|Home | About | Journals | Submit | Contact Us | Français|
Structural imaging based on magnetic resonance is an integral part of the clinical assessment of patients with suspected Alzheimer dementia. Prospective data on the natural history of change in structural markers from preclinical to overt stages of Alzheimer disease are radically changing how the disease is conceptualized, and will influence its future diagnosis and treatment. Atrophy of medial temporal structures is now considered to be a valid diagnostic marker at the mild cognitive impairment stage. Structural imaging is also included in diagnostic criteria for the most prevalent non-Alzheimer dementias, reflecting its value in differential diagnosis. In addition, rates of whole-brain and hippocampal atrophy are sensitive markers of neurodegeneration, and are increasingly used as outcome measures in trials of potentially disease-modifying therapies. Large multicenter studies are currently investigating the value of other imaging and nonimaging markers as adjuncts to clinical assessment in diagnosis and monitoring of progression. The utility of structural imaging and other markers will be increased by standardization of acquisition and analysis methods, and by development of robust algorithms for automated assessment.
Clinical and neuropathological studies have greatly advanced our knowledge of the pathophysiology and progression of Alzheimer disease (AD). This disease is associated with progressive accumulation of abnormal proteins (amyloid-β [Aβ] and hyperphosphorylated tau) in the brain, which leads to progressive synaptic, neuronal and axonal damage. Neurobiological changes occur years before symptoms appear, with a stereotypical pattern of early medial temporal lobe (entorhinal cortex and hippocampus) involvement, followed by progressive neocortical damage.1,2 The delay in emergence of the cognitive correlates of these changes suggests that the toxic effects of tau and/or Aβ progressively erode ‘brain reserve’ until a clinical threshold is surpassed and amnestic symptoms develop. For example, amnestic mild cognitive impairment (MCI)—memory disturbance in the absence of dementia—is followed by more-widespread cognitive deficits in multiple domains until a disability threshold is reached and traditional diagnostic criteria for probable AD are fulfilled.3 The prospect of disease-modifying drugs that can slow or prevent disease progression has prompted increased interest in identifying individuals with AD earlier and more accurately.
Several studies have shown that structural MRI estimates of tissue damage or loss in characteristically vulnerable brain regions, such as the hippocampus and entorhinal cortex, are predictive of progression of MCI to AD. Moreover, the clinical utility of MRI in differentiating AD from other pathologies, such as vascular or non-Alzheimer neurodegeneration, has been established. Finally, MRI-based estimates of progression; for example, atrophy rates, might be used to assess potential disease-modifying drugs in phase II and III trials.
In this article, we review current knowledge on structural MRI changes in AD, focusing particularly on measures of atrophy in typical late-onset sporadic AD. We also address other promising biomarkers that can set structural loss in the broader context of molecular, metabolic and functional changes at different stages of the disease. Current and future methods to measure regional atrophy in clinical settings have been reviewed elsewhere.4–6
MRI-based measures of atrophy are regarded as valid markers of disease state and progression for several reasons. Atrophy seems to be an inevitable, inexorably progressive concomitant of neurodegeneration. The topography of brain tissue loss correlates well with cognitive deficits, both cross-sectionally and longitudinally. Structural brain changes map accurately upstream to Braak stages of neurofibrillary tangle deposition7,8 and downstream to neuropsychological deficits.9,10 The earliest sites of tau deposition and MRI-based atrophic changes typically lie along the perforant (polysynaptic) hippocampal pathway (entorhinal cortex, hippocampus and posterior cingulate cortex), consistent with early memory deficits.11,12 Later, atrophy in temporal, parietal and frontal neocortices is associated with neuronal loss, as well as language, praxic, visuospatial and behavioral impairments.13,14
Rates of change in several structural measures, including whole-brain,15–19 entorhinal cortex,20 hippocampus9,21–23 and temporal lobe volumes,24,25 as well as ventricular enlargement,9,21,23,26 correlate closely with changes in cognitive performance, supporting their validity as markers of disease progression. The appropriate use of an atrophy marker in the clinic requires that its dynamics are known at the different stages of the disease, and that its relationship with the dynamics of other imaging and biological markers is understood. Atrophy measures change with disease progression over a wide range of AD disease severity. From MCI to well into the moderate dementia stage of AD, structural markers are more sensitive to change than are markers of Aβ deposition (as assessed through imaging or cerebrospinal fluid [CSF] analysis).18,27 In the asymptomatic to MCI stages, however, indirect evidence indicates that amyloid markers show more-substantial abnormalities than do structural markers (Figure 1).16,28–33
Macrostructural loss (atrophy) is accompanied by microstructural (dendritic, myelin and axonal) loss and metabolite changes, all of which are measurable with other magnetic resonance-based sequences. Magnetic resonance spectroscopy,34 diffusion-weighted imaging (DWI),35 fiber tracking,36 and magnetization transfer imaging37,38 are all either sensitive to early change or can add complementary information to atrophy measures. other MRI-based techniques, such as tissue perfusion with arterial spin labeling39,40 or functional measures of resting-state networks (particularly the default mode network),41,42 show promise as diagnostic markers, but have not yet been subjected to thorough validation. None of these techniques yet has an established role in clinical practice.
