Keywords
neurodegeneration - neuroimaging - dementia - magnetic resonance imaging (MRI) - functional
magnetic resonance imaging (fMRI) - functional connectivity magnetic resonance imaging
(fcMRI) - diffusion tensor imaging (DTI) - [
18F] fluorodeoxyglucose positron emission tomography (FDG-PET) - amyloid positron emission
tomography - dopamine transporter (DAT) imaging - Alzheimer's disease - dementia with
Lewy bodies - frontotemporal dementia
Rapid advances in our understanding of neurodegenerative dementias in recent years
have translated into a change in the clinical approach to a patient presenting with
changes in cognition. Whereas previous practice frequently entailed “excluding treatable
causes” and lumping the remaining cases together as presenile dementia, senile dementia,
or possibly Alzheimer's disease or multiinfarct dementia, dementia specialists in
the present day actively endeavor to predict a patient's underlying neuropathology
utilizing history, examination, and biomarker data. Advances in neuroimaging techniques,
particularly magnetic resonance imaging (MRI), have contributed immensely not only
to our understanding of clinical/pathologic relationships in the research setting,
but also to our ability to arrive at clinical diagnoses in daily practice.
Most neurodegenerative dementias are characterized by a clinical course of insidious
onset and gradual progression of symptoms over months to years. Thought previously
to engender global cognitive dysfunction even in the early stages of illness,[1] neurodegenerative processes are now recognized to produce focal cognitive dysfunction
initially, corresponding to a circumscribed initial distribution of pathology. Localization
of cognitive dysfunction via history and examination and localization of neuronal
dysfunction and/or neurodegeneration with neuroimaging thus empower clinicians to
distinguish between unique clinical syndromes with characteristic early anatomic signatures
([Table 1]).
Table 1
Neurodegenerative Dementia Clinical Syndromes
|
Syndrome
|
Pathologies
|
Key Clinical Features
|
Anatomy
|
|
Amnestic AD
|
AD ± LB
|
↓ Episodic memory
|
Medial temporal, temporoparietal
|
|
Posterior cortical atrophy
|
AD >> tau, others
|
Visuospatial, somatospatial dysfunction
|
Parietal, occipital
|
|
Logopenic PPA
|
AD > TDP, tau
|
↓ Word retrieval, sentence repetition
|
Left temporoparietal
|
|
Dysexecutive AD
|
N/A[a]
|
↓ Complex attention, EF
|
Frontal, temporoparietal
|
|
DLB/PDD
|
LB ± AD
|
↓ EF, VF; park., VH, fluctuations
|
Midbrain, hypothalamus, SI
|
|
bvFTD
|
TDP ∼ tau >> FUS
|
Disinhibition, apathy, ↓ empathy, perseveration, Δ diet, ↓ EF
|
Frontal, temporal, insula, anterior cingulate
|
|
Agrammatic PPA
|
tau > TDP, AD
|
Agrammatism ± AOS[b]
|
Left posterior frontal, insula
|
|
Semantic dementia
|
TDP > tau, AD
|
↓ Confrontation naming, single word comprehension
|
Left anterior temporal
|
|
PSPS
|
tau >> others
|
Subcortical dementia, axial park., falls, ↓ supranuclear gaze
|
Symmetrical posterior frontal, brainstem
|
|
Corticobasal syndrome
|
tau > TDP, AD
|
Asymmetrical parkinsonism, cortical signs
|
Asymmetrical frontal ± parietal[c]
|
|
Creutzfeldt-Jakob disease
|
PrPsc
|
RPD, myoclonus, ↓ pyramidal, extrapyramidal, cerebellar, VF
|
Cortex, striatum, thalamus
|
Abbreviations: AD, Alzheimer's disease; PPA, primary progressive aphasia; DLB, dementia
with Lewy bodies; PDD, Parkinson's disease dementia; bvFTD, behavioral variant frontotemporal
dementia; PSPS, progressive supranuclear palsy syndrome; LB, Lewy body; tau, frontotemporal
lobar degeneration (FTLD-) tau; TDP, FTLD-TAR DNA-binding protein 43 (TDP); FUS, FTLD-RNA-binding
protein FUS; PrPsc, prion; EF, executive function; VF, visual function; park., parkinsonism;
VH, visual hallucinations; AOS, apraxia of speech; RPD, rapidly progressive dementia;
SI, substantia innominata.
a The syndrome has not been explicitly defined; no clinical/pathologic correlations
available.
b Apraxia of speech is predictive of FTLD-tau pathology.
c Focal frontal atrophy is predictive of FTLD-tau pathology.
Converging evidence from cellular and systems-level neuroscience suggests that the
anatomic signatures of neurodegenerative processes correspond to large-scale networks
in the brain. This concept has been articulated as the network degeneration hypothesis, summarized in [Fig. 1] and by Pievani and colleagues.[2] Neuroimaging techniques have played a central role in exploring this hypothesis.
