimaging, simultaneously acquiring multiparametric quantitative BOLD (mqBOLD) (Christen et al., 2012;Hirsch et al., 2014;Kaczmarz et al., 2020) and functional 18 F-FDG-PET (fPET) ) (Hahn et al., 2016;Rischka et al., 2018Rischka et al., , 2021Villien et al., 2014) data. This allowed us to, for the first time, measure CMRO2 and CMRglc at the same time and under different conditions in one scanning session. ...
... Additionally, inter-session variations can cause inaccuracies in result interpretation. More recently, researchers have developed functional FDG-PET (fPET) (Hahn et al., 2016;Rischka et al., 2018Rischka et al., , 2021Villien et al., 2014), involving a bolus administration followed by a continuous infusion of the radiotracer instead of a single bolus injection The constant infusion maintains a steady-state level of tracer in the bloodstream, allowing for dynamic imaging of glucose metabolism over time. In this way, fPET enables the measurement of multiple conditions within a single scanning session, as successfully applied in previous studies (Hahn et al., 2017(Hahn et al., , 2017Jamadar et al., 2019Jamadar et al., , 2021Rischka et al., 2021). ...
... More recently, researchers have developed functional FDG-PET (fPET) (Hahn et al., 2016;Rischka et al., 2018Rischka et al., , 2021Villien et al., 2014), involving a bolus administration followed by a continuous infusion of the radiotracer instead of a single bolus injection The constant infusion maintains a steady-state level of tracer in the bloodstream, allowing for dynamic imaging of glucose metabolism over time. In this way, fPET enables the measurement of multiple conditions within a single scanning session, as successfully applied in previous studies (Hahn et al., 2017(Hahn et al., , 2017Jamadar et al., 2019Jamadar et al., , 2021Rischka et al., 2021). ...
Preprint
Full-text available
The brain relies on oxidized glucose as its primary fuel. Despite robust coupling of cerebral oxygen and glucose consumption during rest, the oxygen to glucose index (OGI) has been suggested to drop significantly during neuronal activation. However, empirical evidence regarding the extent of this uncoupling is scarce, mainly due to the inability of previous studies to measure CMR O2 and CMR glc concurrently during tasks. Therefore, in the present study, we integrated multiparametric quantitative BOLD (mqBOLD) with functional PET (fPET) to simultaneously quantify cerebral oxygen and glucose metabolism during visual stimulation and rest within a single session. Results show increases in both CMR O2 and CMR glc in visual areas, concomitant with focal blood flow increases. Moreover, OGI values during rest were close to the theoretical value of 6, which is in line with previous literature. In response to visual stimulation, the OGI decreased by 6.6-21.6%, depending on the mask applied. For the first time, the present study demonstrates the feasibility of combining mqBOLD and fPET to study CMR O2 and CMR glc simultaneously. This setup has the potential to be applied to various experimental settings, providing valuable information about the extent of oxidative glucose metabolism in the human brain under different conditions in health and disease.
... For this retrospective analysis, data from our previous test-retest study was used 22 . Thus, full details about the study design and data acquisition can also found there and in related work 1,11 . ...
... To familiarize themselves with the task and the controls, participants completed a 30 s training of each condition before each scanning session. A more detailed description of the task can be found in 22,24 . ...
... Finally, the cerebral metabolic rate of glucose (CMRGlu) was quantified using the lumped constant of 0.89 26 . For a more detailed description please see 22 . ...