The key role of imaging in AD diagnosis is highlighted by the inclusion of imaging markers in proposed new criteria for earlier diagnosis of AD.43 These criteria build on traditional National Institute of Neurological and Communicative Disorders and Stroke–Alzheimer's Disease and Related Disorders Association criteria by keeping the requirement for objective memory deficits but removing the requirement that disability (dementia) must already be present. Instead, at least one of the following three markers is required: medial temporal atrophy, temporoparietal hypometabolism, and abnormal neuronal CSF markers (tau and/or Aβ). These criteria imply that structural imaging and other markers can reliably detect AD before dementia occurs; that is, at an MCI stage (Box 1).
The concept of mild cognitive impairment (MCI) was developed in the 1990s to capture patients with early clinical signs of Alzheimer disease (AD) who did not yet fulfill the criteria for dementia. The amnestic variant of MCI features the following: memory complaints, preferably qualified by an informant; memory impairment for age, as indexed by low cognitive performance in one or more neuropsychological tests that tap into learning abilities (for example, prose recall, word list); preserved general cognitive function (for example, Mini-Mental State Examination score of 24 out of 30 or above); intact activities of daily living; and no dementia. About two-thirds of all patients with amnestic MCI harbor the pathological features of AD and develop the clinical syndrome of Alzheimer dementia within 5 years, whereas the remaining one-third have non-progressive or very slowly progressive causes of cognitive impairment (for example, depression or age-related cognitive impairment). Proposed new diagnostic criteria for AD developed in 200743 suggested that the disease can be recognized at the MCI stage if the patient is positive for at least one of the following four markers: medial temporal atrophy on MRI; temporoparietal cortical hypometabolism on 18F-fluorodeoxyglucose PET; abnormality of cerebrospinal fluid markers (tau, amyloid-β42 or phospho-tau); and positivity on amyloid imaging with PET. These criteria need to be validated before being applied in clinical populations.
Of all the MRI markers of AD (Box 2),44 hippocampal atrophy assessed on high-resolution T1-weighted MRI is the best established and validated. The simplest way to assess atrophy of the medial temporal lobes is by visual inspection of coronal T1-weighted MRI. Several rating scales to quantify the degree of atrophy have been developed and are widely used. Visual rating scales provide ≈80–85% sensitivity and specificity to distinguish patients with AD from those with no cognitive impairment, and only slightly lower sensitivity and specificity levels for diagnosing amnestic MCI. These scales also have good predictive power to anticipate decline in MCI.45–48 Visual rating also correlates well with underlying pathology and has high diagnostic accuracy against a pathologically verified diagnosis of AD.49
Current clinical MRI scanners with 1.5T or 3T magnets allow acquisition of high-resolution digital images of the brain in exquisite structural detail, with excellent tissue contrast and spatial resolution of ≤1 mm. Atrophy of target structures can be estimated through procedures with varying levels of human input. Visual rating scales allow atrophy of medial temporal lobe structures to be categorized into discrete levels of increasing severity. Structures with definite boundaries can be labeled by manual outlining and volumes can be computed. Automated algorithm pipelines can align an individual digital brain to a reference template on a voxel-by-voxel basis and automatically label brain structures on the basis of prior knowledge of a digital atlas. In either case, volumes can be normalized to head size and compared with a normative population. A variety of voxel-based techniques treat the information of each voxel with mathematical models, which allow the production of maps of density, volume or other features of the brain tissue, and derived maps of significance, variance and other statistical measures.
