Deformation-Based Morphometry Reveals Brain Atrophy in Frontotemporal Dementia

ORIGINAL CONTRIBUTION
Deformation-Based Morphometry Reveals Brain
Atrophy in Frontotemporal Dementia
Valerie A. Cardenas, PhD; Adam L. Boxer, MD, PhD; Linda L. Chao, PhD; Maria L. Gorno-Tempini, MD, PhD;
Bruce L. Miller, MD; Michael W. Weiner, MD; Colin Studholme, PhD
Objective: To compare deformation-based maps of local anatomical size between subjects with frontotemporal dementia (FTD) and healthy subjects to identify regions of the brain involved in FTD.
Design: Structural magnetic resonance images were ob-
tained from 22 subjects with FTD and 22 cognitively normal, age-matched controls. We applied deformationbased morphometry and compared anatomy between
groups using an analysis of covariance model that
included a categorical variable denoting group membership and covaried for head size.
Setting: University of California, San Francisco, Memory
and Aging Center, and the San Francisco Veterans Affairs Medical Center.
Patients: Twenty-two subjects with FTD and 22 cognitively normal, age-matched controls.
Interventions: Neurological, neuropsychological, and
functional evaluations and magnetic resonance imaging.
F
Author Affiliations: Magnetic
Resonance Unit, San Francisco
Veterans Affairs Medical Center
(Drs Cardenas, Chao, Weiner,
and Studholme), and
Departments of Radiology
(Drs Cardenas, Chao, Weiner,
and Studholme) and Psychiatry
(Dr Chao) and the Memory and
Aging Center, Department of
Neurology (Drs Boxer,
Gorno-Tempini, and Miller),
University of California,
San Francisco.
Main Outcome Measure: Deformation maps of local
anatomical size.
Results: Patients with FTD showed extensive, signifi-
cant atrophy of the frontal lobes, affecting both gray matter and white matter. Atrophy of similar magnitude but
less significance was observed in the anterior temporal
lobes. The subcortical and midbrain regions, particularly the thalamus, pons, and superior and inferior colliculi, showed strongly significant atrophy of smaller magnitude.
Conclusions: We confirmed frontal and anterior temporal gray matter atrophy in FTD. The observed white
matter loss, thalamic involvement, and midbrain atrophy are consistent with pathological findings in latestage FTD. Dysfunction of ventral-frontal-brainstem circuitry may underlie some of the unique clinical features
of FTD.
Arch Neurol. 2007;64:873-877
RONTOTEMPORAL DEMENTIA
(FTD) is a clinical subtype of
frontotemporal lobar degeneration defined by deficits in
social and personal conduct. Postmortem studies have suggested that brain atrophy in FTD begins
in the frontal lobe, extending into the anterior temporal lobes, basal ganglia, and
the thalamus. White matter (WM), including the corpus callosum, is prominently affected.1
In vivo magnetic resonance imaging
(MRI) structural analyses have compared patients with FTD with controls
using conventional measures of gray matter (GM), WM, or cerebrospinal fluid volumes obtained from computer segmentation and volumes of manually delineated
regions of interest, such as the hippocampus. Using such measurements, atrophy
of the frontal and temporal lobes2 and cor-
(REPRINTED) ARCH NEUROL / VOL 64, JUNE 2007
873
pus callosum3 has been demonstrated in
FTD. Hippocampal atrophy is present in
FTD as compared with controls, but it is
not as severe as in Alzheimer disease.4
Unlike region of interest methods, voxelwise structural image analysis assesses
anatomical variation without prior hypotheses about the location and extent of
the anatomical variation. Early techniques, such as voxel-based morphometry,5 provided regional indications of GM
loss in the frontal and temporal regions.6
Because voxel-based morphometry relies
on the automated segmentation of images into GM, WM, and cerebrospinal
fluid, regions of abnormal WM (prevalent in older populations) may be incorrectly classified as GM by automatic
segmentation. In addition, automatic segmentation of subcortical structures can be
problematic because of the mixing of GM
and WM in these structures. For these
WWW.ARCHNEUROL.COM
©2007 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ on 10/21/2014
reasons, voxel-based morphometry is suboptimal for investigating WM loss or determining subcortical involvement in FTD.