Preprint
Full-text available
Functional Positron Emission Tomography (fPET) has advanced as an effective tool for investigating dynamic processes in glucose metabolism and neurotransmitter action, offering potential insights into brain function, disease progression, and treatment development. Despite significant methodological advances, extracting stimulation-specific information presents additional challenges in optimizing signal processing across both spatial and temporal domains, which are essential for obtaining clinically relevant insights. This study aims to provide a systematic evaluation of state-of-the-art filtering techniques for fPET imaging. Forty healthy participants underwent a single [ ¹⁸ F]FDG PET/MR scan, engaging in the cognitive task Tetris®. Twenty thereof also underwent a second PET/MR session. Eight filtering techniques, including 3D and 4D Gaussian smoothing, highly constrained backprojection (hypr), iterative hypr (Ihypr4D), two MRI-Markov Random Field (MRI-MRF) filters (L=10 and 14 mm neighborhood) as well as static and dynamic Non-Local Means (sNLM and dNLM respectively) approaches, were applied to fPET data. Test-retest reliability (intraclass correlation coefficient), the identifiability of the task signal (temporal signal-to-noise ratio (tSNR)), spatial task-based activation (group level t-values), and sample size calculations were assessed. Results indicate distinct performance between filtering techniques. Compared to standard 3D Gaussian smoothing, dNLM, sNLM, MRI-MRF L=10 and Ihypr4D filters exhibited superior tSNR, while only dNLM and hypr showed improved test-retest reliability. Spatial task-based activation was enhanced by both NLM filters and MRI-MRF approaches. The dNLM enabled a minimum reduction of 15.4% in required sample size. The study systematically evaluated filtering techniques in fPET data processing, highlighting their strengths and limitations. The dNLM filter emerges as a promising choice, with improved performance across all metrics. However, filter selection should align with specific study objectives, considering factors like processing time and resource constraints.
... DS1 includes 52 healthy participants (23.2 ± 3.3 years, 24 females, all right-handed), who were partly also included in previous work [13,[15][16][17]. DS2 comprises 18 healthy participants' data (24.2 ± 4.3 years, 8 females, all righthanded), of which 15 had previously contributed to another study [3]. ...
... Another significant distinction lies in the test-retest variability of the methods. Previous work has indicated higher reliability for fPET than for fMRI [15,27]. This variability implies that robust fMRI disease markers are difficult to establish [28], which contributes to the rare use in clinical routine. ...
Article
Full-text available
Purpose Functional positron emission tomography (fPET) with [¹⁸F]FDG allows quantification of stimulation-induced changes in glucose metabolism independent of neurovascular coupling. However, the gold standard for quantification requires invasive arterial blood sampling, limiting its widespread use. Here, we introduce a novel fPET method without the need for an input function. Methods We validated the approach using two datasets (DS). For DS1, 52 volunteers (23.2 ± 3.3 years, 24 females) performed Tetris® during a [¹⁸F]FDG fPET scan (bolus + constant infusion). For DS2, 18 participants (24.2 ± 4.3 years, 8 females) performed an eyes-open/finger tapping task (constant infusion). Task-specific changes in metabolism were assessed with the general linear model (GLM) and cerebral metabolic rate of glucose (CMRGlu) was quantified with the Patlak plot as reference. We then estimated simplified outcome parameters, including GLM beta values and percent signal change (%SC), and compared them, region and whole-brain-wise. Results We observed higher agreement with the reference for DS1 than DS2. Both DS resulted in strong correlations between regional task-specific beta estimates and CMRGlu (r = 0.763…0.912). %SC of beta values exhibited strong agreement with %SC of CMRGlu (r = 0.909…0.999). Average activation maps showed a high spatial similarity between CMRGlu and beta estimates (Dice = 0.870…0.979) as well as %SC (Dice = 0.932…0.997), respectively. Conclusion The non-invasive method reliably estimates task-specific changes in glucose metabolism without blood sampling. This streamlines fPET, albeit with the trade-off of being unable to quantify baseline metabolism. The simplification enhances its applicability in research and clinical settings.
... In order to fully utilize a real-time motion-tolerant PET imager, one must take advantage of ligand delivery methods that enable multiple task/ baseline periods such as multiple bolus 28 and bolus/infusion (bolus followed by steady-state infusion) techniques 29-32, . Bolus-infusion F 18 -FDG studies in humans using traditional PET scanners have demonstrated task appropriate activity for visual and auditory task paradigms, with a same-session baseline control, and with ON/OFF task period cycles as short as one to two minutes [30][31][32][33] . In terms of physical device changes, the massive heavy detectors that limited PETs motion tolerance can be replaced with lightweight detectors due to advances in detector materials, such as solid-state Silicon Photomultiplier (SiPM) technology 34,35 , which have been used in headdedicated brain scanners 36,37 , although these so far are fixed and do not tolerate head motion 38 . ...