Despite its convoluted structure, the boundaries of the hippocampus (and adjacent CSF spaces) are easier for human operators or automated algorithms to recognize than the amygdala, entorhinal cortex or parahippocampal gyrus. This is because the anatomical boundaries of the hippocampus are distinct on high-resolution T1-weighted MRI scans around most of the surface of this structure. Hippocampal volume measured in vivo by MRI correlates with Braak stage and neuronal counts.50–52 At the mild dementia stage of AD, hippocampal volume is already reduced by 15–30% relative to controls,38 and in the amnestic variant of MCI the volume is reduced by 10–15%53 (a meta-analysis of hippocampal MRI studies is provided elsewhere54). A recent meta-analysis estimated that medial temporal atrophy has ≈73% sensitivity and ≈81% specificity for predicting whether patients with amnestic MCI will convert to dementia.55 If medial temporal atrophy is measured with a continuous metric such as hippocampal volume, specificity might be increased, but at the cost of reduced sensitivity. If hippocampal atrophy is used as an inclusion criterion for clinical trials in MCI, a trade-off ensues between a relatively low proportion of screened negatives with a more-contaminated sample of screened positives and a higher proportion of screened negatives but a ‘cleaner’ group to treat and follow (Figure 2). Indeed, contamination of MCI groups with non-Alzheimer cases might in part explain the failure of some trials with cholinesterase inhibitors in patients with MCI.56 Enrichment of MCI groups with true AD cases in clinical trials of drugs aiming to delay the development of dementia might lead to a significant increase in study power.57
Despite the evidence reported above, medial temporal atrophy is not sufficiently accurate on its own to serve as an absolute diagnostic criterion for the clinical diagnosis of AD at the MCI stage. High specificity is required to minimize a false-positive diagnosis of AD, and medial temporal atrophy by itself lacks the specificity to confidently exclude other dementias.58 To enhance the accuracy of structural markers, other structural and nonstructural measures can be added in an algorithmic formula to diagnose AD. Studies that included pathological confirmation of the diagnosis have shown that parietal atrophy combined with medial temporal lobe atrophy on MRI carries positive predictive value for diagnosing AD.59 Moreover, in 59 patients with amnestic MCI, 33 of whom converted to dementia in 19 months on average, those with both medial temporal atrophy (as rated visually45) and abnormal CSF biomarkers had a fourfold higher risk of progression to dementia than patients with either abnormality alone.60 Prediction of dementia was almost perfect (94% positive predictive value),60 but replication is still needed. CSF and MRI measures provided better prediction than either measure alone, although MRI measures were more accurate in a head-to-head comparison.8
Other atrophy markers have been suggested for early diagnosis of AD or to enrich MCI trials, but their validity and added value for predicting decline remains to be confirmed. These markers include analysis of whole-brain patterns of atrophy through use of support vector machines,61 the AD-specific structural abnormality index (STAND) score,10,62 patterns of hippocampal subfield atrophy63 structural changes in cholinergic nuclei of the basal forebrain,64 deformation-based morphometry of the gray and white matter,24 and measures of the lateral temporal and parietal cortex.65
For any diagnostic marker to be useful in practice, its capacity to separate two or more conditions that can be confused on clinical grounds is a necessary but not sufficient requirement. Clinical usefulness requires that the marker provides incremental benefit over and above that provided by clinical assessment. Remarkably few attempts have been made to investigate the incremental diagnostic value of imaging markers in the differential diagnosis of the dementias,66,67 and the section that follows should be interpreted in the light of this limitation.
Imaging is recognized as having an important role in differentiating the various causes of dementia (Table 1). Several MRI features have positive predictive value for non-AD dementias, and have been incorporated into diagnostic criteria. The National Institute of Neurological Disorders and Stroke–Internationale pour la Recherche et l'Enseignement en Neurosciences criteria for vascular dementia,68 for example, require demonstration of vascular changes on structural imaging.
