The cortical signature of prodromal AD Regional thinning predicts mild AD dementia
We previously used exploratory analyses across the entire cortex to determine that mild Alzheimer disease (AD) is reliably associated with a cortical signature of thinning in specific limbic and association regions. Here we investigated whether the cortical signature of AD-related thinning is present in individuals with questionable AD dementia (QAD) and whether a greater degree of regional cortical thinning predicts mild AD dementia. Participants included 49 older adults with mild impairment consistent with mild cognitive impairment (Clinical Dementia Rating [CDR] = 0.5) at the time of structural MRI scanning. Cortical thickness was measured in nine regions of interest (ROIs) identified previously from a comparison of patients with mild AD and controls. Longitudinal clinical follow-up revealed that 20 participants converted to mild AD dementia (progressors) while 29 remained stable (nonprogressors) approximately 2.5 years after scanning. At baseline, QAD participants showed a milder degree of cortical thinning than typically seen in mild AD, and CDR Sum-of-Boxes correlated with thickness in temporal and parietal ROIs. Compared to nonprogressors, progressors showed temporal and parietal thinning. Using receiver operating characteristic curves, the thickness of an aggregate measure of these regions predicted progression to mild AD with 83% sensitivity and 65% specificity. Thinning in specific cortical areas known to be affected by Alzheimer disease (AD) is detectable in individuals with questionable AD dementia (QAD) and predicts conversion to mild AD dementia. This method could be useful for identifying individuals at relatively high risk for imminent progression from QAD to mild AD dementia, which may be of value in clinical trials.
The cortical signature of prodromal AD
Regional thinning predicts mild AD dementia
Akram Bakkour, BS
John C. Morris, MD,
Bradford C. Dickerson,
Objective: We previously used exploratory analyses across the entire cortex to determine that
mild Alzheimer disease (AD) is reliably associated with a cortical signature of thinning in specific
limbic and association regions. Here we investigated whether the cortical signature of AD-related
thinning is present in individuals with questionable AD dementia (QAD) and whether a greater
degree of regional cortical thinning predicts mild AD dementia.
Methods: Participants included 49 older adults with mild impairment consistent with mild cogni-
tive impairment (Clinical Dementia Rating [CDR] ⫽ 0.5) at the time of structural MRI scanning.
Cortical thickness was measured in nine regions of interest (ROIs) identified previously from a
comparison of patients with mild AD and controls.
Results: Longitudinal clinical follow-up revealed that 20 participants converted to mild AD demen-
tia (progressors) while 29 remained stable (nonprogressors) approximately 2.5 years after scan-
ning. At baseline, QAD participants showed a milder degree of cortical thinning than typically
seen in mild AD, and CDR Sum-of-Boxes correlated with thickness in temporal and parietal ROIs.
Compared to nonprogressors, progressors showed temporal and parietal thinning. Using receiver
operating characteristic curves, the thickness of an aggregate measure of these regions pre-
dicted progression to mild AD with 83% sensitivity and 65% specificity.
Conclusions: Thinning in specific cortical areas known to be affected by Alzheimer disease (AD) is
detectable in individuals with questionable AD dementia (QAD) and predicts conversion to mild
AD dementia. This method could be useful for identifying individuals at relatively high risk for
imminent progression from QAD to mild AD dementia, which may be of value in clinical trials.
AD ⫽ Alzheimer disease; ADT ⫽ AD signature thickness; AUC ⫽ area under the curve; CDR ⫽ Clinical Dementia Rating;
CDR-SB ⫽ CDR Sum-of-Boxes; eTIV ⫽ estimated total intracranial volume; EV ⫽ entorhinal volume; HV ⫽ hippocampal
volume; MCI ⫽ mild cognitive impairment; MCT ⫽ mean cortical thickness; MMSE ⫽ Mini-Mental State Examination; MTL ⫽
medial temporal lobe; MTLT ⫽ medial temporal lobe thickness; OC ⫽ older controls; QAD ⫽ questionable AD dementia; ROC ⫽
receiver operating characteristic; ROI ⫽ region of interest; WBV ⫽ whole brain volume.
Anatomic abnormalities of brain regions known to be early sites of Alzheimer disease (AD)
pathology, such as medial temporal lobe (MTL) regions including the entorhinal cortex and
hippocampal formation, can be detected in prodromal AD prior to dementia.