Functional connectivity MRI (fcMRI), examining interregional correlations in neuronal
activity with blood-oxygen-level-dependent (BOLD) signal, has revolutionized our understanding
of large-scale functional brain networks.[3]
[4] Supporting a relationship between large-scale networks and neurodegeneration, studies
have demonstrated correspondence between distributions of pathology and network anatomy,
for example, overlap between the distribution of fibrillar amyloid-β protein (Aβ)
as revealed by [11C] Pittsburgh Compound-B (PiB) positron emission tomography (PET) and the default network, a set of brain regions including the medial frontal cortex, posterior cingulate,
precuneus, lateral parietal cortex, and medial temporal cortex.[5] Employing resting state fcMRI and voxel-based morphometry (VBM), Seeley and colleagues[6] demonstrated concordance between patterns of volume loss in AD, bvFTD, SD, nonfluent
aphasia, and CBS and dissociable networks comprising regions that covary both functionally
and structurally in healthy control subjects. These studies have laid the foundation
for investigating potential mechanisms underlying network degeneration, which eventually
could help to explain different patterns of neurodegeneration with different types
of neuropathology.[7]
[8]
[9]
Fig. 1 Network degeneration hypothesis. *Discussed in Zhou et al.[9]
Neuroimaging has thus furthered progress toward the ultimate goal of predicting neuropathology
in vivo in single subjects. At present, imaging techniques have value in predicting
progression to dementia in individuals with mild cognitive impairment (MCI), a condition
representing an intermediate stage between cognitive changes present with aging and
those fulfilling criteria for dementia.[10] Given evidence that neuropathologic processes start years in advance of clinical
symptoms[11] and the widely accepted principle that early intervention will confer a greater
likelihood of success with disease-modifying treatments,[12] presymptomatic detection of disease using a combination of neuroimaging and other
biomarkers will likely represent a cornerstone of clinical management in the future.[13]
[14]
This review summarizes characteristic neuroimaging features for more common neurodegenerative
dementia clinical syndromes ([Table 1]). Imaging modalities already approved for and established in clinical use are presented,
along with salient information about selected research imaging techniques that may
have an increasing role in clinical practice in the future ([Table 2]).
Table 2
Imaging Biomarkers of Neurodegenerative Syndromes and Pathology
|
Presymptomatic
|
Mild Cognitive Impairment
|
Dementia
|
|
Neuropathology
|
✓
|
✓
|
✓
|
|
Clinically significant cognitive impairment
|
X
|
✓
|
✓
|
|
Loss of function in usual activities
|
X
|
X
|
✓
|
|
Structural MRI - visual inspection[a]
|
X
|
±
|
✓
|
|
FDG-PET[a]
|
G
[d]
|
±
|
✓
|
|
Quantitative loss of volume or cortical thickness[b]
|
G
[d]
|
✓
|
✓
|
|
Amyloid PET[c]
|
G
[d]
|
±
|
✓
|
|
Task fMRI
|
G
|
G
|
G
|
|
Resting state fcMRI
|
G
|
G
|
G
|
|
Diffusion tensor imaging
|
G
|
G
|
G
|
Abbreviations: ✓, Reliable changes detected in individuals; ± , changes inconsistently
detected in individuals; G, effects detected in groups; effects in individuals not yet established reliably;
X, changes not detected in individuals or groups; MRI, magnetic resonance imaging;
FDG-PET, [18F] fluorodeoxyglucose positron emission tomography; fMRI, functional MRI; fcMRI, functional
connectivity MRI.
a Modalities in clinical use.
b Some automated medial temporal lobe volumetric protocols in clinical use.
c Recently approved by the Food and Drug Administration for clinical use.
d Combined use of multiple modalities may increase predictive value for individual
subjects.
Alzheimer's Disease
AD represents the most common cause of both late-onset dementia (LOD) and early-onset
dementia (EOD),[a] accounting for ∼50–70% of LOD and 20–40% of EOD diagnosed clinically.[15]
[16] The pathologic defining features of AD include extracellular Aβ senile plaques (in
particular, a subset called neuritic plaques) and intracellular neurofibrillary tangles
(NFTs) containing the microtubule-associated protein tau.[17] The average lifetime risk of developing AD is ∼10–12%; although the presence of
a first-degree relative with AD approximately doubles this risk,[18] familial autosomal dominant disease due to mutations in the genes presenilin-1 (PSEN1, chromosome 14), presenilin-2 (PSEN2, chromosome 1), or amyloid precursor protein (APP, chromosome 21) account for less than 2% of all cases of AD.[19] AD usually presents with an insidious onset of difficulties remembering autobiographic
events (progressing from recent to remote) and frequently accompanied to a lesser
extent by problems with language (usually word-finding), executive functions, and
visuospatial functions. Approximately one-third of early-onset cases present atypically
with dysfunction in a cognitive domain other than episodic memory,[20]
[21] consistent with neuropathologic research reporting a younger mean age of onset for
individuals with an atypical distribution of tau pathology affecting the neocortex
and sparing the medial temporal lobes (“hippocampal sparing AD”).[22] The most common extreme nonamnestic phenotypes of AD are a visual variant (PCA),
a language variant (logopenic PPA), and a frontal/dysexecutive variant.