Consensus criteria for frontotemporal lobar degeneration (FTLD) include frontal and/or temporal atrophy as supportive features,64 and relatively good correlations have been observed between the FTLD subtype and the pattern of atrophy. Semantic dementia is associated with anterior (often asymmetrical) temporal lobe atrophy, progressive nonfluent aphasia is associated with left perisylvian loss, and behavioral variant frontotemporal dementia is associated with frontal atrophy.69–71 Focal or asymmetrical frontal or temporal atrophy reduce the likelihood of a diagnosis of AD.61,64
Consensus criteria for the clinical diagnosis of dementia with Lewy bodies (DLB) include relative preservation of medial temporal lobe structures on computed tomography or MRI,72,73 although substantial overlap between DLB and AD with regard to atrophy in this region73 detracts from the usefulness of this marker in individual cases. This overlap contributes to a blurring of the boundary between DLB and AD, but molecular imaging of the dopaminergic system can help to differentiate these two conditions.74
The most recent criteria for multiple system atrophy feature atrophy of the putamen, middle cerebellar peduncle, pons and/or cerebellum, as observed on MRI, as additional features of both the parkinsonian and cerebellar variants.75 Some studies indicate that T2-signal changes in the basal ganglia and brainstem on 1.5 T MRI, including posterior putaminal hypointensity, hyperintense lateral putaminal rim, the ‘hot cross bun’ sign,76 and middle cerebellar peduncle hyperintensities,77,78 could aid the diagnosis of this condition.
In Creutzfeldt–Jakob disease, MRI changes are almost pathognomonic, with characteristic patterns of high signal being observed in the basal ganglia on fluid-attenuated inversion recovery imaging, as well as changes in the striatum or cortical ribbon on DWI.79–81 This remarkable specificity of MRI for prion diseases is emphasized by the fact that variant CJD is associated with a very specific pattern of thalamic changes (the pulvinar sign) that even distinguishes it from sporadic CJD.82
Absence of vascular changes on MRI essentially excludes a diagnosis of vascular dementia according to internationally accepted criteria.68 However, a large proportion of patients with progressive cognitive deterioration show varying degrees of small-vessel disease, which manifests on T2-weighted MRI as white matter changes and one or more lacunes.83 Most individuals with progressive cognitive deterioration probably have a mixed etiology of AD and cerebrovascular changes.83,84 Estimating the proportion of cognitive impairment that is attributable to neurodegenerative versus cerebrovascular components is difficult, but nevertheless important. The larger the contribution of cerebrovascular disease, the greater the therapeutic emphasis on addressing vascular risk factors.
Clinical and epidemiological studies have shown that despite their high prevalence and tendency to progress,85,86 white matter changes account for a small fraction of the massive cognitive impairment in patients with progressive dementia,38,87 and that neurodegeneration is a more relevant determinant of cognitive decline than are white matter changes.88 A crude but meaningful estimate is that the cross-sectional contribution of severe white matter changes is equivalent to ≈0.5 points on the Mini-Mental State Examination (MMSE) in people with cognitive impairment, and the contribution of these changes to progression of cognitive deterioration is around 12 times smaller than the contribution of neurodegenerative changes (Figure 3).87 Lacunes and other silent brain infarcts more than double the risk of dementia occurring within 5 years,89 and could decrease cognitive reserve in patients who are accumulating plaques and tangles.90
Crucially, many studies of cerebrovascular factors in clinical populations of MCI and AD have, by design, excluded individuals with risk factors, history, or presence of indicators of cerebrovascular disease on MRI. To better understand the potential independent or synergistic contributions of neurodegenerative AD and cerebrovascular changes to cognitive impairment, studies focusing on both gray matter and white matter measures must be performed in representative populations. Such studies should include direct comparisons of, for example, hippocampal volume or three-dimensional cortical thinning—and possibly also amyloid imaging on PET—with cerebrovascular disease signal changes.
