HAI Book 2020 - Dec 18B

01

Lemoine, Laetitia

P28: Visualizing brain astrocytes - cha
human AD brain tissue

Laetitia Lemoine1, Mona-Lisa Malarte1, Niina K

1Center of Alzheimer Research, Department of Neurob
Institutet, Stockholm, Sweden
2Theme Aging, Karolinska University Hospital, Stockh
Background: Astrocytosis has been shown to be an e
therapeutic target in AD. However, there are few posi
investigating in vivo astrocyte activation available, th
astrocytic tracers. Presently, the only PET tracer used
deuterium-L-deprenyl (11C-DED). The aim of our stu
reactive astrocytes and neuroinflammation in AD, and
Methods: In vitro characterization of 3H-BU99008 u
single concentration regional distribution binding assa
Autoradiographies on large frozen hemisphere using 3
DED.
Results: Saturation binding studies showed difference
control (Bmax: 58fmol/mg; Kd: 0.65nM). Competitio
nanomolar range for AD and control. Unlabeled depre
reverse was possible. Single concentration regional bi
binding in parietal cortex and cerebellum of AD in co
in hippocampus and temporal cortex. Autoradiograph
regional distribution between 3H-BU99008 and 3H-D
Conclusions: To be able to detect astrocytosis early in
data shows that BU99008, proposed as novel astrocyt
tissue in the nanomolar range, however with some dif
characterization is needed to understand its binding pr
astrocytic biomarkers.
Keywords: Deprenyl, astrocytes, BU99008, autoradio

10

aracterization of 3H-BU99008 in

Koistinen1, Agneta Nordberg1,2

biology, Care Sciences and Society, Karolinska
holm, Stockholm, Sweden
early phenomenon in AD, and seems to be a promising
itron emission tomography (PET) tracers for
erefore there is a growing interest in developing new
for the imaging of astrocyte reactivity in AD is 11C-
udy is to characterize BU99008, in the assessment of
d to compare its characteristics with the DED.
using competition, saturation binding assay as well as
ays was performed on AD and control brain tissues.
3H-BU99008 were performed and compared with 3H-

es between AD (Bmax: 90fmol/mg; Kd: 2.2nM) and
on binding studies showed a similar IC50 in the
enyl could not displace 3H-BU99008 when the
inding distribution showed significantly higher
omparison to control. High binding was also observed
hy on adjacent sections from AD cases shows similar
DED but with different binding intensity.
n AD, specific astrocyte biomarkers are needed. Our
te PET tracer, shows binding in CN and AD brain
fference in affinity with 3H-deprenyl. More
roperties and future use in comparison to existing

ography, PET

02

Levitis, Elizabeth

P29: Application of an epidemic sprea
beta accumulation in autosomal domin
carriers

Elizabeth Levitis1, Jacob W Vogel1, Gregory Ki
Alan C Evans1

1Montreal Neurological Institute, McGill University, M

Previous work has shown that the pattern of amyloid b
sporadic AD spectrum can be reproduced using an epi
spread of Aβ as a diffusive process occurring across n
with autosomal dominant AD and evaluate the utility
prediction.
Co-registered PiB PET scans from the Dominantly Inh
alongside Freesurfer-generated subject-specific DKT
for 78 ROIs using a cerebellar or brainstem reference
and follow-up, respectively. The ESM was first used t
trajectory of Aβ spread across a template structural co
relationship between two initial timepoints for each su
patterns of 53 MC who had subsequent visits (Fig 1).
accumulators based on their mean rate of change in co
relationship between model performance and Aβ spre
The ESM reproduced baseline aggregated deposition
posterior cingulate and caudal anterior cingulate regio
parameters fit between the first two timepoints, the ES
accuracy of 24% for subjects consistently accumulatin
This work demonstrates that Aβ spread may be descri
model. Efforts to predict out-of-sample regional patter
heterogeneous Aβ changes over the disease progressio

10

ading model to characterize amyloid
nant Alzheimer’s disease mutation

iar1, Thomas Funck1, Yasser Iturria-Medina1,

Montreal, QC, Canada
beta (Aβ) deposition in individuals spanning the
idemic spreading model (ESM), which simulates the
neuronal connections. We apply the ESM on subjects
of the ESM for longitudinal evaluation and

herited Alzheimer's Network dataset were used
segmentations to quantify Aβ binding probabilities
for 215 and 121 mutation carriers (MC) at baseline
to identify candidate seed regions and simulate the
onnectome. We then applied the ESM to learn the
ubject and extrapolated this relationship to predict Aβ
Categorizing subjects as accumulators or non-
ortical Aβ binding allowed us to explore the
ead.
patterns with an accuracy of 40% (Fig 2). The
ons were found as the most likely seeds. Using
SM learned the rate of Aβ change with an average
ng Aβ (Fig 3).
ibed in familial AD mutation carriers using a diffusion
rns using the ESM are limited due in part to
on.