Although it is
well known that early in its clinical course AD affects non-MTL neocortical association re-
there has been little investigation of neocortical anatomy in prodromal AD prior to
dementia. The few investigations of cortical anatomic abnormalities in mild cognitive impair-
or prodromal AD
have used techniques that involve exploratory mapping of
Address correspondence and
reprint requests to Dr. Brad
Dickerson, MGH Gerontology
Research Unit, 149 13th St.,
Suite 2691, Charlestown, MA
Supplemental data at
Editorial, page 1038
e-Pub ahead of print on December 24, 2008, at www.neurology.org.
From the Departments of Psychiatry (A.B.) and Neurology (B.C.D.), Massachusetts Alzheimer’s Disease Research Center and Athinoula A. Martinos
Center for Biomedical Imaging, Massachusetts General Hospital and Harvard Medical School, Boston; Division of Cognitive and Behavioral
Neurology (B.C.D.), Department of Neurology, Brigham & Women’s Hospital, Boston, MA; and Department of Neurology and Alzheimer’s Disease
Research Center (J.C.M.), Washington University School of Medicine, St. Louis, MO.
Supported by grants from the NIA K23-AG22509, R01-AG29411, R21-AG29840, P50-AG05681, and P01-AG03991, NCRR P41-RR14075, U24-
RR021382, the Alzheimer’s Association, Howard Hughes Medical Institute, and the Mental Illness and Neuroscience Discovery (MIND) Institute.
Disclosure: The authors report no disclosures.
1048 Copyright © 2009 by AAN Enterprises, Inc.
the entire cortex. While these techniques are
well suited to disorders in which the localiza-
tion of pathology is unknown, it is possible to
use in vivo imaging and postmortem patho-
logic data from patients with AD dementia to
predict the localization of anatomic abnor-
malities in prodromal AD. Such an approach
employing disease-signature cortical regions
of interest (ROIs) is powerful in that it en-
ables hypotheses to be tested about the se-
quence of involvement of cortical regions in
AD progression through comparison of effect
sizes from cross-sectional data from patients at
different clinical stages in the disease.
Furthermore, there are questions about ab-
normalities of cortical anatomy in prodromal
AD that, to our knowledge, have received no
investigation. How do cortical anatomic ab-
normalities relate to symptoms in the mildest
predementia phases of the disease? Which
neocortical measures are best in terms of sen-
sitivity and specificity for early diagnosis, and
do they improve upon the predictive power of
well-accepted morphometric measures, such
as volumes of the hippocampal formation, en-
torhinal cortex, or whole brain?
To address those questions, we studied a
sample of individuals with questionable de-
mentia (Clinical Dementia Rating [CDR]
0.5) whose symptoms and signs were largely
similar to MCI. Longitudinal clinical data af-
ter the MRI scan were used to classify subjects
into progressors to mild AD dementia (CDR
1) vs nonprogressors. A set of AD signature
cortical ROIs—found previously in a separate
sample of patients with mild AD (CDR 1) to
be consistently affected
—was used to extract
regional thickness measures which were em-
ployed in subsequent analyses. ROI measures
were compared in the group of MCI progres-
sors vs nonprogressors, and were also com-
pared to controls and patients with mild AD.
In addition, receiver operating characteristic
(ROC) curves were generated for the various
novel anatomic measures obtained, which
were compared to ROC curves for standard
morphometric measures. Across all subjects,
ROI measures were studied in relation to clin-
ical measures for purposes of clinical valida-
tion. Finally, an analysis was performed to
demonstrate a map of the spatial pattern of
thinning associated with prodromal AD
across the cerebral cortex.
METHODS Participants, clinical assessment, and
MRI data acquisition.
Forty-nine volunteer participants (age
64–90, 20 women) were studied, data from some of whom have
been published previously.
Participants were recruited from
the ongoing longitudinal sample of the Washington University
Alzheimer’s Disease Research Center using procedures approved
by Washington University’s human subjects committee, and as-
sessed using procedures previously described.
For the purposes of the present study, the clinical diagnostic
categories included questionable dementia of the Alzheimer type
(QAD) (CDR Rating ⫽ 0.5) and mild AD (CDR Rating ⫽ 1).
All participants were classified as QAD at the time of baseline
MRI scan. Based on annual clinical follow-up, 29 individuals
remained with a CDR Rating of 0.5 (nonprogressors) while
20 converted to mild AD (declined to a CDR Rating ⫽ 1;
Multiple (three or four) structural T1-weighted magnetization-
prepared rapid gradient echo images were acquired on a 1.5 T
Siemens Vision scanner (Siemens Medical Systems, Erlangen,
Germany) using the following parameters: repetition time/echo
time/inversion time/delay ⫽ 9.7/4/20/200 msec, flip angle ⫽
10º, 256 ⫻ 256 (1 mm ⫻ 1 mm) in-plane resolution, 128 sagit-
tal 1.25 mm slices without gaps. These images were motion-
corrected and averaged together during processing. These data
are openly available to the community, thanks in part to re-
sources provided by the Washington University Alzheimer’s Dis-
ease Research Center (http://www.oasis-brains.org/).