An intriguing imaging development has been the recognition of so-called microbleeds—small dot-like lesions with low signal in T2 images that indicate hemosiderin deposition. Studies have indicated that the prevalence of microbleeds in AD is at least 20%,91 or possibly higher if more-sensitive MRI sequences or higher field strengths are used, and may have prognostic significance.92 The contribution of microbleeds to the patient's cognitive profile and decline is poorly understood. A correlation between microbleeds and CSF Aβ levels was recently documented,93 but the relative roles of amyloid and non-amyloid angiopathy in cognitive impairment remain a matter for debate.63
The search for a valid marker to track disease progression should be viewed in the context of the development of drugs with potential disease-modifying effects. A valid marker of disease activity that has higher measurement precision than the currently used outcomes (cognitive and functional scales) might provide a surrogate outcome measure. For a biomarker to be accepted as a surrogate outcome in a clinical drug trial, it must both be correlated with the clinical outcome and fully capture the net effect of the intervention on the clinical efficacy outcome.94,95 Imaging outcomes could potentially allow meaningfully powered phase II and III clinical trials with significantly smaller patient groups and/or shorter follow-up times than are currently possible,96 thereby providing a boost to the efficacy of drug development programs. No widely accepted, valid surrogate outcome of disease progression currently exists, although preliminary data indicate that imaging measures could provide adequate power to clinical trials with far smaller samples of patients than are required if traditional cognitive and functional measures are used.21,24,96
The sensitivity of a marker to track disease progression depends on the steepness of the slope of change during the disease stage of interest, intrinsic measurement precision, and its statistical effect size: markers that have plateaued to maximal impairment or have not yet changed appreciably (ceiling and floor effects, respectively) are likely to be poor markers of progression (Figure 1).32 Sample sizes increase with the square of the SD of the rate of change of measurements in the relevant clinical group, so precision and reduced variance are key requirements to reduce sample sizes. For use in clinical trials, markers should be sensitive to change, but should also have high biological plausibility and be related to the core clinical or biological features of the disease. The available evidence indicates that structural markers fulfill many of these requirements and are, therefore, reasonable candidates for monitoring disease progression.9,15–19,21–25,97 In one clinical trial, a putative disease modifier reduced the rate of decline on serial MRI but provided no significant clinical benefit with regard to cognition.98 This finding is, perhaps, consistent with the suggestion that brain atrophy is a more precise indicator of disease progression than are clinical scales, but it raises questions regarding the clinical significance of small changes in imaging markers.23,96,99 For validation purposes, imaging changes suggesting a disease-modifying effect will need to predict longer-term clinical outcomes.
In mild AD (for example, MMSE score >20), hippocampal atrophy rates are 3–6% per year, compared with 0.3–2.2% per year in normal aging (Table 2). Hippocampal atrophy rates diverge from normal ≈5.5 years before the dementia threshold is crossed;100 that is, at a time when patients are at the MCI stage101 or soon before.28,102 The rate of atrophy correlates with CSF tau protein concentrations at baseline, offering a valuable indicator of disease progression.103 Different landmarks and tracing procedures have led to different hippocampal volume estimates,104 however, making drug effects difficult to compare across clinical trials. The variance and estimation error for atrophy rates depends on the interscan time interval: shorter interscan intervals (<12 months) give larger variances.105 Typically reported SDs of hippocampal atrophy rates in AD are 2.5–3.5% per year for a 1 year study, falling to 2.0–2.5% per year for an 18–24 month study.106 Many published studies used manual outlining of the hippocampus; however, semi-automated measures have been introduced in recent years.107,108 Fully automated (template-based) measures109–111 have not yet been used in large trials.
Rates of whole-brain atrophy in AD diverge from normal ≈4 years before the dementia threshold is crossed100 and, with few exceptions, have been estimated at 1.4–2.2% per year, whereas rates of atrophy during normal aging (for a mean age of 70 years) do not usually exceed 0.7% per year (Table 2). As with hippocampal atrophy, this differential between disease-related atrophy and normal aging offers the possibility of providing evidence for a therapeutic effect. The SD of the brain atrophy rate in AD is typically ≈1% per year for a single-center study when assessed over 1 year or more.106 The seemingly surprising finding of greater whole-brain atrophy rates in the treated patients of the AN1792 vaccination trial112 can be interpreted in light of the limited prior knowledge of the structural effects of the drug in an Aβ-laden brain.113 Ongoing efforts aim to bridge this knowledge gap by developing disease markers in animal models that are homologous to the markers shown to be sensitive to disease progression in human trials.114
Ventricular enlargement is a highly reproducible measure of disease progression, owing to the high contrast between the CSF and the surrounding brain tissue on T1-weighted images. The ventricles expand by 5–16% per year in patients with AD and 1.5–3.0% per year in healthy elderly individuals (Table 2). The relatively low error variance gives this measure excellent power to detect consistent changes over short follow-up intervals (for example, 6 months),99,105 but this relative advantage over whole-brain measures dissipates if the follow-up interval is extended to 1 year or more.105 The high accuracy and feasibility of ventricular dilation measurement, even on CT,115,116 make this an excellent candidate measure for trials where MRI is not available or feasible.