Article
Full-text available
Background Mobile upright PET devices have the potential to enable previously impossible neuroimaging studies. Currently available options are imagers with deep brain coverage that severely limit head/body movements or imagers with upright/motion enabling properties that are limited to only covering the brain surface. Methods In this study, we test the feasibility of an upright, motion-compatible brain imager, our Ambulatory Motion-enabling Positron Emission Tomography (AMPET) helmet prototype, for use as a neuroscience tool by replicating a variant of a published PET/fMRI study of the neurocorrelates of human walking. We validate our AMPET prototype by conducting a walking movement paradigm to determine motion tolerance and assess for appropriate task related activity in motor-related brain regions. Human participants (n = 11 patients) performed a walking-in-place task with simultaneous AMPET imaging, receiving a bolus delivery of F¹⁸-Fluorodeoxyglucose. Results Here we validate three pre-determined measure criteria, including brain alignment motion artifact of less than <2 mm and functional neuroimaging outcomes consistent with existing walking movement literature. Conclusions The study extends the potential and utility for use of mobile, upright, and motion-tolerant neuroimaging devices in real-world, ecologically-valid paradigms. Our approach accounts for the real-world logistics of an actual human participant study and can be used to inform experimental physicists, engineers and imaging instrumentation developers undertaking similar future studies. The technical advances described herein help set new priorities for facilitating future neuroimaging devices and research of the human brain in health and disease.
... For instance, conventional magnetic resonance imaging (MRI) can exclude other brain pathologies such as brain tumors or strokes that may cause similar symptoms. Techniques like functional MRI (fMRI) (Shi et al., 2022(Shi et al., , 2023 and positron emission tomography (PET) can detect patterns of activity and metabolic changes in the brain, providing real-time information about the functioning of different brain regions (Rischka et al., 2021;Malén et al., 2022). This helps to understand the abnormal patterns of brain activity in PD patients, particularly during the execution of motor and cognitive tasks (Haq et al., 2020). ...
Article
Full-text available
Objective The purpose of this study is to explore whether machine learning can be used to establish an effective model for the diagnosis of Parkinson's disease (PD) by using texture features extracted from cerebellar gray matter and white matter, so as to identify subtle changes that cannot be observed by the naked eye. Method This study involved a data collection period from June 2010 to March 2023, including 374 subjects from two cohorts. The Parkinson's Progression Markers Initiative (PPMI) served as the training set, with control group and PD patients (HC: 102 and PD: 102) from 24 global sites. Our institution's data was utilized as the test set (HC: 91 and PD: 79). Machine learning was employed to establish multiple models for PD diagnosis based on texture features of the cerebellum's gray and white matter. Results underwent evaluation through 5-fold cross-validation analysis, calculating the area under the receiver operating characteristic curve (AUC) for each model. The performance of each model was compared using the Delong test, and the interpretability of the optimized model was further augmented by employing Shapley additive explanations (SHAP). Results The AUCs for all pipelines in the validation dataset were compared using FeAture Explorer (FAE) software. Among the models established by Kruskal-Wallis (KW) and logistic regression via Lasso (LRLasso), the AUC was highest using the “one-standard error” rule. 'WM_original_glrlm_GrayLevelNonUniformity' was considered the most stable and predictive feature. Conclusion The texture features of cerebellar gray matter and white matter combined with machine learning may have potential value in the diagnosis of Parkinson's disease, in which the heterogeneity of white matter may be a more valuable imaging marker.