MRI morphometric data analysis: Automated tissue
segmentation and surface reconstruction and align-
ment of participants.
These methods have been previously
described in detail, including the surface-based cortical thickness
processing and spherical registration to align subjects’ cortical
The methods are summarized in appendix e-1 on
Web site at www.neurology.org. The Freesurfer
software used to analyze and visualize data in this study is freely
MRI morphometric data analysis: Quantification of
magnitude of atrophy within regions of interest.
mated volume measures for whole brain, hippocampus, and en-
torhinal cortex were adjusted for estimated total intracranial
volume (eTIV, calculated as described previously
) by dividing
regional volume by eTIV. Nine regions of interest (ROIs) con-
stituting the cortical signature of AD derived from an explor-
atory cortical thickness analysis comparing 115 older controls to
29 subjects with AD
were used to generate regional thickness
measures for each subject in this study. These adjusted volume
and regional thickness measures were used for descriptive pur-
poses and to calculate the mean difference in values of each ROI
between the progressor and nonprogressor groups, the percent
thinning in progressors relative to nonprogressors ([mean thick-
ness of nonprogressor group ⫺ mean thickness of progressor
group]/mean thickness of nonprogressor group), and the Cohen
d effect sizes of progression-related differences for each ROI. In
addition, we analyzed a measure of the average thickness of these
9 ROIs (average AD-signature ROI thickness), previously shown
to be sensitive to early presymptomatic effects of AD.
Statistical analyses. The eTIV-adjusted volume and regional thickness
measures were normalized to a sample of age-matched controls by calcu-
Neurology 72 March 24, 2009 1049
lating a Z score based on the mean and SD of an age-matched group of
older controls (OC; CDR ⫽ 0) as follows: Z ⫽ (x ⫺
. This Z score was used in all statistical analyses. All nor-
malized (for age) thickness measurements were examined by analysis of
variance, with clinical outcome as the independent variable (progressor vs
nonprogressor). Thickness is not correlated with intracranial volume (data
not shown), so is not adjusted for eTIV. Measures from both hemispheres
were pooled to make one measure per region for each subject. Pearson
correlations and a multiple linear regression analysis were performed to
examine relationships between regional anatomic measures and the mea-
sures of symptom severity (the CDR Sum-of-Boxes [CDR-SB
]) and that
of severity of impairment on cognitive testing (Mini-Mental State Exami-
). ROC curves were used to assess the sensitivity and
specificity of different measures. The ROC performance of particular
morphometric measures was compared graphically, and for simplicity we
highlight here the peak classification performance in which both sensitiv-
ity and specificity were greater than 50% (i.e., top left peak in non-shaded
area in figure 2A). These statistical analyses were performed using SPSS
13.0 (SPSS, Chicago, IL).
RESULTS Clinical features. Participant characteris-
tics are presented in table 1. For the entire sample,
the average duration of clinical follow-up after the
MRI scan was 2.7 (SD ⫽ 1.7) years with the progres-
sor and nonprogressor groups being followed for 2.3
(SD ⫽ 1.1) and 3 (SD ⫽ 1.4) years. The progressor
and nonprogressor groups did not differ on the basis
of age [F(1) ⫽ 2.7, p ⫽ 0.11], but they did differ on
CDR-SB [F(1) ⫽ 19.9, p ⬍ 0.001] and MMSE
scores [F(1) ⫽ 10.1, p ⬍ 0.005].
Morphometric correlates of symptom severity. Within
the entire sample, several cortical thickness measures
correlated with the relative severity of cognitive im-
pairment as measured by CDR-SB, including medial
temporal (r ⫽⫺0.36, p ⬍ 0.01), inferior temporal
(r ⫽⫺0.30, p ⬍ 0.04), and superior parietal cortex
(r ⫽⫺0.32, p ⬍ 0.03), with trends (p ⬍ 0.1) for
temporal pole (r ⫽⫺0.24) and angular gyrus (r ⫽
⫺0.28). In addition, CDR-SB correlated with aver-
age AD-signature thickness (r ⫽⫺0.38, p ⬍ 0.01),
and with mean thickness across the entire cortical
mantle (r ⫽⫺0.32, p ⬍ 0.03).