03

10

04

Keywords: Alzheimer’s disease, amyloid beta, comput
10

tational modelling, DIAN
05

Lois, Cristina

P30: Phantom-based harmonization o
parameters between the Siemens ECA
MI PET/CT scanners

Cristina Lois1, Kira Grogg1, Georges El Fakri1,

1Massachusetts General Hospital, Boston, MA, US
Background:Accuracy of PETquantification depends
corrections. Systematic inter-scanner variability hamp
multiple scanners, and harmonization of imaging proc
The Siemens ECAT HR+ PET scanner (HR+) has bee
a GE Discovery MI (GE-DMI) PET/CT with increase
installed in our center.
Objective: To identify optimal reconstruction parame
values between data acquired using the GE-DMI and
combine data acquired in different individuals on thes
Methods: Jaszczak and Hoffman phantoms (total acti
scanners. Our typical 18F-AV1451 protocol (6 5-min f
filter) was applied for the HR+. For the GE-DMI, list
frame lengths (5, 2.5,1.5, 1-min). Different reconstruc
and without PSF and TOF, and various filters. Region
phantom sections (speheres, vials, background for Jas
ROI metrics were computed: contrast-to-noise ratio (C
Results and Conclusions: The GE-DMI reconstructio
quality and quantification (Fig.1). GE-DMI data recon
min frames, 7 mm filter, presented an optimal balance
similar to the HR+ (Hoffman voxel-wise correlation, r
human tau PET data set is ongoing.
[1] Joshi et al. Neuroimage 2009, 46(1)
[2] Pan et al. Med Phys. 2019

10

of PET image reconstruction
AT HR+ PET and the GE Discovery

Keith Johnson1, Julie Price1

s on detector hardware, image reconstruction and
pers the ability to compare and combine data from
cedures is needed to maximize study power [1].
en of primary use by our group for 20 years. Recently,
ed sensitivity, PSF and TOF capability [2], was

eters that achieve the most comparable quantitative
the legacy HR+ data. The ultimate goal is to robustly
se two scanners in our ongoing studies.
ivity ~50MBq18F-FDG) were imaged in both
frames, OSEM (16 subsets 4 iterations (16s4i), 5mm
t-mode phantom data were rebinned into varying
ctions were evaluated: OSEM 34s2i and 16s4i, with
ns-of-interest (ROI) were delineated in multiple
szcak; cortical, subcortical for Hoffman). Quantitative
CNR), signal-to-noise (SNR), contrast recovery (CR).
ons evaluated in phantoms had large impact in image
nstructed with OSEM 34s2i (no PSF, no TOF), 1.5-
e of CNR, SNR, and CR (Fig.1) and were the most
r2= 0.98, Fig.2) . Evaluation of these results in a

06

Keywords: Harmonization, quantification, phantoms
10

07

Martersteck, Adam

P31: Age prediction and amyloid depo

Adam Martersteck1,2, Pierre Besson2,3, Jaiashre
Parrish2, Aggelos Katsaggelos4, M.-Marsel Mes

1Mesulam Center for Cognitive Neurology and Alzhei
School of Medicine, Chicago, IL, US
2Department of Radiology, Northwestern University F
3Department of Neurological Surgery, Northwestern U
US
4Department of Electrical & Computer Engineering, N
5Department of Neurology, Northwestern University F
6Department of Psychiatry and Behavioral Sciences, N
Medicine, Chicago, IL, US

Background: Machine learning techniques to predict
be a potential biomarker for accelerated pathologic ag
potential protective mechanisms of resilience and resi
over age 80 with episodic memory at least as good as
previously been shown to have higher cortical thickne
over time.