Structural Imaging in Alzheimer's Disease
Structural imaging with computed tomography (CT) and MRI represents the most widely
used imaging techniques to assist in the diagnosis of AD. High-resolution T1-weighted
sequences such as magnetization prepared rapid gradient echo (MP-RAGE) and spoiled
gradient echo (SPGR) MRI sequences allow for more exquisite detection of brain atrophy
and also lend themselves to quantitative morphometric techniques for measuring volume
and cortical thickness. The medial temporal lobes (primarily hippocampus and entorhinal
cortex) are typically the first location to demonstrate atrophy, and remain the most
severely affected region as illness progresses.[23] Atrophy spreads to paralimbic and association neocortices, in particular temporal
and parietal association cortex, with relative sparing of primary sensory and motor
cortices until late in the course of illness ([Fig. 2]).
Fig. 2 Atrophy in clinical variants of Alzheimer's disease (AD). (A) Typical (amnestic)
AD: Medial temporal atrophy out of proportion to other cortical areas. (B) Posterior
cortical atrophy due to AD: Severe bilateral parietooccipital atrophy. (C) Logopenic
primary progressive aphasia due to AD: Left > right temporoparietal atrophy.
Newer research criteria for the clinical diagnosis of AD have incorporated atrophy
as a feature to potentially assist in making an earlier diagnosis and in increasing
diagnostic confidence. For example, Dubois and colleagues[24] proposed criteria allowing for a diagnosis of probable AD in individuals with cognitive
dysfunction limited to episodic memory provided one obtains certain specified biomarkers
of AD, one of which is medial temporal atrophy (MTA) utilizing qualitative ratings.[b] The MTA scale represents one well-validated method toward this end, delineating
five levels of atrophy (MTA 0 through 4) based on manual inspection of coronal T1-weighted
images.[25] The National Institute on Aging-Alzheimer's Association (NIA-AA) workgroups introduced
diagnostic constructs including “probable AD dementia with evidence of the AD pathophysiological
process,” “possible AD dementia with evidence of the AD pathophysiological process,”
and “MCI due to AD,” incorporating atrophy in medial, basal, and lateral temporal
lobe and medial parietal cortex on MRI as evidence of downstream neuronal degeneration
or injury characteristic of AD.[26]
[27]
[c]
Quantitative structural imaging techniques provide a greater degree of sensitivity,
specificity, and reproducibility in both cross-sectional and longitudinal formats
than qualitative techniques, and are thus preferred for use in clinical treatment
trials.[23] Methods applied to single subject data have included manual, semiautomated, and
automated tracings of the hippocampus or entorhinal cortex,[28]
[29] longitudinal quantification of global brain volume over time,[30] and the use of support vector machine algorithms to extract information on the pattern
and severity of atrophy from a subject's three-dimensional (3-D) MRI dataset and to
compare it with information from large databases of scanned controls and subjects
with AD.[31]
[32]
[33]
[34] Quantitative measurements of hippocampal volumes in subjects with amnestic MCI have
not only revealed associations between smaller hippocampal volumes and/or more rapid
hippocampal atrophy and progression to a clinical diagnosis of dementia due to AD,[35]
[36]
[37] but also an association between subregional hippocampal atrophy and progression
from cognitively normal status to amnestic MCI.[38]
Additional insights into the neuroanatomy of AD have been obtained via quantitative
structural imaging methods such as VBM, cortical thickness mapping, and tensor-based
morphometry applied to groups of subjects with AD compared with healthy controls or
other conditions.[39] VBM represents a widely used technique for investigating topographic differences
in gray matter and/or white matter volumes across the entire brain between groups
of subjects.[40] Studies in AD have consistently demonstrated volumetric reductions in the medial
and inferior temporal lobes, posterior cingulate, precuneus, anterior cingulate, and
temporoparietal association cortex relative to control subjects and subjects with
nonprogressive mild memory impairment.[41]
[42]
[43]
Measurement of cortical thickness using automated or semiautomated algorithms[44] provides an advantage of capturing a physical property of the brain that can be
measured in an individual person in vivo or postmortem, with thinning of the cortex
corresponding to pathologic observations including cellular shrinkage, neuropil loss,
and reduction of intracortical myelin.[39]
[45] A consistent pattern of cortical thinning in AD comprises the medial temporal lobes;
inferior and anterior temporal association cortices; superior, inferior, and medial
parietal association cortices; along with superior and inferior frontal association
regions.[46]
[47]
[48] Dickerson and colleagues have demonstrated high reliability of the “cortical signature
of AD,” a set of regions with thinning across patient samples and scanner platforms
that also predicts progression to AD dementia in subjects with MCI[49] and older individuals without cognitive impairment.[50]
Importantly, structural neuroimaging measures of atrophy appear to correlate well
with the magnitude and distribution of NFT pathology. Consistent with the aforementioned
relationships between hippocampal atrophy and clinical progression to amnestic MCI
and AD, hippocampal volumes measured antemortem in demented and nondemented subjects
have been shown to correlate well with postmortem Braak NFT scores.[51] Similarly, the magnitude and pattern of gray matter volume loss across the cerebrum
as determined by VBM corresponds with the severity and pattern of NFT pathology as