... Although blood oxygenation level-dependent (BOLD) fMRI has been predominantly used to study brain activation, arterial spin labeling (ASL) fMRI has been demonstrated to be capable of detecting brain activation with great sensitivity [37]. Recent technical development of dynamic ASL fMRI has been implemented with heavy background suppression, which has shown superior performance in detecting resting state networks in healthy volunteers [38] and changes in resting state networks from meditation training [28]; however, the background suppressed ASL has not been compared with BOLD for its performance in investigating the FAM effects on brain activation while performing an attentional task. ...
Article
Full-text available
Focused attention meditation (FAM) training has been shown to improve attention, but the neural basis of FAM on attention has not been thoroughly understood. Here, we aim to investigate the neural effect of a 2-month FAM training on novice meditators in a visual oddball task (a frequently adopted task to evaluate attention), evaluated with both ASL and BOLD fMRI. Using ASL, activation was increased in the middle cingulate (part of the salience network, SN) and temporoparietal (part of the frontoparietal network, FPN) regions; the FAM practice time was negatively associated with the longitudinal changes in activation in the medial prefrontal (part of the default mode network, DMN) and middle frontal (part of the FPN) regions. Using BOLD, the FAM practice time was positively associated with the longitudinal changes of activation in the inferior parietal (part of the dorsal attention network, DAN), dorsolateral prefrontal (part of the FPN), and precentral (part of the DAN) regions. The effect sizes for the activation changes and their association with practice time using ASL are significantly larger than those using BOLD. Our study suggests that FAM training may improve attention via modulation of the DMN, DAN, SN, and FPN, and ASL may be a sensitive tool to study the FAM effect on attention.
Article
The dopaminergic system is a central component of the brain’s neurobiological framework, governing motor control and reward responses and playing an essential role in various brain disorders. Within this complex network, the nigrostriatal pathway represents a critical circuit for dopamine neurotransmission from the substantia nigra to the striatum. However, stand-alone functional magnetic resonance imaging is unable to study the intricate interplay between brain activation and its molecular underpinnings. In our study, the use of a functional [fluorine-18]2-fluor-2-deoxy- d -glucose positron emission tomography approach, simultaneously with blood oxygen level–dependent functional magnetic resonance imaging, provided an important insight that demonstrates an active suppression of the nigrostriatal activity during optogenetic stimulation. This result increases our understanding of the molecular mechanisms of brain function and provides an important perspective on how dopamine influences hemodynamic responses in the brain.
Article
Full-text available
Identifying brain biomarkers of disease risk is a growing priority in neuroscience. The ability to identify meaningful biomarkers is limited by measurement reliability; unreliable measures are unsuitable for predicting clinical outcomes. Measuring brain activity using task functional MRI (fMRI) is a major focus of biomarker development; however, the reliability of task fMRI has not been systematically evaluated. We present converging evidence demonstrating poor reliability of task-fMRI measures. First, a meta-analysis of 90 experiments ( N = 1,008) revealed poor overall reliability—mean intraclass correlation coefficient (ICC) = .397. Second, the test-retest reliabilities of activity in a priori regions of interest across 11 common fMRI tasks collected by the Human Connectome Project ( N = 45) and the Dunedin Study ( N = 20) were poor (ICCs = .067–.485). Collectively, these findings demonstrate that common task-fMRI measures are not currently suitable for brain biomarker discovery or for individual-differences research. We review how this state of affairs came to be and highlight avenues for improving task-fMRI reliability.
Article
Full-text available
The ability to solve cognitive tasks depends upon adaptive changes in the organization of whole-brain functional networks. However, the link between task-induced network reconfigurations and their underlying energy demands is poorly understood. We address this by multimodal network analyses integrating functional and molecular neuroimaging acquired concurrently during a complex cognitive task. Task engagement elicited a marked increase in the association between glucose consumption and functional brain network reorganization. This convergence between metabolic and neural processes was specific to feedforward connections linking the visual and dorsal attention networks, in accordance with task requirements of visuo-spatial reasoning. Further increases in cognitive load above initial task engagement did not affect the relationship between metabolism and network reorganization but only modulated existing interactions. Our findings show how the upregulation of key computational mechanisms to support cognitive performance unveils the complex, interdependent changes in neural metabolism and neuro-vascular responses.