Objective: We examined whether SuperAgers resist a
whether their predicted brain age is lower than their ch

Methods: We used a previously established graph con
healthy individual’s publicly available T1-weighted im
entirely based on the geometric shape of the brain. Br
surfaces: pial and white, left and right hemispheres, 10
Using 5-fold cross validation on 8,046 participants, th
average absolute error of 4.58 years. The model was i
florbetapir PET scans. Amyloid PET positivity was as
regions and full cerebellar reference region.

Results: gCNN predicted brain age from 11 SuperAg
(p<0.0001; mean difference 20.76 years ± 8.65 [s.d.])
amyloid positive (SUVR > 1.17).

Conclusions: SuperAgers show a mismatch between
our previous reports. AD pathology is present in appro
only <10% of SuperAgers. Both results suggest Super
related changes.

Keywords: Deep learning, successful aging, amyloid

10

osition in SuperAgers

Sridhar1, Beth Makowski-Woidan1, Todd
sulam1,5, Kathleen Bandt3, Emily Rogalski1,6

imer’s Disease, Northwestern University Feinberg
Feinberg School of Medicine, Chicago, IL, US
University Feinberg School of Medicine, Chicago, IL,
Northwestern University, Chicago, IL, US
Feinberg School of Medicine, Chicago, IL, US
Northwestern University Feinberg School of

t biologic brain age versus chronologic brain age may
ging. Here, we use these same techniques to identify
istance to brain aging in SuperAgers. SuperAgers are

that of average middle-age adults. SuperAgers have
ess than age-matched peers and resist cortical atrophy

amyloid accumulation using 18F-florbetapir PET and
hronologic age using a deep learning framework.
nvolutional neural network (gCNN), trained on 8,046
mages (age=6-89 years). The gCNN predicts age
riefly, it takes the x,y,z coordinates of four Freesurfer
0,242 vertices each, affinely registered to MNI-space.
he gCNN predicted age with Pearson’s r=0.93 and
implemented on MRI scans from 11 SuperAgers with
ssessed with 6 bilateral Clark et al. 2011 cortical

gers was significantly lower than their chronologic age
). Only one of the 80+-year-old SuperAgers was

chronologic and biologic brain age, consistent with
oximately 30% of cognitive healthy adults >65, but
rAging may inform mechanisms of resistance to age-

08

Mayblyum, Danielle

P32: PiB tracer delivery decreases ove
amyloid pathology measurement

Danielle Mayblyum1, J. Alex Becker1, Aaron Sc
Rubinstein1, Justin Sanchez1, Emma Thibault1, R
Bernard Hanseeuw1,3

1Massachusetts General Hospital, Boston, MA, US
2Brigham and Women Hospital, Boston, MA, US
3Cliniques Universitaires Saint-Luc, Brussels, Belgium
Background: PET measures of relative tracer deliver
progression and/or to better estimate PET measures of
PiB-R1, a measure of tracer delivery derived from the
dataset available in the Harvard Aging Brain Study (H
Methods: We analyzed baseline (N=269) and longitu
PiB-data from clinically normal HABS participants. W
SRTM2 in the same neocortical aggregate (FLR) with
between PiB-R1, demographics, and subsequent chan
Composite (PACC-5). We also evaluated cross-sectio
PiB-DVR.

Results: Baseline PiB-R1 was lower at older ages (Fig
not linearly associated with PiB-DVR (p=0.412), APO
baseline cognition (Fig.2A, p=0.879), or subsequent c

10

ertime in preclinical AD – impact for

chultz1, Michelle Farrell1, Heidi Jacobs1, Zoe
Reisa Sperling1,2, Julie Price1, Keith Johnson1,2,

m
ry are potentially useful to predict Alzheimer’s disease
f Aβ pathology. Therefore, we aimed to characterize
e PiB-PET full dynamic scan, in the longitudinal
HABS).
udinal (N=212, median follow-up: 4 years) dynamic
We calculated PiB-R1 and 40-60min PiB-DVR using
h cerebellar gray reference. We explored associations
nges in the Preclinical Alzheimer’s Cognitive
onal and longitudinal associations between PiB-R1 and

g.1, p<0.001) and in males (p=0.003). Overall, it was
OE e4 carriage (p=0.162), education years (p=0.968),
cognitive decline (Fig.2B, p=0.396). Of note, PiB-
09