determined by NFT counts at Braak stages III and IV and Braak stage itself at stages
V and VI.[52] Employing a support vector machine algorithm, Vemuri and colleagues likewise obtained
excellent correlation between structural MRI changes in single subjects antemortem
(as reflected in the Structural Abnormality Index (STAND) score) and portmortem Braak
NFT scores.[53]
Functional and Molecular Imaging in Alzheimer's Disease
MRI, PET, and single-photon emission computed tomography (SPECT) offer tools for investigating
changes in the functioning and/or molecular composition of brain tissues, particularly
valuable because such changes may precede atrophy detectable by structural imaging
methods. [18F] Fluorodeoxyglucose positron emission tomography (FDG-PET) and brain perfusion SPECT
are currently in widespread clinical use for diagnosis of and discrimination between
different types of neurodegenerative dementia. AD typically produces a characteristic
pattern of hypometabolism or hypoperfusion in bilateral temporoparietal and posterior
cingulate/precuneus cortex with later involvement of the frontal lobes and sparing
of primary sensorimotor cortices. FDG-PET offers higher sensitivity and spatial resolution
than does SPECT.[54] Images may be analyzed qualitatively, with volume-of-interest techniques, or with
voxel-based analyses in stereotactic space.[55] Studies investigating the use of qualitative criteria to classify PET scans as reflective
or not reflective of AD in groups of subjects evaluated for possible memory impairment
with eventual pathologically confirmed diagnoses suggest a diagnostic sensitivity
of 84–94% and specificity of 74–78%.[56]
[57] Ongoing efforts to optimize and standardize methodologies for acquiring, analyzing,
and interpreting FDG-PET data will further increase its value as a biomarker in the
clinical setting and in treatment trials for AD.[58]
Amyloid PET imaging with agents such as PiB and [18F] florbetapir (AV-45) has revolutionized the practice of predicting AD pathology
in vivo. Amyloid PET agents detect β-sheet rich fibrillar deposits of Aβ as typically
present in compact/cored plaques, to varying degrees in diffuse plaques, and in cerebrovascular
amyloid.[59] Preliminary studies correlating in vivo PiB or florbetapir PET results with postmortem
pathologic diagnosis suggest a high sensitivity of amyloid PET for detecting probable
or definite pathologic AD by CERAD (Consortium to Establish a Registry for Alzheimer's
Disease[60]) criteria and/or intermediate or high likelihood of pathologic AD by NIA/Reagan
Institute (National Institute on Aging and Reagan Institute Working Group on Diagnostic
Criteria for the Neuropathological Assessment of Alzheimer's Disease[61]) criteria, and a high specificity for amyloid deposition.[62]
[63] [d] The United States Food and Drug Administration's recent approval of florbetapir
PET for clinical use stands to increase clinicians' diagnostic accuracy for AD (as
confirmed pathologically), particularly in less-specialized settings, if used appropriately.[64] Although amyloid imaging appears promising as a tool for predicting progression
of MCI to AD dementia[62] and even for detecting AD in its early presymptomatic stages,[65] continued research is required to identify and actualize its full clinical utility.
Research studies utilizing BOLD functional MRI (fMRI) have contributed a great deal
to our understanding of AD and other neurodegenerative dementias, and offer additional
potential biomarkers for clinical use in treatment trials and beyond. Task fMRI studies
employing block or event-related memory paradigms in subjects with AD and healthy
older controls have consistently revealed disease-related decreases in activity in
MTL structures (hippocampus, parahippocampus), prefrontal regions, superior parietal
lobule, precuneus, cingulate, and lingual gyrus (see Schwindt and Black[66] for meta-analysis). In contrast, numerous investigators have reported increases
in MTL activity during encoding in subjects with MCI or asymptomatic individuals with
genetic or family risk factors who remain able to perform the tasks well (reviewed
in Sperling et al[67]). Although this may reflect compensatory mechanisms, longitudinal studies have suggested
that increased MTL activity at baseline predicts rapid cognitive decline and decreased
MTL function.[68]
Functional connectivity MRI offers advantages of not requiring or having to account
for task performance, not requiring additional hardware, and of potentially generating
reproducible data across scanner platforms with relatively short acquisition sequences.[69] Studies employing fcMRI with either “seed-based” or independent component analysis
(ICA) methods have suggested that, relative to controls, groups of subjects with AD
have exhibited reduced intrinsic functional connectivity between default network regions,[70]
[71] a finding that also appears to discriminate between subjects with MCI who progress
to dementia and those who do not.[72] Additional research is needed to validate functional MRI techniques for potential
use as a biomarker of AD in single subjects.[68]
[73]
Atypical Presentations of Alzheimer's Disease
As described above, AD occasionally presents atypically, with prominent primary dysfunction
in language, attention/executive processing, or higher-order visual processing. In
such cases, neuroimaging frequently helps to confirm the presence of a neurodegenerative
syndrome with a characteristic pattern of atrophy involving specific neocortical regions
and sparing the medial temporal lobes ([Fig. 2]). PCA frequently causes parietal and occipital symptoms such as visual dysfunction