Improvements in image alignment allow purely deformation-based morphometry (DBM) to be used to provide more direct quantitative maps of anatomical variation.7 By avoiding the need for image segmentation and
using robust registration methods, DBM may be more suitable for investigating anatomical variation of WM and
subcortical structures.
In this study, our goal was to compare deformationbased maps of local anatomical size between subjects with
FTD and healthy subjects to identify regions of the brain
involved in FTD. We hypothesized that subjects with FTD
would show atrophy in frontal and temporal GM and WM
and subcortical structures, including the basal ganglia and
thalamus.
then analyzed using statistical parametric mapping. Using a general linear model, we compared the FTD and CN groups with
a categorical variable coding group membership and head size
(defined as the average Jacobian determinant within the intracranial vault delineated on the reference anatomy) included as
a covariate, as shown in the equation, where |J(x)| is the Jacobian evaluated at voxel x, and a(x), b(x), and the intercept c(x)
are estimated at each voxel. Because statistics were computed
independently at each voxel, we calculated the corrected P⬍.05
peak threshold using permutation testing,12 where we permuted the group membership 1000 times and recalculated the
voxel statistics to build the null distribution, and also the method
of Bonferroni13 (using all voxels within the average brain). We
used fMRIstat14 to identify in FTD clusters of contracting or
expanding voxels, where both magnitude (all voxels within the
cluster must be significant at P⬍.001 uncorrected) and spatial
extent (larger clusters are less likely to be false positives) were
jointly used to assess corrected significance.
RESULTS
METHODS
PARTICIPANTS
All subjects underwent neurological, neuropsychological, and
functional evaluations at the University of California, San Francisco, Memory and Aging Center. Cognitively normal (CN) controls (mean±SD age, 63±7 years; n=22; 7 women) had cognitive test scores within the normal age and education-adjusted
range. All subjects with FTD (mean±SD age, 63±6 years; n=22;
7 women; mean ± SD 6.6 ± 3.7 years since disease onset) met
Neary8 criteria for FTD. In addition, 5 subjects with FTD also
met El Escorial criteria9 for possible or probable amyotrophic
lateral sclerosis. Diagnoses were blinded to neuroimaging results to avoid confounding future neuroimaging analyses.6 Pathological verification of diagnosis was obtained in 5 subjects (2,
Pick disease; 2, FTD-ubiquitin; 1, FTD–motor neuron disease). The Clinical Dementia Rating (CDR) scaled and sum of
boxes scores were 0 for both measures for CN controls and
mean±SD 1.12±0.69 (scaled) and 6.7±3.8 (sum) for the FTD
group. The mean±SD Mini-Mental State Examination score for
the FTD group was 23.1±7.0 and for the CN controls, 29.3±2.2.
The study received institutional review board approval and all
participants gave informed consent before the study.
MAGNETIC RESONANCE IMAGING
The MRI data were acquired at the San Francisco Veterans
Affairs Medical Center on a clinical 1.5-T MRI scanner (Vision;
Siemens Medical Systems, Iselin, NJ). Coronal T1-weighted
images were acquired using a magnetization-prepared rapidacquisition gradient-echo sequence (repetition time, 9 milliseconds; inversion time, 300 milliseconds; echo time, 4 milliseconds; 1⫻1 mm2 in-plane resolution; 1.5-mm slabs); images were
acquired orthogonal to the long axis of the hippocampus.