DVR had a quadratic association with PiB-R1 (Fig.3A
increased, and then decreased with increasing levels o
predicted subsequent decline in PiB-R1 (Fig.3B, p=0.
similar when using Logan-DVR instead of SRTM2-D

Conclusion: PiB tracer delivery decreases with age in
with high-Aβ burden. Overall, the association with Aβ
between Aβ and tracer delivery at low retention levels
not associated with concurrent or subsequent cognitio
Keywords: R1, PiB, cognitive decline

11

A, p=0.034), suggesting that PiB-R1 initially
of PiB-DVR. Longitudinally, baseline PiB-DVR
.002), adjusting for demographics. Results were
DVR as both metrics were highly correlated (R2=0.96).

n clinically normal adults, specifically in individuals
β measures is weak although the relationships
s deserve further attention. The PiB-R1 parameter is
on.

10

Tudorascu, Dana

P33: MRI and PET data harmonizatio

Davneet Minhas1, Zixi Yang1, John Muschelli2,
Ellison3, Sterling Johnson3, Chester Mathis1, Ci
Handen1, Bradley Christian3, Ann Cohen1, Dana

1University of Pittsburgh, Pittsburgh, PA, US
2Johns Hopkins University, Baltimore, MD, US
3University of Wisconsin-Madison, Madison, WI, US

Introduction: Harmonization of MR data, structural s
quantification from multisite studies presents a challen
RAVEL1method to normalize T1-MRI intensities coll
populations. Effects on cortical thickness and on MR-
evaluated.

Methods: MRI and [11C]PiB scans for forty participa
from two sites were used. MR intensity histograms pr
FreeSurfer (v5.3.0) was used with and without RAVE
thicknesses and PiB SUVRs. Neuroradiologists rated
RAVEL. Effect sizes were computed for differences b
method.

Results: Figure 1 shows per-scan histograms of MRI
site, with and without RAVEL. PIB SUVRs showed v
and anteroventral striatum (AVS) regions (ICC’s >0.9
hippocampus segmentation were higher after RAVEL
5% to 40% for cortical thickness (Table 1).

Discussion: RAVEL was effective at MRI intensity n
individual intensity histograms were widely spread wi
observed with RAVEL. Neuroradiologist evaluations
improves segmentation. Harmonization using RAVEL
measures than on PET SUVR measures. Large differe
data from different sites/scanners should be properly a

11

on in a two-site Down syndrome study

, Charles Laymon1, Matthew Zammit3, Paul
iprian Crainiceanu2, William Klunk1, Benjamin
a Tudorascu1

segmentation, and MR-based PET SUVR
nging problem. In this study we used the
lected at two sites in Down syndrome (DS)
-based quantification of [11C]PiB SUVR were

ants (mean age: 39.13±6.60), age- and sex- matched
re and post RAVEL processing were evaluated.
EL for MR parcellation, producing regional cortical
FreeSurfer parcellation accuracy before and after
between PiB(+) and PiB(-) subjects within each

intensity classified by PiB global-SUVR status and
very good agreement between methods for both global
99, not shown). Neuroradiological ratings of the
L (Figure 2). Differences in effect sizes ranged from

normalization in this population. As seen in Figure 1
ithout RAVEL, while almost perfect overlap was
also demonstrated that RAVEL normalization
L had a much greater impact on cortical thickness
ences observed in effect sizes suggest that combining
addressed.

11

11

12

11

13

Keywords: PET Harmonization, MRI, Down syndrom
11

me,
14

Minhas, Davneet

P34: Sex-specific differences in [18F]A

Davneet Minhas1, Zheming Yu1, Charles Laymo
Tudorascu3, Howard Aizenstein4, Oscar Lopez2

1Department of Radiology, University of Pittsburgh, P
2Department of Neurology, University of Pittsburgh, P
3Department of Medicine, University of Pittsburgh, Pi
4Department of Psychiatry, University of Pittsburgh, P
Background: [18F]AV-1451 PET imaging has recentl
early tau deposition. The objective of this work was to
1451 retention, which may confound comparisons of t
cohorts.
Methods: For the UPitt cohort, [18F]AV-1451 PET an
demented” (CDR global: 0 or 0.5) Caucasian subjects
For the ADNI cohort, fully processed [18F]AV-1451 P
from the ADNI database for 159 non-demented Cauca
for the UPitt cohort were created using FreeSurfer cer
specific templates were created using the SPM12 DAR
to respective cohort templates using subject-specific f
smoothed by 6mm; no additional smoothing was appl
ANCOVAs of [18F]AV-1451 SUVR between sexes w
global amyloid status ([11C]PiB for UPitt; [18F]AV45