Article
Full-text available
A 194-cm-long total-body positron emission tomography/computed tomography (PET/CT) scanner (uEXPLORER), has been constructed to offer a transformative platform for human radiotracer imaging in clinical research and healthcare. Its total-body coverage and exceptional sensitivity provide opportunities for innovative studies of physiology, biochemistry, and pharmacology. The objective of this study is to develop a method to perform ultrahigh (100 ms) temporal resolution dynamic PET imaging by combining advanced dynamic image reconstruction paradigms with the uEXPLORER scanner. We aim to capture the fast dynamics of initial radiotracer distribution, as well as cardiac motion, in the human body. The results show that we can visualize radiotracer transport in the body on timescales of 100 ms and obtain motion-frozen images with superior image quality compared to conventional methods. The proposed method has applications in studying fast tracer dynamics, such as blood flow and the dynamic response to neural modulation, as well as performing real-time motion tracking (e.g., cardiac and respiratory motion, and gross body motion) without any external monitoring device (e.g., electrocardiogram, breathing belt, or optical trackers).
Article
Full-text available
Studies of task-evoked brain activity are the cornerstone of cognitive neuroscience, and unravel the spatial and temporal brain dynamics of cognition in health and disease. Blood oxygenation level dependent functional magnetic resonance imaging (BOLD-fMRI) is one of the most common methods of studying brain function in humans. BOLD-fMRI indirectly infers neuronal activity from regional changes in blood oxygenation and is not a quantitative metric of brain function. Regional variation in glucose metabolism, measured using [18-F] fluorodeoxyglucose positron emission tomography (FDG-PET), provides a more direct and interpretable measure of neuronal activity. However, while the temporal resolution of BOLD-fMRI is in the order of seconds, standard FDG-PET protocols provide a static snapshot of glucose metabolism. Here, we develop a novel experimental design for measurement of task-evoked changes in regional blood oxygenation and glucose metabolism with high temporal resolution. Over a 90-min simultaneous BOLD-fMRI/FDG-PET scan, [18F] FDG was constantly infused to 10 healthy volunteers, who viewed a flickering checkerboard presented in a hierarchical block design. Dynamic task-related changes in blood oxygenation and glucose metabolism were examined with temporal resolution of 2.5sec and 1-min, respectively. Task-related, temporally coherent brain networks of haemodynamic and metabolic connectivity were jointly coupled in the visual cortex, as expected. Results demonstrate that the hierarchical block design, together with the infusion FDG-PET technique, enabled both modalities to track task-related neural responses with high temporal resolution. The simultaneous MR-PET approach has the potential to provide unique insights into the dynamic haemodynamic and metabolic interactions that underlie cognition in health and disease.
Article
Full-text available
Absolute quantification of PET brain imaging requires the measurement of an arterial input function (AIF), typically obtained invasively via an arterial cannulation. We present an approach to automatically calculate an image-derived input function (IDIF) and cerebral metabolic rates of glucose (CMRGlc) from the [18F]FDG PET data using an integrated PET/MRI system. Ten healthy controls underwent test–retest dynamic [18F]FDG-PET/MRI examinations. The imaging protocol consisted of a 60-min PET list-mode acquisition together with a time-of-flight MR angiography scan for segmenting the carotid arteries and intermittent MR navigators to monitor subject movement. AIFs were collected as the reference standard. Attenuation correction was performed using a separate low-dose CT scan. Assessment of the percentage difference between area-under-the-curve of IDIF and AIF yielded values within ±5%. Similar test–retest variability was seen between AIFs (9 ± 8) % and the IDIFs (9 ± 7) %. Absolute percentage difference between CMRGlc values obtained from AIF and IDIF across all examinations and selected brain regions was 3.2% (interquartile range: (2.4–4.3) %, maximum < 10%). High test–retest intravariability was observed between CMRGlc values obtained from AIF (14%) and IDIF (17%). The proposed approach provides an IDIF, which can be effectively used in lieu of AIF.