Similarly, a number of cortical thickness measures
correlated with MMSE score: angular gyrus (r ⫽
0.45, p ⬍ 0.001), superior frontal gyrus (r ⫽ 0.34,
p ⬍ 0.02), superior parietal lobule (r ⫽ 0.38, p ⬍
0.01), supramarginal (r ⫽ 0.32, p ⬍ 0.03), mean
cortical thickness (r ⫽ 0.34, p ⬍ 0.02), and average
AD-signature thickness (r ⫽ 0.35, p ⬍ 0.02).
As for volumetric measures, the only correlation
with CDR-SB or MMSE was observed with entorhi-
nal cortex volume, which showed a trend-level (r ⫽
⫺0.24, p ⬍ 0.1) effect.
Morphometric differences between progressors and
Mean cortical thickness was 3.2%
thinner in progressors than nonprogressors [nonpro-
gressors ⫽ 2.11 (SD ⫽ 0.10) mm, progressors ⫽
2.05 (SD ⫽ 0.14) mm; F(1) ⫽ 4.6, p ⫽ 0.037, Co-
hen d ⫽ 0.54].
As mentioned above, figure 1 (top) illustrates the
localization of the nine AD-signature cortical ROIs,
derived from a separate sample of 115 OC vs 29
Thinning in prodromal AD (the progressor
group) was most prominent in the rostral medial
temporal cortex, with thinning (compared to non-
progressors) of 0.28 mm (10.1%, Cohen d ⫽ 0.79).
Three other regions differed (p ⬍ 0.05) between the
two groups: inferior temporal gyrus (0.17 mm
[6.3%] thinning, Cohen d ⫽ 0.73), superior parietal
lobule (0.12 mm [5.8%] thinning, Cohen d ⫽ 0.70),
and temporal pole (0.12 mm [4.7%] thinning, Co-
hen d ⫽ 0.51). Two regions showed trends (p ⬍ 0.1)
toward thinning in prodromal AD: precuneus (0.15
mm [7.03%] thinning, Cohen d ⫽ 0.52), and angu-
lar gyrus (0.1 mm [4.43%] thinning, Cohen d ⫽
0.48). Weaker effects in the same direction were ob-
served in the other three ROIs. Group mean and
standard error data are shown in figure 1. Additional
details are shown in table 2.
The progressor and nonprogressor groups did not
differ for whole brain volume (p ⬎ 0.3), but both
hippocampal [F(1) ⫽ 4.5, p ⬍ 0.04, Cohen d ⫽ 0.4]
and entorhinal [F(1) ⫽ 6.4, p ⬍ 0.02, Cohen d ⫽
0.6] volumes were smaller in the progressors than
nonprogressors, as expected.
ROC analysis. The goal of this analysis was to provide
data to assist in the comparison of the cortical thickness
measures of primary interest in this study with morpho-
metric measures employed more widely in previous
studies of the prediction of progression to dementia.
Rather than discriminant function or logistic regression
analysis, ROC was chosen to explore the full spectrum
of performance across a range of sensitivity and specific-
ity. Full curves are presented in figures, and here we
compare measures by focusing on the peak performance
point (highlighted in the figures) that optimized both
sensitivity and specificity.
Of the standard volumetric measures, both hip-
pocampal and entorhinal volume outperformed
whole brain volume across a wide range of the ROC
curve, with the best peak performance by entorhinal
cortex (sensitivity and specificity of 72.4% and 65%;
hippocampal sensitivity and specificity were 83%
Table 1 Demographic and clinical characteristics of participants at baseline
Group No. Age, y
(M/F) MMSE CDR (0/0.5/1)
Nonprogressors 29 77 (60–90) 14/7 28 (21–30) 0/29/0 2 (1–5)
Progressors 20 73.5 (64–86) 6/10 26 (19–30)* 0/20/0 3 (2–5)*
Values represent median (range).
*p ⬍ 0.005 (Different from nonprogressors).
MMSE ⫽ Mini-Mental State Examination; CDR ⫽ Clinical Dementia Rating.
1050 Neurology 72 March 24, 2009
and 50%; figure 2A). Area under the curve (AUC)
measures for whole brain volume (WBV), hippocam-
pal volume, and entorhinal volume are 0.59, 0.65,
Since AD is primarily a gray matter disease with
important degenerative effects on the cerebral cortex,
we hypothesized that, among global measures, mean
cortical thickness would be more sensitive than
Figure 1 Regions of interest (ROIs) derived from previous exploratory analysis that identified foci of
thinning in mild Alzheimer disease,
and thinning in cortical ROIs in patients with questionable AD
who later progressed to mild AD dementia (QAD-P) compared to QAD nonprogressors (QAD-NP)
Top: ROIs derived from previous exploratory analysis that identified foci of thinning in mild AD.