Results: Demographic data for the UPitt and ADNI c
[18F]AV-1451 SUVR in females was observed in men
lateral temporal cortical surfaces in both cohorts (Figu
significantly greater SUVR in meningeal tissue along
temporal cortex. Significantly greater retention in UPi

Conclusion: These findings suggest sex comparisons
reflective of differences in off-target meningeal bindin
tau deposition in females may, in fact, be confounded

11

AV-1451 off-target retention

on1, Brian Lopresti1, Beth Snitz2, Dana
2, Chester Mathis1, William Klunk4, Ann Cohen4

Pittsburgh, PA, US
Pittsburgh, PA, US
ittsburgh, PA, US
Pittsburgh, PA, US
ly been used to evaluate sex-specific differences in
o assess voxel-level differences in off-target [18F]AV-
tau deposition between sexes, in two independent

nd T1 MR images were acquired in 162 “non-
s (95F,67M;74±5y) at the University of Pittsburgh.
PET SUVR and T1 MR images were downloaded
asian subjects (83F,76M;76±8y). SUVR PET images
rebellar gray matter (GM) as reference. Cohort-
RTEL toolbox, and SUVR PET images were warped
flow fields. Warped UPitt SUVR images were
lied to warped ADNI SUVR images. Voxel-wise
were performed using age, MMSE, education, and
for ADNI) covariates.
cohorts are presented in Table 1. Significantly greater
ningeal tissue on the frontal, parietal, occipital, and
ure 1). In the UPitt cohort, females also demonstrated

the midline and anterior and posterior to the medial
itt cohort males was observed in the basal ganglia.
of tau deposition using [18F]AV-1451 may be
ng rather than actual tau deposition. Early detection of
d by [18F]AV-1451 meningeal binding.

15

Keywords: AV1451, flortaucipir, tau, off-target, sex
11

16

Moore, Annah

P35: Polygenic risk for Cerebrospinal
amyloid in clinically normal older adu

Annah Moore1,2,3, Logan Dumitrescu1,2, Aaron S
Richard Mayeux5,6,7, Reisa Sperling4, Timothy H

1Vanderbilt Memory and Alzheimer’s Center, Vanderb
2Vanderbilt Genetics Institute, Vanderbilt University M
3Department of Pharmacology, Vanderbilt University
4Department of Neurology, Massachusetts General H
5Department of Neurology, Columbia University Med
New York, NY, US
6The Taub Institute for Research on Alzheimer’s Disea
Medical Center and The New York Presbyterian Hosp
7The Institute for Genomic Medicine, Columbia Unive
Hospital, New York, NY, New York, NY, US
8Department of Neurology and Neurological Sciences
CA, US

Objective: Early detection may be key to the effectiv
polygenic risk scores (PRS) may increase sensitivity t
older individuals.

Methods: Two PRS were calculated using genome-w
(1) clinical AD diagnosis (PRS-ADdx; Jansen et al, 20
amyloid β42 (PRS-CSFAβ42; Deming et. al, 2017, n=3
computed for an independent cohort of 2,988 Non-his
screened for the Anti-Amyloid Treatment in Asympto
non-APOE SNPs (<1Mb) that were significant at the p
included in the PRS calculation. Linear regression mo
continuous amyloid levels covarying for age, sex, edu
interactions with PRS were also assessed.