in the absence of ocular disease, elements of the syndromes of Bálint and/or Gerstmann,[e] constructional apraxia, visual field defects, and environmental disorientation[74] (see Crutch et al[75] for review). Structural imaging characteristically reveals prominent, relatively
symmetric atrophy in the parietal and occipital lobes.[76] Logopenic PPA is characterized by significant word-retrieval difficulties, poor
repetition of phrases and sentences, phonologic errors in speech, but preserved grammar,
motor output, and repetition and comprehension of single words.[77] Imaging evidence consists of atrophy and/or hypoperfusion or hypometabolism in left
posterior perisylvian temporal and parietal cortices.[78] Dysexecutive or “frontal” AD has been less explicitly characterized as a clinical
syndrome distinct from frontotemporal dementia (FTD), but appears likely to involve
early frontal and temporoparietal atrophy.[79]
[80]
Imaging studies have suggested that, though clinically distinct, these nonamnestic
phenotypes share a common core temporoparietofrontal network.[81] Not only have voxel-based techniques revealed substantial overlap in patterns of
atrophy across groups of patients with PCA, logopenic PPA, and early-onset AD,[82] they have been used to discriminate between groups of patients with pathologically
confirmed AD (both amnestic and nonamnestic) and frontotemporal lobar degeneration
regardless of clinical diagnosis.[80]
[83] Similarly, PET studies comparing atypical or early-onset AD with typical AD have
revealed similar patterns of fibrillar amyloid deposition between atypical and typical
groups as well as hypometabolism in a core distribution of regions including temporoparietal,
posterior cingulate, and precuneus.[84]
[85]
Dementia with Lewy Bodies
Dementia with Lewy Bodies
Dementia with Lewy bodies and Parkinson's disease dementia (PDD) are the second most
common type of neurodegenerative dementia behind AD, with estimated prevalences of
0.7% and 0.3%, respectively, in the population of individuals aged 65 and older. [86] [f] In addition to dementia, central features of DLB (and frequently PDD) include fluctuations
in attention, alertness or cognition, visual hallucinations, and parkinsonism (see
McKeith et al[87] for consensus clinical criteria). Early changes in cognition frequently involve
attention, executive functions, and higher-order visual functions.[88]
[89] Additional commonly encountered features include neuropsychiatric symptoms (depression,
anxiety, apathy, and delusions), rapid eye movement (REM) sleep behavior disorder,
severe sensitivity to neuroleptic medications, and autonomic dysfunction.
Lewy bodies, the pathologic hallmark of DLB, consist of filamentous intraneuronal
protein inclusions containing the protein α-synuclein distributed throughout the central,
peripheral, and autonomic nervous systems.[87] To what degree the quantity and distribution of Lewy body pathology correlates with
symptom type, severity, and progression in Lewy body disorders remains a matter of
debate.[90] Synaptic dysfunction and/or cell loss may provide better correlates of clinical
dysfunction. Although most patients with DLB have high levels of senile plaques containing
Aβ, the level of neocortical NFT pathology is more variable and more directly influences
the presence of AD-like versus DLB-like phenotypic manifestations.[87]
[91]
Structural Neuroimaging in Dementia with Lewy Bodies
Given the pathologic heterogeneity of DLB, it is perhaps not surprising that studies
investigating cortical and subcortical atrophy in DLB have yielded inconsistent results
(see Taylor and O'Brien[92] for review). Structural neuroimaging demonstrating relative sparing of the medial
temporal lobes from atrophy supports a clinical diagnosis of DLB rather than AD in
the correct context.[93]
[94] Notably, this principle was supported by a recent MRI volumetric study with neuropathologic
classification of subject groups revealing increasing antemortem hippocampal and amygdalar
volumes with increasing likelihood of DLB pathology and lower volume of the dorsal
mesopontine gray matter in the group with high likelihood DLB compared with healthy
control subjects and with subjects with AD.[95]
Functional and Molecular Imaging in Dementia with Lewy Bodies
Several types of PET or SPECT studies can be used to support a diagnosis of DLB. Low
dopamine transporter (DAT) uptake as measured by [18F] fluorodopa PET or [123I] ioflupane SPECT is useful to distinguish degenerative from nondegenerative causes
of parkinsonism.[96] A large multicenter study comparing ioflupane SPECT against a gold standard of clinical
diagnosis of DLB or non-DLB dementia established longitudinally, suggested diagnostic
sensitivity of ∼80% and specificity of ∼90% for SPECT,[97] findings corroborated by a smaller study yielding sensitivity of 88% and specificity
of 100% against a pathologic gold standard of DLB versus non-DLB dementia.[98] Perhaps of greater clinical utility, ioflupane SPECT is also reliable in confirming
cases of clinically possible DLB,[99] supporting the inclusion of reduced DAT uptake as a “suggestive feature” of DLB
in formal criteria.[87]
Hypometabolism or hypoperfusion in the occipital and posterior temporoparietal regions
on PET or SPECT also occurs frequently, although sensitivities and specificities are
debated.[92] Kantarci and colleagues[100] demonstrated the value of utilizing multiple imaging biomarkers to distinguish between
clinical DLB and AD, achieving an area under the receiver operating characteristic
(ROC) curve of 0.98 by combining measures of hippocampal volume, global PiB retention,
and occipital metabolism. Additional studies with multiple modes of imaging and pathologic
correlation will likely shed additional light on the relationship between amyloid
pathology, network function, neurodegeneration, and clinical symptoms in DLB.