FULLY AUTOMATED DBM
A B-Spline free-form deformation algorithm driven by normalized mutual information10 was used to register individual scans
to a 72-year-old female reference atlas, chosen to retain the finest anatomical structures for accurate registration.7 The Jacobian determinant of this transformation at each point, giving
the pattern of volume change required to force the individual
anatomy to conform to the reference, was mapped and smoothed
using an intensity-consistent filtering approach.11 These voxel
maps of relative local anatomical size of each individual were
(REPRINTED) ARCH NEUROL / VOL 64, JUNE 2007
874
DEFORMATION-BASED MORPHOMETRY
Figure 1 shows regions where patients with FTD dem-
onstrate brain volume reductions compared with CN controls. Regions where patients with FTD show significant (P⬍.01 uncorrected, or T=2.70) atrophy are overlaid
on the average spatially normalized MRI. The threshold
at P=.05 corrected for multiple comparisons using permutation testing is T=5.0 and the corrected threshold
using the method of Bonferroni is T=6.62; these 2 thresholds are marked on the Figure 1 color bar (as PT P=.05
and BF [Bonferroni over brain voxels] P=.05, respectively). Figure 1 reveals many voxels where patients with
FTD show significant atrophy relative to CN controls even
after correction for multiple comparisons, including a large
region of the pons and midbrain, the right superior and
inferior colliculus, thalamus, left superior frontal GM, anterior frontal WM, and a ventromedial frontal WM region. At lower significance, patients with FTD showed
extensive atrophy of frontal and anterior temporal WM
and GM.
Cluster analysis using fMRIstat revealed a single connected cluster of contraction in patients with FTD vs CN
controls encompassing all the regions mentioned earlier. Figure 2 shows the T statistic map overlaid on the
average spatially normalized MRI, where voxels belonging to this significant cluster of contraction are outlined
in red. Negative T values that indicate atrophy in patients with FTD compared with CN controls are in green
and blue; positive T values that indicate expansion in patients with FTD relative to CN controls (all located within
cerebrospinal fluid) are in yellow and red.
To estimate the magnitude of atrophy in patients with
FTD, the brainstem (including midbrain), thalamus, and
ventromedial frontal lobe were defined anatomically on
the average image, and we then averaged the estimates
a(x) from the equation at all statistically significant voxels x (t⬎5) within each region of interest. We found that
tissue volume was reduced by 10% in the brainstem region in patients with FTD compared with CN controls,
by 26% in the thalamic region, and by 34% in the ventromedial frontal region. Although no single voxel in the
WWW.ARCHNEUROL.COM
©2007 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ on 10/21/2014
UC P =.01
UC P =.001
UC P =.0001
PT P =.05
PT P =.001
BF P =.05
PT P =.01
T = 2.70
T = 6.62
Figure 1. Coronal sections moving from anterior to posterior showing the pattern of the T statistic of spatially consistent contractions related to frontotemporal
dementia diagnosis. Color scales are marked with significance levels. BF indicates Bonferroni over brain voxels; PT, permutation testing; and UC, uncorrected.
temporal lobes reached significance after correction for
multiple comparisons, the regions of the temporal lobes
were part of the significant cluster of contraction. To determine whether the relatively small T statistics in the
anterior temporal lobe were due to a small volume difference between groups or due to variability within groups,
we averaged the estimates a(x) from the equation at all
voxels x in the anterior temporal lobe that belonged to
the significant cluster of contraction. This showed that
tissue volume was reduced by 35% in the anterior temporal region in patients with FTD compared with CN controls, as large a reduction as observed in the frontal lobes.
To validate DBM, we calculated region of interest measures on 22 consecutive subjects (11 with FTD and 11 CN
controls), measuring frontal and temporal lobe volumes and
brainstem volume (midbrain, pons, and medulla) on a single
midsagittal slice.15 Volumes were then normalized to account for global head size differences and expressed as a
percentage of intracranial volume, as shown in the Table.
Similar to DBM, we found significant volume differences
between patients with FTD and CN controls within the frontal lobe and brainstem and no significant difference within
the temporal lobe. The magnitude of the reduction within
the frontal and temporal lobes was smaller, probably because these regions of interest included the entire frontal
or temporal lobe, whereas our DBM estimates were within
only significant subregions.
(REPRINTED) ARCH NEUROL / VOL 64, JUNE 2007
875
COMMENT
We used DBM to examine brain structure differences between patients with FTD and CN controls. The major findings of this study are (1) patients with FTD showed significant atrophy in the frontal lobes, affecting both WM
and GM, (2) significant atrophy was observed within the
thalamus and adjacent WM in FTD, (3) the brainstem,
including the midbrain and pontine tegmentum as well
as the superior and inferior colliculi, showed significant
tissue volume reduction in FTD, and (4) regions of the
anterior temporal lobes were atrophied in FTD.