Results: PRS-ADdx showed a significantly albeit low
models, both PRS were associated with amyloid (p-va
both PRS as simultaneous predictors, both remained s
together explained an additional 0.7% of variance bey
PRS (p<0.001), whereby the association was only pre
There was no age interaction for PRS-ADdx. There w

Discussion: Despite low correlation between CSF and
associated with amyloid deposition during preclinical
across different endophenotypes of AD may aide in im
and help to identify the highest risk patients for develo

11

l Fluid (CSF) amyloid predicts brain
ults

Schultz4, Neha Raghavan5,6,7, Keith Johnson4,
Hohman1,2, Elizabeth Mormino8

bilt University Medical Center, Nashville, TN, US
Medical Center, Nashville, TN, US
y Medical Center, Nashville, TN, US
Hospital, Nashville, TN, US
dical Center and The New York Presbyterian Hospital,

ase and The Aging Brain, Columbia University
pital, New York, NY, US
ersity Medical Center and The New York Presbyterian

s, Stanford University School of Medicine, Palo Alto,

ve treatment of Alzheimer’s Disease (AD), and
to detect preclinical disease among cognitively normal

wide association study (GWAS) summary statistics for
019, n=455,258), and (2) cerebrospinal fluid (CSF)
3,146). PRS derived from these GWAS studies were
spanic White, clinically normal older adults that were
omatic Alzheimer’s (A4) Study (age=71.4 ± 4.8). All
p=0.01 level in each parent GWAS study were
odels assessed the association between PRS and
ucation, APOE-ε4, and APOE-ε2. Age and sex

w correlation with PRS-CSFAβ42 (r=0.19). In separate
alues<0.001, Figure 1). In a competitive model with
significantly associated with amyloid (p<0.001) and
yond covariates. Age interacted with the CSF Aβ42
esent in those 70 and older (p<0.0001, Figure 2).
was no significant sex interaction with either score.

d clinical AD PRS, both scores independently are
phases of disease. Aggregate genetic risk scores
mproving prediction models for amyloid deposition
oping brain amyloid pathology.

17

11

18

Keywords: Polygenic Risk, Amyloid, Preclinical disea
11

ase, Age interaction
19

O’Dell, Ryan

P36: Association between cerebral am
density in Alzheimer’s disease: A mult

Ryan O’Dell1,2, Adam Mecca1,2, Ming-Kai Chen
Bartlett1,2, Emmie Banks1,2, Victoria Kominek1,
Toyonaga3, Yihuan Lu3, Nabeel Nabulsi3, Brent
Arnsten5, Yiyun Huang3, Richard Carson3, Chri

1Alzheimer’s Disease Research Unit, Yale University
2Department of Psychiatry, Yale University School of
3Department of Radiology and Biomedical Imaging, Y
US
4Program on Aging, Yale University School of Medici
5Department of Neuroscience, Yale University School
6Department of Neurology, Yale University School of

Introduction: Synaptic loss is an early pathology in A
correlate of cognitive impairment. Using [11C]UCB-J
synaptic loss in medial temporal and neocortical brain
further investigate the association of global and local

Methods: [11C]UCB-J and [11C]PiB binding were me
normal participants aged 55-85 years. For [11C]PiB an
were generated using SRTM2 with cerebellum as a re
ratios (DVR). For [11C]UCB-J,BPNDwas first compute
reference region and converted to DVRwith a cerebell

Results:The AD group showed significant reductions
temporal lobe, but also widespread across neocortical
both higher global and local amyloid was associated w
the AD group only, there were no significant associati
synaptic density in any ROI.

Conclusions: Using [11C]UCB-J PET, widespread red
participants. Cortical regions associated with the least
temporal lobe) were associated with the most pronoun
negative associations between both global and local am
were driven by group differences, and not significant
consistent with postmortem studies describing the pat
AD.

Funding Sources:P50-AG047270 (CHvD,MKC), K2
R01-AG062276 (CHvD), Dana Foundation (MKC), R
Thomas P. Detre Fellowship (RSO).

Keywords: Alzheimer’s disease, amyloid, synaptic den

12

myloid accumulation and synaptic
titracer PET study

n3, Tyler Godek1,2, Joanna Harris1,2, Hugh
,2, Wenzhen Zhao1,2, Mika Naganawa3, Takuya
t Vander Wyk4, Pradeep Varma3, Amy
istopher van Dyck1,2,5,6

School of Medicine, New Haven, CT, US
Medicine, New Haven, CT, US
Yale University School of Medicine, New Haven, CT,

ine, New Haven, CT, US
l of Medicine, New Haven, CT, US
Medicine, New Haven, CT, US

Alzheimer’s disease (AD) and the major structural
–PET, we have recently demonstrated extensive
n regions of participants with AD. In this study, we
amyloid burden with synaptic density.