Frontotemporal Lobar Degeneration
Frontotemporal Lobar Degeneration
Frontotemporal lobar degeneration (FLTD) comprises several distinct subtypes of pathology
that have a predilection to involve the frontal and/or temporal lobes early in the
course of illness. The more commonly occurring subtypes of FTLD are those distinguished
by tau inclusions (FLTD-tau) and those by inclusions containing TAR DNA-binding protein
43 (FTLD-TDP). FTLD-tau pathologies are further classified into 3R, 4R, or combined
3R/4R subtypes on the basis of the number of microtubule-binding repeats in the tau
protein, a consequence of differential mRNA splicing.[101] FTLD-TDP can be classified into types A, B, C, and D based upon neuropathologic
features.[102] The remaining minority of FTLD cases are characterized by inclusions containing
RNA-binding protein FUS (FTLD-FUS) or other rare pathologies.
Frontotemporal dementias, the variety of clinical syndromes arising in the setting
of FTLD pathology, include bvFTD, agrammatic PPA, SD, FTD with motor neuron disease
(FTD-MND), PSPS, and CBS. FTD represents a common cause of early-onset dementia, with
an incidence and prevalence similar to that of AD.[103] An estimated 25–50% of cases have a family history of dementia; an estimated 10–25%
of cases demonstrate a pattern of inheritance suggestive of an autosomal dominant
gene mutation.[104]
[105] The most common genes associated with familial autosomal dominant FTD are the tau
gene (MAPT; associated with FTLD-tau), the progranulin gene (GRN; associated with FTLD-TDP), and the chromosome 9 open reading frame 72 gene (C9ORF72; associated with FTLD-TDP), a relatively common cause of both familial FTD and familial
amyotrophic lateral sclerosis (ALS).[106]
[107] Clinical/pathologic associations vary for different FTD syndromes; for example,
PSPS is almost always associated with 4R FTLD-tau pathology,[g] whereas CBS and bvFTD are pathologically more heterogeneous.
Neuroimaging in Behavioral Variant Frontotemporal Dementia
Unlike most other neurodegenerative clinical syndromes, bvFTD produces early dramatic
changes in personality and behavior often out of proportion to cognitive deficits
detectable by standard neuropsychologic measures (see Piguet et al[108] for review). Symptomatic hallmarks, reflected in the most recent consensus clinical
diagnostic criteria, include disinhibition, apathy/inertia, loss of sympathy/empathy,
perseverative/compulsive behaviors, hyperorality/change in food preference, and executive
dysfunction.[109] Approximately 50% of bvFTD cases demonstrate FTLD-TDP pathology (all subtypes, most
commonly type A), ∼40% FTLD-tau pathology, and the majority of the remaining cases
FTLD-FUS pathology.[110]
In its earliest stages, bvFTD can be difficult to discriminate from developmental
or psychiatric conditions that may present in similar fashion, a phenomenon described
in the literature on nonprogressive bvFTD “phenocopies.”[111] Atrophy in nodes of the salience network, a paralimbic circuit comprising anterior insular (right > left), medial frontal,
orbitofrontal, and temporopolar regions (among others) is reliably associated with
progressive, degenerative bvFTD.[112] Kipps and colleagues[113] developed a qualitative visual rating scale for FTD with high inter-rater reliability
on the basis of atrophy in the frontal lobes, anterior temporal lobes, and posterior
temporal lobes. VBM has been used to demonstrate the pattern and progression of atrophy
from very mild bvFTD through moderate to severe disease,[114] as well as to delineate multiple distinct patterns of atrophy via hierarchical agglomerative
cluster analysis.[115] Algorithms for pattern-based classification of individual MR images as FTD, AD,
or healthy older individuals have yielded promising results.[116]
Visual interpretation of perfusion SPECT or FDG-PET scans has been shown to improve
diagnostic accuracy in distinguishing between FTD and AD based on clinical information.