Article
2-Deoxy-2-[18F]fluoro-D-glucose (2-FDG) with positron emission tomography (2-FDG-PET) is undeniably useful in the clinic, among other uses, to monitor change over time using the 2-FDG standardized uptake values (SUV) metric. This report suggests some potentially serious caveats for this and related roles for 2-FDG PET. Most critical is the assumption that there is an exact proportionality between glucose metabolism and 2-FDG metabolism, called the lumped constant, LC. This report describes that LC is not constant for a specific tissue and may be variable before and after disease treatment. The purpose of this work is not to deny the clinical value of 2-FDG PET; it is a reminder that when one extends the use of an appropriately qualified imaging method, new observations may arise and further validation would be necessary. Current understanding of glucose-based energetics in vivo is based on the quantification of glucose metabolic rates with 2-FDG PET, a method that permits the non-invasive assessment in various human disorders. However, 2-FDG is only a good substrate for facilitated-glucose transporters (GLUTs) but not for sodium-dependent glucose co-transporters (SGLTs), which have recently been shown to be distributed in multiple human tissues. Thus, the GLUT-mediated in vivo glucose utilization measured by 2-FDG PET would be blinded to the potentially substantial role of functional SGLTs in glucose transport and utilization. Therefore, in these circumstances the 2-FDG LC used to quantify in vivo glucose utilization should not be expected to remain constant. 2-FDG LC variations have been especially significant in tumors, particularly at different stages of cancer development, affecting the accuracy of quantitative glucose measures and potentially limiting the prognostic value of 2-FDG, as well as its accuracy in monitoring treatments. SGLT-mediated glucose transport can be estimated using α-methyl-4-deoxy-4-[18F]fluoro-D-glucopyranoside (Me-4FDG). Utilizing both 2-FDG and Me-4FDG should provide a more complete picture of glucose utilization via both GLUT and SGLT transporters in health and disease stages. Given the widespread use of 2-FDG PET to infer glucose metabolism, appreciating the potential limitations of 2-FDG as a surrogate for glucose metabolic rate and the potential reasons for variability in LC is relevant. Even when the readout for the 2-FDG PET study is only an SUV parameter, variability in LC is important, particularly if it changes over the course of disease progression (e.g., an evolving tumor).
Article
This study evaluated the performance of the Biograph Vision digital PET/CT system according to the NEMA NU 2-2012 standard (published by the National Electrical Manufacturers Association [NEMA]) to allow for a reliable, reproducible, and intersystem-comparable performance measurement. Methods: The new digital PET/CT system features silicon photomultiplier-based detectors with 3.2-mm lutetium oxyorthosilicate crystals and full coverage of the scintillator area. The PET components incorporate 8 rings of 38 detector blocks, and each block contains 4 × 2 mini blocks. Each mini block consists of a 5 × 5 lutetium oxyorthosilicate array of 3.2 × 3.2 × 20 mm crystals coupled to a silicon photomultiplier array of 16 × 16 mm, resulting in an axial field of view of 26.1 cm. In this study, PET/CT system performance was evaluated for conformation with the NEMA NU 2-2012 standard, with additional measurements described in the new NEMA NU 2-2018 standard. Spatial resolution, sensitivity, count-rate performance, accuracy of attenuation and scatter correction, image quality, coregistration accuracy, and time-of-flight performance were determined. Measurements were directly compared with results from its predecessor, the Biograph mCT Flow, using existing literature. Moreover, feasibility to comply with the European Association of Nuclear Medicine Research Ltd. (EARL) criteria was evaluated, and some illustrative patient PET images were obtained. Results: The Biograph Vision showed a transverse and axial spatial resolution of 3.6 and 3.5 mm, respectively, in full width at half maximum at a 1-cm offset from the center of the field of view (measured with a 22Na 0.25-mm point source), a NEMA sensitivity of 16.4 kcps/MBq, and a NEMA peak noise-equivalent count-rate of 306 kcps at 32 kBq/mL. Time-of-flight resolution varied from 210 to 215 as count-rate increased up to the peak noise-equivalent count-rate. The overall image contrast seen with the NEMA image quality phantom ranged from 77.2% to 89.8%. Furthermore, the system was able to comply with the current and future EARL performance criteria. Conclusion: The Biograph Vision outperforms the analog Biograph mCT Flow, and the system is able to meet European harmonizing performance standards.