easured in 38 participants with AD and 20 cognitively
nalysis, parametric images of binding potential(BPND)
eference region and converted to distribution volume
ed using SRTM2 with a shrunken centrum semiovale
lum reference region.

s of synaptic density, most pronounced in medial
and subcortical regions. Among pooled participants,

with lower synaptic density in the majority of ROIs. In
ions between either global or local amyloid and

ductions of synaptic density were observed in AD
t robust accumulation of fibrillar amyloid (medial
nced reductions in synaptic density. Significant
myloid and synaptic density in the pooled sample
within the AD group alone. Our findings are
tterns of fibrillar amyloid burden and synaptic loss in

23-AG057784 (APM), R01-AG052560 (REC, CHV),
R25-MH071584 (RSO), T32-MH019961 (RSO), and

nsity

20

Ottoy, Julie

P37: A clinically applicable proxy for
amyloid-beta in Alzheimer’s disease: B
AV45 PET

Julie Ottoy1, Min Su Kang2,4, Jeroen Verhaeghe
Hsiao4, Jenna Stevenson2, Nesrine Rahmouni2, D
Paul Soucy4, Serge Gauthier1,2, Pedro Rosa-Net

1Molecular Imaging Center Antwerp, University of An
2Translational Neuroimaging Laboratory, The McGill
Montréal, QC, Canada
3Department of Nuclear Medicine, Antwerp Universit
4Montreal Neurological Institute, Montréal, QC, Cana

Introduction: The SUVR has been widely applied to
trials of Alzheimer’s disease. However, it is prone to b
blood flow alterations. Here, we report on a clinically
imaging technique to accurately quantify Aβ.

Methods: Five PSEN1 carriers (MMSE 26.4±3) unde
with continuous intravenous tracer infusion, and conti
distribution based on B/I (VT,B/I) was determined as th
concentration in tissue (CPET) to 3 unmetabolized arter
against the full pharmacokinetic ‘gold standard’ VT,2T
metabolite-corrected plasma input functions. Addition
determined as the VT target-to-reference ratios (refere
(WM), or pons).

Results: The B/I method demonstrated true equilibriu
rate of tracer concentration change (e.g., Cparietal_lobe: 2
VT,B/I was similar to the full-kinetic VT,2TCM and VT,Lo
and 3.3±0.7%, respectively) (Fig.2).
DVRB/I normalized to the WM or pons was positively
pons: R=0.7,P<0.001). The WM-DVRB/I showed +6.2
compared to their DVR2TCM. Cerebellar plaques were
mutation (Fig.3), resulting in a negative correlation be

Conclusion: A 20-min static scan in combination with
to reproduce the full-kinetic VT. Considering a suitabl
quantification without the need for blood sampling. T
clinical trials.

12

gold standard quantification of
Bolus plus constant infusion 18F-

e1, Sigrid Stroobants3, Reda Bouhachi4, Chris
Daniel Chartrand4, Gassan Massarweh4, Jean-
to2,4, Steven Staelens1

ntwerp, Antwerp, Belgium
l University Research Centre for Studies in Aging,

ty Hospital, Antwerp, Belgium
ada
quantify amyloid-beta (Aβ) in the clinic and anti-Aβ
bias induced from reference region selection and
y feasible bolus-plus-constant-infusion (B/I) PET

erwent dynamic 90-min [18F]-AV45-PET on HRRT
inuous and manual arterial sampling. Total volume of
he equilibrium ratio (50-70min) of tracer
erial plasma samples (CP). The VT,B/I was validated
TCM and the graphical VT,Logan using individual
nally, distribution volume ratios (DVR) were
ence: cerebellar grey (CB), subcortical white matter

um conditions as both CPET and CP demonstrated low
2.0±1.4%, CCB: 3.1±3.4%, and CP: 5.6±3.5%) (Fig.1).
ogan (average difference over cortical lobes: 2.9±0.8

y correlated to the full-kinetic VT of cortical areas (e.g.
2±1.3 and the pons-DVRB/I +6.9±1.2% difference
evident in two subjects with the same PSEN1
etween DVRB/I and the full-kinetic VT of all subjects.

h 3 metabolite-corrected arterial samples is sufficient
le reference region, the B/I method provides accurate
This method is promising for wide adaptation in

21

12

22

Keywords: Florbetapir, Positron emission tomograph
12

hy, Alzheimer’s disease, constant infusion, equilibrium
23

Manber, Richard

P38: Amyloid PET variability due to v
anatomical segmentation

Alessandro Palombit1, Richard Manber1, Richar

1IXICO plc., London, UK

Background: Variability in MRI acquisition protocol
anatomical segmentation differences, introducing add
amyloid beta PET (Ab-PET) SUVR.