Sample sizes required for brain, hippocampal and temporal lobe measures are similar. Around 100–200 individuals per treatment arm should provide 90% power to detect a 20% slowing of atrophy.24,25,105,117 By also screening out AD patients with low CSF tau or high Aβ42 levels or high hippocampal volumes, and by accounting for baseline covariates (age, baseline hippocampal volume, and apolipoprotein E genotype), the necessary sample size for a 1 year trial that uses hippocampal atrophy as an outcome measure has been estimated to be as low as 50 individuals per arm.118 Rates of whole-brain and hippocampal atrophy and ventricular enlargement have been included in several clinical trials of candidate disease modifiers in AD26,98,113,119–121 To support disease-modification claims, virtually all phase III clinical trials now include serial structural MRI scans from which rates of atrophy are computed. However, power estimates for trials and drug registration are still based on traditional clinical cognitive and functional tests, owing to the requirements of regulatory agencies and concerns over surrogate outcomes. This current situation can be partly attributed to the lack of a close relationship between markers and the fundamental features of AD neuropathology; that is, plaques and tangles (Table 3). Indeed, hippocampal atrophy, ventricular dilation and whole-brain atrophy are not specific to AD, or even to neurodegeneration, as these findings are also present to some degree in healthy aging.99,122,123 The three-dimensional pattern of gray matter loss or cortical thinning might, however, be more disease specific (Figure 4), and cortical thinning correlates cross-sectionally and prospectively with clinical symptoms.124,125
Contrary to the assumptions of many studies performed to date, atrophy is not a linear phenomenon. In familial AD mutation carriers, whole-brain atrophy accelerates by ≈0.3% per year in the years leading up to diagnosis (for example, before the MMSE score falls below 23 out of 30).117 Whole-brain and hippocampal atrophy accelerate in sporadic AD during the transition from normality to cognitive impairment,126 and as individuals with MCI progress to AD dementia.127 Cross-sectional data indicate that the pattern of accelerated atrophy coincides with functional networks that are progressively affected by the disease.128 The acceleration of atrophy may be followed by deceleration.14,128 Mapping and modelling the dynamics of the three-dimensional pattern of atrophy over the disease course by use of nonlinear models is likely to provide an in vivo marker of the progression of neuropathology.
Structural imaging based on MRI is an integral component of the clinical assessment of patients with suspected AD. Structural MRI markers now support earlier and more-precise diagnosis and measurement of progression. The presence of atrophy of medial temporal structures is a partially validated candidate marker for early diagnosis of the disease at the MCI stage. Rates of whole-brain and hippocampal atrophy are sensitive and powerful markers of progression of neurodegeneration and, as a result, are increasingly used, along with clinical metrics, as outcomes in clinical trials of potential disease-modifying therapies. Measures of cortical thinning and automated classification approaches that assess the overall pattern of atrophy seem to show promise for the diagnosis of AD.10
The newly proposed diagnostic criteria for AD43 must be validated in multiple large data sets. Positive test results for all imaging and biochemical markers might be highly specific for AD,60,129,130 but the yield of combinations of the less invasive techniques (for example, structural MRI, 18F-fluorodeoxyglucose PET131 and PET amyloid imaging132), less costly approaches (for example, structural MRI and CSF biomarkers), or other relatively new or less validated MRI markers (for example, spectroscopy of the posterior cingulate cortex133 and diffusion changes in the temporal lobe134) will need to be evaluated empirically. Standardized operating procedures—for example, for manual hippocampal volumetry135—and automated algorithm pipelines will facilitate the comparison and integration of results across laboratories and studies, and will enable the generation of normative values. Such values are needed to achieve the ultimate goal of individual diagnosis with a single cross-sectional MRI scan by comparing it with a large number of normal scans, as has been done in other imaging modalities.136 Large multicenter studies will be needed to compare different imaging and nonimaging markers.137
The natural evolution of structural brain changes, and their relationship with nonstructural markers, should be investigated at the asymptomatic stage preceding MCI. Descriptive studies of the longitudinal changes in both structural and nonstructural markers in patients who have undergone accurate clinical assessment will help to identify the most accurate combination of markers for early diagnosis, and to estimate the time window within which the slope of change (and variance in that slope) renders the marker suitable for tracking disease progression and acting as an outcome measure in clinical trials. Validation of the many automated algorithms claiming to provide accurate marker measures will contribute to wider clinical use and, ultimately, replacement of human ratings and tracing. Defining the time course of marker change will also shed light on the contributions to cognitive impairment of genetic and/or neurodevelopmental effects, normal aging, and neurodegenerative changes. The development of homologous animal and human markers of disease progression will aid the accurate prediction of drug effects in human trials.