Differentiating between a pattern of hypometabolism in frontal association cortex,
anterior temporal cortex, and anterior cingulate cortex from a pattern of hypometabolism
in posterior cingulate and posterior association cortex improved clinical diagnostic
accuracy from 79% to 90% in an FDG-PET study with pathologic confirmation.[117] Visual interpretation of perfusion SPECT improved specificity for a clinical diagnosis
of FTD (vs AD) and sensitivity for a diagnosis of AD (vs FTD) from 77% to 84% in a
separate study with pathologic confirmation.[118]
Diffusion tensor imaging (DTI) and fcMRI have offered exciting tools for probing the
integrity of white matter tracts and functional networks in vivo in bvFTD.[119] Reduced fractional anisotropy (FA), reflecting reduced integrity of white matter,
has been detected with DTI in bvFTD versus healthy controls in frontal and temporal
tracts including the anterior corpus callosum, anterior and descending cingulum tracts,
and the uncinate fasciculus.[120] Reduced FA in the anterior cingulum tracts has been shown to predict executive dysfunction
(left tract) and deficits in visuospatial attention and working memory (right tract)
in bvFTD.[121] Consistent with these DTI results and with the pattern of regional atrophy in bvFTD,
fcMRI studies to date have reported decreases in salience network connectivity.[122]
Neuroimaging in Semantic Dementia and Agrammatic PPA
SD and agrammatic PPA usually present with changes in language out of proportion to
other cognitive domains. The left temporal variant of SD causes a fluent aphasia with
impairments in single-word comprehension and confrontation naming, reduced vocabulary,
and frequently associated features including impaired object knowledge, and inability
to read or spell orthographically irregular words[78]
[123]
[124](see Hodges and Patterson[124] for review). With time, behavioral features such as irritability, emotional withdrawal,
disinhibition, and behavioral rigidity usually develop.[123]
[125] The less frequently encountered right temporal variant frequently presents with
changes in emotional and social functioning such as apathy, emotional blunting, loss
of extroversion, in addition to difficulty recognizing people and objects and later
on the anomic aphasia similar to that seen in the left temporal variant.[125]
[126]
[127]
[128] A strong majority of SD cases show FTLD-TDP type C pathology, with much fewer cases
exhibiting FTLD-tau pathology or AD pathology (reviewed in Grossman[129]). SD very reliably produces asymmetrical atrophy, hypoperfusion, or hypometabolism
of the anterior temporal lobes ([Fig. 3]).[77]
[78]
[130]
[131]
[132]
[133]
Fig. 3 Atrophy in frontotemporal dementia. (A) Semantic dementia: Severe asymmetric anterior
temporal atrophy. (B) Behavioral variant frontotemporal dementia: Frontal, temporal,
paralimbic (insula, anterior cingulate) cortex. (C) Agrammatic primary progressive
aphasia: Left posterior frontal/insular atrophy.
In contrast to SD, agrammatic PPA causes a nonfluent aphasia with agrammatism, effortful,
halting speech, and frequently, impaired comprehension of syntactically complex sentences.[78] Cases can be classified based on whether there is superimposed apraxia of speech,
an oral motor speech disorder characterized by difficulty initiating utterances, abnormal
rhythm, stress, and intonation, inconsistent articulation errors, effortful trial
and error with groping, and self-correction of errors.[134]
[135] Although agrammatic PPA can be associated with FTLD-tau, AD, DLB, and FTLD-TDP type
A pathology,[129] Josephs and colleagues have demonstrated that the presence of apraxia of speech
predicts FTLD-tau pathology.[134] Neuroimaging characteristics of agrammatic PPA include atrophy or hypometabolism
in the posterior left frontal/insula region, including inferior frontal gyrus, frontal
operculum, insula, premotor and supplementary motor areas ([Fig. 3]).[77]
[133]
[134]
[136]
Neuroimaging in Progressive Supranuclear Palsy and Corticobasal Syndromes
CBS and PSPS are syndromes characterized by atypical parkinsonism with varying levels
of cognitive impairment and dementia, originally recognized in the setting of distinctive
4R tau pathologies, corticobasal degeneration (CBD), and PSP.[137]
[138] Clinical features of CBS include asymmetrical apraxia and extrapyramidal symptoms
(bradykinesia and rigidity ± tremor, dystonia) accompanied by signs of cortical involvement
such as alien limb syndrome, cortical sensory loss, hemisensory neglect, myoclonus,
and/or impairments in speech and language, visuospatial function, or executive function.[139]
[140]
[141] PSPS comprises core features of severe early postural instability and falls, supranuclear
gaze impairment that manifests with decreased saccade velocity or ophthalmoplegia,
and an extrapyramidal syndrome with prominent axial rigidity.[142] A majority of patients with PSPS also develop a dementia syndrome with frontal and
subcortical features such as disinhibition or impulsivity, apathy, social withdrawal,
stimulus bound/imitative/utilization behavior, psychomotor slowing, executive dysfunction,
and pseudobulbar affect. As might be predicted by the multiple phenotypes cause by
both PSP and CBD pathologies, it is possible to see patients with overlapping features
of CBS, PSP, agrammatic PPA, bvFTD, and other syndromes.[143]
[144]
[145]
[146]
MRI in CBS characteristically reveals asymmetric atrophy of the perirolandic cortex
and basal ganglia, whereas PSPS has been associated consistently with atrophy of the
brainstem (particularly midbrain and superior cerebellar peduncle) and frontal white