Aims: To quantify the variability of Ab-PET attributa
variation in MR acquisition protocol.

Methods: We employ two T1w-MRI datasets 1) singl
protocols; HZ) and 2) four subjects scanned with diffe
NHZ).

Once bias field-corrected and skull-stripped, automati
and a subject-specific T1w template constructed with

Ab-PET images for 20 ADNI2 (http://adni.loni.ucla.e
non-linearly warped to each subject-specific template.

Inter-segmentation Ab-PET variability was measured
segmentations (within-subject and ROI) divided by th

Intra-segmentation Ab-PET variability was quantified
within each ROI, averaged across segmentations and d

Results: Ab-PET variability was not significantly diff
datasets, Figure 2A; paired t-test, p > 0.05, Cohen’s d
comparable to scan-rescan level (5-9%).

Intra-segmentation variability (uptake variability with
segmentation-related variability condition (Figure 2B;
18.0-28.6%, significantly different between harmoniz
t-test, p > 0.05) with however small effect size (Cohen

Conclusions: We found that the variability of regiona
differences is within the physiological scan-rescan ran
suggesting the tolerability of using compliant MR data

12

variation in MRI protocol and

rd Joules1, Robin Wolz1

l (e.g. voxel-size, contrast parameters) cause
ditional sources of cross-sectional variance in regional

able to brain segmentation differences arising from

le volunteer on different scanners (18 harmonised
erent acquisitions (eight non-harmonised protocols;

ic whole brain parcellation was performed with LEAP
ANTs from skull-stripped T1w-MRI (see Figure 1).
edu) subjects, representative of the AD spectrum, were
.
by the standard deviation of regional uptake across
he global average.
d by the standard deviation of the uptake distribution
divided by the global within-ROI average.
ferent between harmonization levels (HZ vs NHZ
< 0.01) with an inter-quartile range 1.6-5.3%,

hin-ROI) was an order of magnitude higher than any
; paired t-test, p < 0.05) with an inter-quartile range of
zation groups (HZ vs NHZ datasets, Figure 2A; paired
n’s d = 0.23).
al Ab-PET introduced by MRI-related segmentation
nge and lower than the within-ROI variability
a with minor impact in Ab-PET studies.

24

Keywords: PET, MRI, Amyloid, Segmentation,
12

25

Manber, Richard

P39: The impact of automatic Tau PE
and power analysis in AD

Alessandro Palombit1, Richard Manber1, Richar

1IXICO plc., London, UK

Background: Tau protein-targeting Positron Emissio
PET) offer promises in Alzheimer’s disease (AD) clin
variability is unclear.

Aims: Explore the variability introduced by different

Methods: A total of 96 ADNI subjects (34 normal co
randomly selected (http://adni.loni.ucla.edu) obtaining
along Tau-PET regional SUVR measures distributed f

T1w-MRI data was anatomically segmented with LEA
separately by means of IRTK or ANTs tools.

Regional Tau-PET SUVR measures, obtained using th
finally aggregated to merge LEAP regions into three B

Results: Braak-SUVR estimates obtained with the tw
them with determination coefficient R2= 0.97 and wit
R2=0.89/0.85 with IRTK or ANTs-based pipelines).

Paired-condition SUVR absolute difference was not c
0.01,p>0.05).

Relative SUVR difference within-region (ANTs – IRT
0.84% (75% of values differed less than 3.31% in abs
suggesting a possible ANTs-related underestimation o

Conclusions: Automatic Tau-PET processing require
repeatable endpoint generation.

Here we found that the variability introduced by differ
typical scan-rescan variability providing alternative pr
on sample size estimation, and therefore on trial desig

Braak-SUVR measures in this test cohort (NC-AD) di
0.185 with IRTK and ANTs (d=-0.255 with ADNI as

A power analysis based on these figures for example w
groups) 563 and 459 subjects respectively with IRTK

Keywords: PET, MRI, Tau, Power analysis, Braak

12