(A) Medial temporal cortex,
(B) inferior temporal gyrus, (C) temporal pole, (D) angular gyrus, (E) superior frontal gyrus, (F) superior parietal lobule, (G)
supramarginal gyrus, (H) precuneus, (I) inferior frontal sulcus. Bottom: Thinning in cortical ROIs in QAD-P compared to
QAD-NP. Bar graphs show mean cortical thickness within each ROI in the two groups (middle two bars) and, for comparison
purposes, in a sample of 115 older controls (OC, left bar) and 29 patients with mild AD (AD, right bar). Error bars indicate 1
SEM. *p ⬍ 0.05, **p ⬍ 0.005. See table 2 for additional detail.
Neurology 72 March 24, 2009 1051
whole brain volume, which reflects white matter vol-
ume as well as gray matter volume. Therefore, an
ROC comparison was made of the predictive classifi-
cation performance of mean cortical thickness vs
whole brain volume. Mean cortical thickness demon-
strated predictive classification performance that was
better than whole brain volume, with a peak sensitiv-
ity of 69% and specificity of 65% (figure 2B) as com-
pared to the peak sensitivity and specificity for whole
brain volume of 62% and 55%. AUC measures for
WBV and mean cortical thickness are 0.59 and 0.67.
Next, we performed similar analyses of the re-
gional cortical thickness measures. Among all re-
gional thickness measures (figure e-1 and appendix
e-1), superior parietal, precuneus, and inferior fron-
tal cortices showed relatively high sensitivity and me-
dial temporal, temporal pole, and superior frontal
cortices showed relatively high specificity. Of the
nine cortical ROI measures, the medial temporal
lobe thickness had the best peak performance, with a
sensitivity and specificity of 83% and 55%.
Finally, we compared the predictive classification
performance of the average AD-signature thickness
vs MTL thickness alone vs entorhinal volume (figure
2C). The AD signature measure performed best,
with a sensitivity of 83% and specificity of 65%.
AUC measures for hippocampal volume, MTL
thickness, and mean AD signature thickness are
0.65, 0.72, and 0.73.
An exploratory analysis of cortical thickness across
the entire mantle (figure 3) revealed a pattern of thin-
ning in prodromal AD similar to, but expressed to a
lesser degree than, that seen in mild AD.
DISCUSSION In this study, we used existing
knowledge about regional cortical thinning in a sam-
ple of patients with mild AD dementia to investigate,
using an a priori ROI-based approach, regional corti-
cal thinning in a sample of patients with questionable
AD dementia, many characteristics of which are sim-
ilar to those of patients with MCI. The magnitude of
thinning of certain regions correlates with the relative
severity of very mild symptoms in daily life (as mea-
sured by CDR-SB) and signs on performance test-
ing (MMSE). Furthermore, the thickness of
particular regions of the cerebral cortex known to
undergo thinning in mild AD provides a useful
measure for the prediction of progression from
questionable AD dementia (CDR 0.5) to mild AD
dementia (CDR 1).
While extensive investigation has been performed
in MCI of certain morphometric measures, such as
the volumes of the whole brain, ventricular system,
hippocampal formation, and entorhinal cortex,
there is scant evidence regarding anatomic abnor-
malities in other cortical brain regions. Most of
the literature on cortical atrophy in MCI has fo-
cused on exploratory mapping of atrophy in MCI
compared to controls or patients with AD,
in MCI converters to mild AD dementia com-
pared to nonconverters.
In the former studies of atrophy in MCI in which
the diagnostic outcome is unknown, atrophy pat-
terns are largely similar to, although of lesser magni-
tude, than those of mild AD dementia. Of the
previous investigations that have followed patients
clinically after scanning, two demonstrated distrib-
uted atrophy patterns in the MCI progressors that
are typical of those of patients with AD when com-
paring MCI progressors to controls,
one study showed more widespread atrophy when
comparing MCI progressors to nonprogressors,
while the other showed much less widespread differ-
ences involving only supramarginal, inferior frontal,
and hippocampal regions.