The frontal and anterior temporal GM atrophy observed in FTD is consistent with previous postmortem and
voxel-based morphometry studies. The frontal and temporal volumes were comparably reduced, although the reduction in the temporal lobes was not as significant. A quantitative validation7 showed excellent agreement between
our deformation-derived (reference image was the same
72-year-old subject as in this study) and manually delineated temporal volumes, so it is unlikely that this lower
significance in the temporal lobe arises because of poor
alignment. Instead, we believe that some patients with FTD
have much smaller temporal lobes than CN individuals
but that in others the FTD disease does not involve or has
not progressed to the stage where the temporal lobes are
WWW.ARCHNEUROL.COM
©2007 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ on 10/21/2014
Corrected P = .001
FTD < CN
Corrected P = .001
FTD > CN
Uncorrected
P = .01
Figure 2. The T statistic map is overlaid on the spatially normalized average magnetic resonance image (N = 44). The red contour encompasses the contraction
voxels that were contained in the large cluster that included regions of the frontal and temporal lobes, subcortical structures, and brainstem. FTD indicates
frontotemporal dementia; CN, cognitively normal.
Table. Region of Interest Volumes Expressed as a Percentage of Intracranial Volume
Intracranial Volume, Mean ± SD, %
Frontal lobe
Temporal lobe
Brainstem midsagittal
Cognitively Normal
Controls
Patients With
Frontotemporal Dementia
Reduction, %
P
Value
34.5 ± 1.0
16.3 ± 1.0
0.086 ± 7.65E-05*
31.9 ± 2.27
16.3 ± 1.0
0.076 ± 7.46E-05†
7.5
0
11.6
.003
.85
.006
*Expression of scientific notation. Alternatively, 7.65E-05 can be expressed as 7.65 ⫻ 10−6 or 0.0000765.
†Expression of scientific notation. Alternatively, 7.46E-05 can be expressed as 7.46 ⫻ 10−6 or 0.0000746.
greatly affected. Such an inconsistent spatial pattern
within the FTD group would explain the lowered significance of atrophy in the anterior temporal region, and this
interpretation is highly consistent with the considerable
variability in clinical and neuroimaging features observed in FTD.16 Consistent with other reports, we did
not observe any significant atrophy in parietal or occipital lobes in the patients with FTD.
In addition to frontal and temporal GM tissue loss,
we observed frontal WM and thalamic atrophy in
patients with FTD compared with CN controls. Thalamic volume loss of up to 37% has previously been
described in neuropathological studies of FTD.17 In a
proposed scheme for staging pathological disease severity in FTD, WM and thalamic atrophy are thought to
occur at later stages, only after the frontal and temporal
lobes are severely affected,1 roughly corresponding to
CDR scores of 3 to 5. Our data suggest that these
changes are measurable in patients with FTD at earlier
stages of disease (mean±SD CDR score 1.2±0.68). Our
(REPRINTED) ARCH NEUROL / VOL 64, JUNE 2007
876
study also may have had greater sensitivity to these
changes because of a more homogeneous clinical
sample (all had FTD) and a smaller interval between
CDR and brain atrophy measurement, as well as the
more quantitative nature of the DBM analysis over
visual atrophy measurements.
Although midbrain atrophy has not previously been
emphasized in imaging studies in FTD, FTD-related
pathological features are frequently found in the substantia nigra and other brainstem structures,18 and this
finding supports the known clinical overlap between FTD
and progressive supranuclear palsy (PSP). Neuroimaging and pathological features of PSP show atrophy in the
midbrain, basal ganglia, and other structures.19-21 Cases
of clinically diagnosed FTD have been found to have PSP
pathological features at autopsy,22 and clinically diagnosed PSP cases have been described with FTDubiquitin pathological features involving the midbrain.23 Recent evidence suggests that patients with FTD
have measurable saccade abnormalities, which may in part
WWW.ARCHNEUROL.COM
©2007 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ on 10/21/2014
reflect damage to the superior colliculus and/or WM tracts
that connect it to the frontal lobes and basal ganglia,24
consistent with our results. Midbrain and thalamic pathological features may underlie some of the deficits in social function and emotion perception identified in FTD25
since integrity of this region is likely to be necessary for
function of a midbrain-thalamus-amygdala pathway implicated in emotion perception.26 Finally, there is an extensive literature on midbrain involvement in FTD at a
pathology level. For example, in the original study by
Knopman et al27 on dementia lacking distinctive histological features, the authors noted that 79% of these patients had pathological features in the midbrain. In previous voxel-based morphometry work from our center,
we also found atrophy in the midbrain, and DBM is a better technique for delineating these changes.