Structural imaging changes lie at the crossroads between the molecular pathology of AD and the clinical and cognitive decline that follow from that pathology. Structural imaging is well placed to contribute to improved early diagnosis of AD and to the search for treatments to slow or prevent this devastating disease.
Articles were selected on the basis of the authors' personal knowledge and the following PubMed searches: “(MR OR MRI) AND atrophy AND Alzheimer*[ti] AND (diagnosis[ti] OR diagnostic[ti])”; “(MR OR MRI) AND atrophy AND (“frontotemporal dementia”[ti] OR “frontal lobe dementia”[ti] OR “frontotemporal lobar degeneration”[ti] OR “dementia with Lewy bodies”[ti] OR “Lewy body dementia”[ti] OR “progressive supranuclear”[ti] OR “Steel–Richardson”[ti] OR “multiple system atrophy”[ti] OR “Parkinson dementia”[ti])”; and “(MR OR MRI) AND (longitudinal OR prospective OR serial) AND atrophy AND Alzheimer*[ti]”; and “(MR OR MRI) AND (“subcortical cerebrovascular” OR (“white matter” [ti] AND (hyperintensities[ti] OR changes[ti] OR lesions[ti]))) AND cognit*[ti]”. Reference lists of the identified papers were examined for further leads. The search was limited to full-text manuscripts published in English over the past 15 years. The final selection was based on relevance, as judged by the authors.
Anna Caroli, Enrica Cavedo, Rossana Ganzola, Marco Lorenzi, Michela Pievani, Annapaola Prestia, Alberto Redolfi and Cristina Scarpazza helped in the literature review, Rossana Ganzola helped in the typesetting of the manuscript, and Alberto Redolfi produced the maps of Figure 4. G. B. Frisoni was supported by FP7 neuGRID, outGRID and IMI Pharma-Cog. N. C. Fox is supported by Medical Research Council grants G0801306 and G0601846, NIH grant U01 AG024904, Alzheimer Research Trust grant ART/RF/2007/1 and the National Institute for Health Research. C. R. Jack is supported by National Institute on Aging grant AG11378 and the Alexander Family Alzheimer's Disease Research Professorship. P. M. Thompson was supported by NIH grants EB007813, EB008281, EB008432, HD050735, AG020098, P41 RR013642 and M01 RR000865.
Competing interests: N. C. Fox declares associations with the following companies: Abbott Laboratories, Elan Pharmaceuticals, Eisai, Eli Lilly, GE Healthcare, IXICO, Lundbeck, Pfizer, Sanofi-Aventis, Wyeth Pharmaceuticals. See the article online for full details of the relationships. The other authors declare no competing interests.
Giovanni B. Frisoni, IRCCS Centro San Giovanni di Dio FBF, via Pilastroni 1, 25125 Brescia, Italy.
Nick C. Fox, Dementia Research Center, UCL Institute of Neurology, Box 16, Queen Square, London WC1N 3BG, UK.
Clifford R. Jack, Jr, Mayo Clinic, 200 1st Street SW, Rochester, MN 55905, USA.
Philip Scheltens, Alzheimer Center and Department of Neurology, VU University Medical Center, Postbus 7057, 1007 MB Amsterdam, The Netherlands.
Paul M. Thompson, Laboratory of Neuro Imaging, Department of Neurology, 225E Neuroscience Research Building, 635 Charles Young Drive, UCLA School of Medicine, Los Angeles, CA 90095, USA.