A third investigation identi-
Table 2 Quantitative metrics of thinning by region
Mean thickness, mm (SD)
effect sizeNonprogressors Progressors
Medial temporal 2.80 (0.33) 2.52 (0.38) 0.28* 10.1 0.79
Inferior temporal 2.64 (0.22) 2.47 (0.23) 0.17† 6.3 0.73
Temporal pole 2.59 (0.23) 2.47 (0.25) 0.12† 4.7 0.51
Angular gyrus 2.22 (0.17) 2.12 (0.24) 0.10 4.4 0.48
Superior parietal 2.01 (0.16) 1.90 (0.17) 0.12† 5.8 0.70
Supramarginal 2.24 (0.21) 2.19 (0.26) 0.05 2.3 0.22
Precuneus 2.19 (0.28) 2.03 (0.31) 0.15 7.0 0.52
Superior frontal 2.50 (0.23) 2.41 (0.20) 0.09 3.8 0.44
Inferior frontal 2.10 (0.14) 2.02 (0.21) 0.07 3.4 0.40
*p ⬍ 0.005.
†p ⬍ 0.05.
1052 Neurology 72 March 24, 2009
fied focal ventromedial temporal atrophy.
There are no
previous comparable data regarding the magnitude of
atrophy (% difference) between progressors vs nonpro-
gressors. The present data are consistent with previous
data indicating that, when AD symptoms are still incip-
ient or very mild, atrophy is already present in a set of
heteromodal and limbic cortical regions. These data
build on previous results by showing that although a
lesser magnitude of atrophy is present in these very
mildly affected patients, the spatial pattern is predict-
able based on what is known about the cortical regions
affected in mild AD dementia (when defined using an
Besides comparisons between diagnostic patient
groups, we also investigated the relationships be-
tween regional cortical thickness and relative severity
of mild symptoms in daily life (as measured by CDR-
SB) and signs of cognitive impairment (as measured
by MMSE) within this group of participants. To our
knowledge, only one study has investigated the rela-
tionship of MMSE to cortical atrophy within a sam-
ple restricted to MCI or prodromal AD, and no
regions of correlation were found.
We observed ro-
bust relationships between the average thickness of
the nine AD-signature regions and MMSE perfor-
mance as well as CDR-SB. Such similar relation-
ships, while intuitive, are not inevitable given the
frequent dissociation between symptoms in daily life
and signs on cognitive testing, particularly in the
mildest prodromal stages of the illness.
tions were anatomically distinct, however, in that
CDR-SB correlates most prominently with ventro-
medial temporal and superior parietal thickness
whereas MMSE correlates with superior frontal and
lateral parietal thickness. Overall, the strength of the
correlations was relatively subtle, as has been the case
in previous reports focusing on medial temporal lobe
regions. Further investigation will be necessary to de-
termine whether consistent localized brain–behavior
relationships can be illuminated through the investi-
Figure 2 Receiver operating characteristic curves of the performance by various morphometric measures for classifying questionable
Alzheimer dementia (QAD) progressors vs QAD nonprogressors
Optimal performance maximizing both sensitivity and specificity was determined by identifying the curve with the highest leftward point (white region of
top graphs); bottom graphs show zoomed in views of areas in white from top graphs. (A) Comparing three standard volumetric variables, EV demonstrated
highest sensitivity (72%) and specificity (65%). EV ⫽ entorhinal volume; HV ⫽ hippocampal volume; WBV ⫽ whole brain volume. (B) Comparing two global
variables, mean cortical thickness demonstrated highest sensitivity (69%) and specificity (65%). MCT ⫽ mean cortical thickness. (C) Comparing two
regional cortical thickness measures with EV, mean AD signature cortical thickness demonstrated highest sensitivity (83%) and specificity (65%). ADT ⫽
AD signature thickness; MTLT ⫽ medial temporal lobe thickness.
Neurology 72 March 24, 2009 1053
gation of regional cortical thinning in neurodegen-
Most importantly, measures of the thickness of
specific cortical regions known to be affected by AD
are useful in predicting progression from question-
able dementia to mild AD dementia within a few
years. These findings build on a large body of work
focused on morphometric measures of the medial
temporal lobe in predicting progression from incipi-
ent dementia/MCI to AD dementia, expanding now
to include regions of the temporoparietal and frontal
cortices. The use of ROC analysis illuminates the
utility of these measures across a broad spectrum of
sensitivity-specificity tradeoffs, allowing the clinician
to choose cutpoints depending on the question at
hand. Also, by comparing the novel cortical thickness
methods to more widely used volumetric measures
(the results of which are similar to those of previous
), these analyses further demonstrate the
potential utility of thickness measures.
Of the global measures, the average thickness of
the cerebral cortex outperforms whole brain volume
substantially with respect to predicting progression.