These cases were defined on the basis of their clinical
features and only a small number have been autopsy confirmed. Because FTD is a pathologically heterogeneous
disorder,28 it will be of particular interest to confirm these
results in histopathologically diagnosed FTD and to assess whether subcortical and brainstem atrophy is associated with specific biochemical FTD phenotypes, such
as tau protein inclusions, which might further strengthen
the link between FTD and PSP.
Taken together, our findings suggest that dysfunction of a frontal-subcortical-brainstem circuit may underlie some of the unique clinical features of FTD and
that DBM measures of volume may be useful for exploring these brain-behavior relationships.
Accepted for Publication: September 5, 2006.
Correspondence: Valerie A. Cardenas, PhD, University
of California, San Francisco, Department of Veterans Affairs Medical Center, 4150 Clement St, 114M, San Francisco, CA 94121 ([email protected]).
Author Contributions: All authors had full access to all
of the data in the study and take responsibility for the
integrity of the data and the accuracy of the data analysis. Study concept and design: Cardenas, Boxer, Chao,
Gorno-Tempini, and Weiner. Acquisition of data: GornoTempini. Analysis and interpretation of data: Cardenas,
Boxer, Chao, Gorno-Tempini, Miller, Weiner, and Studholme. Drafting of the manuscript: Cardenas and Boxer.
Critical revision of the manuscript for important intellectual content: Cardenas, Boxer, Chao, Gorno-Tempini,
Miller, Weiner, and Studholme. Statistical analysis: Cardenas and Gorno-Tempini. Obtained funding: Miller,
Weiner, and Studholme. Administrative, technical, and material support: Cardenas, Boxer, Chao, Weiner, and Studholme. Study supervision: Boxer, Chao, Gorno-Tempini,
Weiner, and Studholme.
Financial Disclosure: None reported.
Funding/Support: This work was supported in part by
National Institutes of Health grants P01 AG19724, R01
AG10897, and R01 MH65392.
Acknowledgment: We would like to acknowledge Diana Truran, BA, and Erin Clevenger, BA, for magnetic
resonance imaging, image quality control, and data organization, and Joanna Hellmuth, BS, for database assistance.
(REPRINTED) ARCH NEUROL / VOL 64, JUNE 2007
877
REFERENCES
1. Broe M, Hodges JR, Schofield E, Shepherd CE, Kril JJ, Halliday GM. Staging disease severity in pathologically confirmed cases of frontotemporal dementia.
Neurology. 2003;60:1005-1011.
2. Boccardi M, Laakso MP, Bresciani L, et al. The MRI pattern of frontal and temporal brain atrophy in fronto-temporal dementia. Neurobiol Aging. 2003;24:
95-103.
3. Kaufer DI, Miller BL, Itti L, et al. Midline cerebral morphometry distinguishes frontotemporal dementia and Alzheimer’s disease. Neurology. 1997;48:978-985.
4. Frisoni GB, Laakso MP, Beltramello A, et al. Hippocampal and entorhinal cortex
atrophy in frontotemporal dementia and Alzheimer’s disease. Neurology. 1999;
52:91-100.
5. Ashburner J, Friston KJ. Voxel-based morphometry—the methods. Neuroimage.
2000;11:805-821.
6. Rosen HJ, Gorno-Tempini ML, Goldman WP, et al. Patterns of brain atrophy in frontotemporal dementia and semantic dementia. Neurology. 2002;58:198-208.
7. Studholme C, Cardenas V, Blumenfeld R, et al. Deformation tensor morphometry of semantic dementia with quantitative validation. Neuroimage. 2004;21:
1387-1398.
8. Neary D, Snowden JS, Gustafson L, et al. Frontotemporal lobar degeneration: a
consensus on clinical diagnostic criteria. Neurology. 1998;51:1546-1554.