Article
Introduction: The brain's energy budget can be non-invasively assessed with different imaging modalities such as functional MRI (fMRI) and PET (fPET), which are sensitive to oxygen and glucose demands, respectively. The introduction of hybrid PET/MR systems further enables the simultaneous acquisition of these parameters. Although a recently developed method offers the quantification of task-specific changes in glucose metabolism (CMRGlu) in a single measurement, direct comparison of the two imaging modalities is still difficult because of the different temporal resolutions. Thus, we optimized the protocol and systematically assessed shortened task durations of fPET to approach that of fMRI. Methods: Twenty healthy subjects (9 male) underwent one measurement on a hybrid PET/MRI scanner. During the scan, tasks were completed in four blocks for fMRI (4 × 30 s blocks) and fPET: participants tapped the fingers of their right hand repeatedly to the thumb while watching videos of landscapes. For fPET, subjects were randomly assigned to groups of n = 5 with varying task durations of 10, 5, 2 and 1 min, where task durations were kept constant within a measurement. The radiolabeled glucose analogue [18F]FDG was administered as 20% bolus plus constant infusion. The bolus increases the signal-to-noise ratio and leaves sufficient activity to detect task-related effects but poses additional challenges due to a discontinuity in the tracer uptake. First, three approaches to remove task effects from the baseline term were evaluated: (1) multimodal, based on the individual fMRI analysis, (2) atlas-based by removing presumably activated regions and (3) model-based by fitting the baseline with exponential functions. Second, we investigated the need to capture the arterial input function peak with automatic blood sampling for the quantification of CMRGlu. We finally compared the task-specific activation obtained from fPET and fMRI qualitatively and statistically. Results: CMRGlu quantified only with manual arterial samples showed a strong correlation to that obtained with automatic sampling (r = 0.9996). The multimodal baseline definition was superior to the other tested approaches in terms of residuals (p < 0.001). Significant task-specific changes in CMRGlu were found in the primary visual and motor cortices (tM1 = 18.7 and tV1 = 18.3). Significant changes of fMRI activation were found in the same areas (tM1 = 16.0 and tV1 = 17.6) but additionally in the supplementary motor area, ipsilateral motor cortex and secondary visual cortex. Post-hoc t-tests showed strongest effects for task durations of 5 and 2 min (all p < 0.05 FWE corrected), whereas 1 min exhibited pronounced unspecific activation. Percent signal change (PSC) was higher for CMRGlu (∼18%-27%) compared to fMRI (∼2%). No significant association between PSC of task-specific CMRGlu and fMRI was found (r = 0.26). Conclusion: Using a bolus plus constant infusion protocol, the necessary task duration for reliable quantification of task-specific CMRGlu could be reduced to 5 and 2 min, therefore, approaching that of fMRI. Important for valid quantification is a correct baseline definition, which was ideal when task-relevant voxels were determined with fMRI. The absence of a correlation and the different activation pattern between fPET and fMRI suggest that glucose metabolism and oxygen demand capture complementary aspects of energy demands.
Article
The haemodynamic responses to neural activity that underlie the blood-oxygen-level-dependent (BOLD) signal used in functional magnetic resonance imaging (fMRI) of the brain are often assumed to be driven by energy use, particularly in presynaptic terminals or glia. However, recent work has suggested that most brain energy is used to power postsynaptic currents and action potentials rather than presynaptic or glial activity and, furthermore, that haemodynamic responses are driven by neurotransmitter-related signalling and not directly by the local energy needs of the brain. A firm understanding of the BOLD response will require investigation to be focussed on the neural signalling mechanisms controlling blood flow rather than on the locus of energy use.