This is not surprising given that whole brain volume
likely reflects large classes of tissue that are relatively
unaffected by AD. Of the regional volume measures,
entorhinal volume performed better than hippocam-
pal volume, as has been seen previously.
thickness performed better yet in comparison to en-
torhinal volume, which may relate in part to past
observations that thickness appears to be a purer re-
flection of the effects of AD on cortical morphology
while surface area, and therefore volume, are altered
substantially by aging.
Finally, ROC analysis dem-
onstrated that the average thickness of all nine AD-
signature ROIs was the most powerful measure for
predicting progression to mild dementia. Compared
to MTL ROI measures, such a disease composite
measure probably increases sensitivity by including
individuals with thinning in additional brain regions
involved in the disease and increases specificity by
excluding individuals with relative preservation of
thickness in those regions (whose relatively isolated
MTL thinning may be due to other pathologies).
Several limitations of this study deserve mention.
The diagnosis of AD was not confirmed by autopsy
or any other imaging markers. Longitudinal MRI
measures were not performed. Finally, to translate
these measures into biomarkers for use in individual
patient diagnostic classification, as has been done
with other approaches to cortical morphometric
analysis in AD,
further work is necessary. Yet the
robust effects observed provide optimism that this
approach to cortical anatomic measurement may be
able to achieve such goals.
Dr. Dickerson and Mr. Bakkour performed all statistical analyses.
The authors thank Dr. Randy Buckner for his insight as well as the faculty
and staff of the Washington University Alzheimer’s Disease Research
Center. They also thank the participants in this study and their families
for their valuable contributions, without which this research would not
have been possible.
Received June 18, 2008. Accepted in final form October 3, 2008.
1. Jack CR Jr., Petersen RC, Xu YC, et al. Prediction of AD
with MRI-based hippocampal volume in mild cognitive
impairment. Neurology 1999;52:1397–1403.
2. Killiany RJ, Gomez-Isla T, Moss M, et al. Use of structural
magnetic resonance imaging to predict who will get Alz-
heimer’s disease. Ann Neurol 2000;47:430–439.
3. Dickerson BC, Goncharova I, Sullivan MP, et al. MRI-
derived entorhinal and hippocampal atrophy in incipient
Figure 3 Magnitude of cortical thinning in prodromal Alzheimer disease (AD)
in millimeters, derived from sample of 29 questionable Alzheimer
dementia (QAD) nonprogressor vs 20 progressor subjects
Map shows parameter estimate of amount of thinning across cerebral cortex from general
linear model analysis of 49 subjects, showing areas where cortex is at least 0.1 mm thinner
in prodromal AD dementia group (progressors) than in nonprogressors. Maps are presented
on the semi-inflated cortical surface of an average brain with dark gray regions represent-
ing sulci and light gray regions representing gyri. Non-neocortical regions and regions that
are not part of the cortical mantle (such as the corpus callosum and thalamus) have been
excluded from the analysis. Color scale shows magnitude of thinning from 0.1 mm (red)
through 0.2 mm (yellow).
1054 Neurology 72 March 24, 2009
and very mild Alzheimer’s disease. Neurobiol Aging 2001;
4. Braak H, Rub U, Schultz C, Del Tredici K. Vulnerability
of cortical neurons to Alzheimer’s and Parkinson’s diseases.
J Alzheimer Dis 2006;9:35–44.
5. Scheff SW, Price DA. Alzheimer’s disease-related alter-
ations in synaptic density: neocortex and hippocampus.
J Alzheimer Dis 2006;9:101–115.
6. Singh V, Chertkow H, Lerch JP, Evans AC, Dorr AE,
Kabani NJ. Spatial patterns of cortical thinning in mild
cognitive impairment and Alzheimer’s disease. Brain 2006;
7. Whitwell JL, Shiung MM, Przybelski SA, et al. MRI patterns
of atrophy associated with progression to AD in amnestic
mild cognitive impairment. Neurology 2008;70:512–520.
8. Chetelat G, Landeau B, Eustache F, et al. Using voxel-
based morphometry to map the structural changes associ-
ated with rapid conversion in MCI: a longitudinal MRI
study. Neuroimage 2005;27:934–946.
9. Bozzali M, Filippi M, Magnani G, et al. The contribution
of voxel-based morphometry in staging patients with mild
cognitive impairment. Neurology 2006;67:453–460.
10. Morris JC, Ernesto C, Schafer K, et al. Clinical dementia
rating training and reliability in multicenter studies: the
Alzheimer’s Disease Cooperative Study experience. Neu-
11. Dickerson BC, Bakkour A, Salat DH, et al. The cortical
signature of Alzheimer’s disease: regionally specific cortical
thinning relates to symptom severity in very mild to mild
AD dementia and is detectable in asymptomatic amyloid-
positive individuals. Cereb Cortex Epub 2008 Jul 16.