matter with lesser involvement of the frontal cortex.[147]
[148]
[149] Specific measurement of the areas of the midbrain tegmentum and the pons on midsagittal
MRI and computation of the midbrain/pons ratio has been shown to discriminate between
patients with a clinical diagnosis of PSP and those with diagnoses of Parkinson's
disease or multiple system atrophy of the Parkinson type.[149]
[150] In subjects with CBS focal atrophy of the posterior frontal cortex has been found
to predict CBD (FTLD-tau) pathology, more widespread frontotemporal atrophy to predict
FTLD-TDP pathology, and more widespread temporoparietal atrophy to predict AD pathology.[151] Similarly, an AD-like pattern of hypoperfusion or hypometabolism with SPECT or PET
imaging may predict CBS due to AD versus CBS due to other pathologies.[152]
[153]
Imaging and Prediction of Molecular Pathology in FTLD
Quantitative morphometric MRI studies have generated the most robust body of data
thus far with respect to delineating the spatial patterns of FTLD subtypes (see Whitwell
and Josephs[154] for review). In a large retrospective study employing cluster analysis of pathologic
groups, Rohrer and colleagues[155] proposed a neuroanatomic framework for understanding profiles of atrophy, differentiating
between pathology associated with relatively symmetric, localized temporal or extratemporal
atrophy (FTLD-tau caused by MAPT mutations, CBD, FTLD-FUS), pathology associated with relatively asymmetric, localized
temporal atrophy (FTLD-TDP type C), and pathology associated with relatively asymmetric,
distributed atrophy (Pick's disease, FTLD-TDP type A caused by GRN mutations). Somewhat consistent with this framework, Whitwell, Josephs, and colleagues[154] identify relatively focal posterior frontal and subcortical atrophy in 4R FTLD-tau
(more consistently symmetric in PSP than CBD), relatively focal and symmetrical temporal
atrophy in FTLD-tau caused by MAPT mutations, and a broader pattern of frontal, temporal, and paralimbic atrophy in
Pick's disease, with more bilaterality in their sample.[156] They further highlight asymmetrical widely distributed frontotemporoparietal atrophy
in FTLD-TDP type A, mild frontal and medial temporal atrophy in FTLD-TDP type B (a
subtype strongly associated with FTD-MND), and focal, asymmetrical temporal atrophy
in FLTD-TDP type C).[157] FTLD-FUS, associated with a syndrome of very early-onset bvFTD in patients 20–45
years old,[158] characteristically produces severe atrophy of the caudate nucleus.[154]
[155]
[158]
DTI correlates of patterns of white matter involvement in FTLD subtypes have not been
as well delineated, but should have value given the distinctive white matter features
of different subtypes.[119] A combination of regional differences in both gray matter volume and white matter
integrity discriminated between groups of patients with FTLD and AD pathology with
87% sensitivity and 83% specificity, with reduced FA in the corpus callosum characteristic
of FTLD versus AD.[159] Similar methods were used to accurately discriminate a group with FTLD-TDP from
a group with FTLD-tau with greater reductions in FA in the superior longitudinal fasciculus
in the FTLD-tau group corroborated by postmortem pathologic examination.[160] Reduced FA in the left uncinate fasciculus and left inferior occipitofrontal fasciculus
has been reported in asymptomatic GRN mutation carriers in a four-generation FTLD pedigree, illustrating the potential
value of this technique in early stages of illness.[161]
FcMRI has been similarly applied to study preclinical familial FTLD due to mutations
in both GRN and MAPT. In contrast to reductions in salience network connectivity detected in symptomatic
patients, presymptomatic mutation carriers demonstrated increased connectivity, possibly
indicative of compensatory mechanisms.[162] Preclinical MAPT mutation carriers have demonstrated altered default network connectivity (including
reduced connectivity in lateral temporal and medial prefrontal cortex and increased
connectivity in the medial parietal lobe), changes similar to those seen in a group
clinically diagnosed with bvFTD.[163]
Summary and Conclusions
Diagnosis of neurodegenerative dementia is currently predicated on identifying insidious
onset and gradual progression characteristic patterns of cognitive dysfunction, involving
memory (amnestic AD), language (agrammatic and logopenic PPA, SD), visual processing
(PCA, DLB), executive dysfunction (bvFTD, DLB), or social/emotional cognition (bvFTD,
SD), with or without parkinsonism or motor neuron disease. Structural and functional
neuroimaging techniques have advanced our understanding of how patterns of cognitive
dysfunction localize to large-scale networks, which appear to serve as substrates
for the pathologic neurodegenerative processes themselves. At present, high-resolution
structural MRI, FDG-PET, perfusion SPECT, and DAT imaging are widely used modalities
that contribute to clinical diagnosis. Ongoing research developing and refining these
techniques and newer techniques such as amyloid PET, fcMRI, and DTI promises to improve
our ability to establish pathologic diagnoses in vivo in early stages of illness.
Imaging biomarkers are likely to be most powerful when used together, with nonimaging
biomarkers, and with sophisticated computer algorithms for pattern recognition.[83] Large, multicenter longitudinal studies including the Alzheimer Disease Neuroimaging
Initiative,[164] the Dominantly Inherited Alzheimer Network,[165] and biomarker studies in FTLD continue to propel these efforts.