9. Brooks BR. El Escorial World Federation of Neurology criteria for the diagnosis
of amyotrophic lateral scleroris. J Neurol Sci. 1994;124(suppl):96-107.
10. Studholme C, Novotny E, Zubal IG, Duncan JS. Estimating tissue deformation
between functional images induced by intracranial electrode implantation using
anatomical MRI. Neuroimage. 2001;13:561-576.
11. Studholme C, Cardenas V, Maudsley A, Weiner M. An intensity consistent filtering approach to the analysis of deformation tensor derived maps of brain shape.
Neuroimage. 2003;19:1638-1649.
12. Nichols TE, Holmes AP. Nonparametric permutation tests for functional neuroimaging: a primer with examples. Hum Brain Mapp. 2002;15:1-25.
13. Glantz S. Primer of Biostatistics. 4th ed. New York, NY: McGraw-Hill; 1981.
14. Worsley KJ, Liao CH, Aston J, et al. A general statistical analysis for fMRI data.
Neuroimage. 2002;15:1-15.
15. Bloomer CW, Langleben DD, Meyerhoff DJ. Magnetic resonance detects brainstem changes in chronic, active heavy drinkers. Psychiatry Res. 2004;132:
209-218.
16. Boxer AL, Trojanowski JQ, Lee VY-M, Miller MJ. Frontotemporal lobar degeneration.
In: Beal MF, Lang AE, Ludolph AC, eds. Neurodegenerative Diseases: Neurobiology, Pathogenesis and Therapeutics. Cambridge, England: Cambridge University Press; 2005.
17. Mann DM, South PW. The topographic distribution of brain atrophy in frontal
lobe dementia. Acta Neuropathol (Berl). 1993;85:334-340.
18. Dickson DW. Neuropathology of Pick’s disease. Neurology. 2001;56(suppl 4):
S16-S20.
19. Litvan I, Hauw JJ, Bartko JJ, et al. Validity and reliability of the preliminary NINDS
neuropathologic criteria for progressive supranuclear palsy and related disorders.
J Neuropathol Exp Neurol. 1996;55:97-105.
20. Verny M, Jellinger KA, Hauw JJ, Bancher C, Litvan I, Agid Y. Progressive supranuclear palsy: a clinicopathological study of 21 cases. Acta Neuropathol (Berl).
1996;91:427-431.
21. Boxer AL, Geschwind MD, Belfor N, et al. Patterns of brain atrophy that differentiate corticobasal degeneration syndrome from progressive supranuclear palsy.
Arch Neurol. 2006;63:81-86.
22. Rippon GA, Boeve BF, Parisi JE, et al. Late-onset frontotemporal dementia associated with progressive supranuclear palsy/argyrophilic grain disease/
Alzheimer’s disease pathology. Neurocase. 2005;11:204-211.
23. Paviour DC, Lees AJ, Josephs KA, et al. Frontotemporal lobar degeneration with
ubiquitin-only-immunoreactive neuronal changes: broadening the clinical picture to include progressive supranuclear palsy. Brain. 2004;127:2441-2451.
24. Meyniel C, Rivaud-Pechoux S, Damier P, Gaymard B. Saccade impairments in
patients with fronto-temporal dementia. J Neurol Neurosurg Psychiatry. 2005;
76:1581-1584.
25. Rankin KP, Kramer JH, Mychack P, Miller BL. Double dissociation of social functioning in frontotemporal dementia. Neurology. 2003;60:266-271.
26. Liddell BJ, Brown KJ, Kemp AH, et al. A direct brainstem-amygdala-cortical ‘alarm’
system for subliminal signals of fear. Neuroimage. 2005;24:235-243.
27. Knopman DS, Mastri AR, Frey WH II, Sung JH, Rustan T. Dementia lacking distinctive histologic features: a common non-Alzheimer degenerative dementia.
Neurology. 1990;40:251-256.
28. Josephs KA, Petersen RC, Knopman DS, et al. Clinicopathologic analysis of frontotemporal and corticobasal degenerations and PSP. Neurology. 2006;66:4148.
WWW.ARCHNEUROL.COM
©2007 American Medical Association. All rights reserved.
Downloaded From: http://jama.jamanetwork.com/ on 10/21/2014