12. Salat DH, Buckner RL, Snyder AZ, et al. Thinning of the
cerebral cortex in aging. Cereb Cortex 2004;14:721–730.
13. Buckner RL, Head D, Parker J, et al. A unified approach
for morphometric and functional data analysis in young,
old, and demented adults using automated atlas-based
head size normalization: reliability and validation against
manual measurement of total intracranial volume. Neuro-
14. Buckner RL, Snyder AZ, Shannon BJ, et al. Molecular, struc-
tural, and functional characterization of Alzheimer’s disease:
evidence for a relationship between default activity, amyloid,
and memory. J Neurosci 2005;25:7709–7717.
15. Fotenos AF, Snyder AZ, Girton LE, Morris JC, Buckner
RL. Normative estimates of cross-sectional and longitudi-
nal brain volume decline in aging and AD. Neurology
16. Storandt M, Grant EA, Miller JP, Morris JC. Longitudinal
course and neuropathologic outcomes in original vs revised
MCI and in pre-MCI. Neurology 2006;67:467–473.
17. Kuperberg GR, Broome MR, McGuire PK, et al. Region-
ally localized thinning of the cerebral cortex in schizophre-
nia. Arch Gen Psychiatry 2003;60:878–888.
18. Rosas HD, Liu AK, Hersch S, et al. Regional and progres-
sive thinning of the cortical ribbon in Huntington’s dis-
ease. Neurology 2002;58:695–701.
19. Fischl B, Sereno MI, Dale AM. Cortical surface-based
analysis. II: Inflation, flattening, and a surface-based coor-
dinate system. Neuroimage 1999;9:195–207.
20. Dale AM, Fischl B, Sereno MI. Cortical surface-based
analysis. I. Segmentation and surface reconstruction. Neu-
21. Fischl B, Dale AM. Measuring the thickness of the human
cerebral cortex from magnetic resonance images. Proc Natl
Acad Sci USA 2000;97:11050–11055.
22. Fischl B, Salat DH, Busa E, et al. Whole brain segmenta-
tion: automated labeling of neuroanatomical structures in
the human brain. Neuron 2002;33:341–355.
23. Folstein MF, Folstein SE, McHugh PR. “Mini-mental
state”: a practical method for grading the cognitive state of
patients for the clinician. J Psychiatr Res 1975;12:189–
24. de Leon MJ, Golomb J, George AE, et al. The radio-
logic prediction of Alzheimer disease: the atrophic hip-
pocampal formation. AJNR Am J Neuroradiol 1993;14:
25. Pennanen C, Testa C, Laakso MP, et al. A voxel based
morphometry study on mild cognitive impairment. J Neu-
rol Neurosurg Psychiatry 2005;76:11–14.
26. Apostolova LG, Steiner CA, Akopyan GG, et al. Three-
dimensional gray matter atrophy mapping in mild cogni-
tive impairment and mild Alzheimer disease. Arch Neurol
27. Whitwell JL, Shiung MM, Przybelski SA, et al. MRI pat-
terns of atrophy associated with progression to AD in am-
nestic mild cognitive impairment. Neurology 2008;70:
28. Dickerson BC, Sperling RA, Hyman BT, Albert MS,
Blacker D. Clinical prediction of Alzheimer disease de-
mentia across the spectrum of mild cognitive impairment.
Arch Gen Psychiatry 2007;64:1443–1450.
29. Devanand DP, Pradhaban G, Liu X, et al. Hippocampal
and entorhinal atrophy in mild cognitive impairment: pre-
diction of Alzheimer disease. Neurology 2007;68:828–
30. Killiany RJ, Hyman BT, Gomez-Isla T, et al. MRI mea-
sures of entorhinal cortex vs hippocampus in preclinical
AD. Neurology 2002;58:1188–1196.
31. Dickerson BC, Feczko E, Augustinack JC, et al. Differen-
tial effects of aging and Alzheimer’s disease on medial tem-
poral lobe cortical thickness and surface area. Neurobiol
Aging Epub 2007 Sep 13.
32. Vemuri P, Gunter JL, Senjem ML, et al. Alzheimer’s dis-
ease diagnosis in individual subjects using structural MR
images: validation studies. Neuroimage 2008;39:1186–
33. Lerch JP, Pruessner J, Zijdenbos AP, et al. Automated cor-
tical thickness measurements from MRI can accurately
separate Alzheimer’s patients from normal elderly controls.
Neurobiol Aging 2008;29:23–30.
Neurology 72 March 24, 2009 1055