Structural Imaging in Early Pre-States of Dementia

Abstract

In this review focus is on structural imaging in the Alzheimer’s disease pre-states, particularly cognitively normal (CN) persons at future dementia risk. Findings in mild cognitive impairment (MCI) are described here only for comparison with CN. Cited literature evidence and commentary address issues of structural imaging alterations in CN that precede MCI and AD, regional patterns of such alterations, and the time relationship between structural imaging alterations and the appearance of symptoms of AD, issues relevant to the conduct of future AD prevention trials.

1. Introduction

Interpretation of structural imaging in normal persons depends critically on beliefs concerning the individuals being imaged. Clarity in key general terms is needed before proceeding with a review of this topic, particularly because the same term may be used in both a theoretical and a practical sense in the literature without distinction. Cognitive normality, for example, may denote concepts of a particular range of expected behaviors attributed to brain performance together with theoretical factors contributing to variability in this performance. In practice cognitive normality may be assessed using interviews that probe relevant behaviors based on clinical knowledge and experience, or defined by scores on cognitive tests [1-3].

The mutual interaction between the theoretical and practical sense of cognitive normality (CN) in persons being imaged is revealed in its application to aging [4]. Is CN the ideal preservation of adult brain performance at any time point, in which case psychometric test norms might be independent of age, or should CN include acknowledgement of universally observed declines with age, particularly in certain domains such as processing speed, and hence interviews and cognitive tests should somehow adjust for age [5]?

A second general term relevant to the interpretation of structural imaging in cognitively normal persons at risk of AD is atrophy. Atrophy as a medical concept refers to the reduction of brain substance relative to a presumed prior healthy normal state, and in addition implies this reduction is an effect of disease rather than from a normal process, e.g., pruning due to developmental changes during puberty. Some studies are designed to assess longitudinal change and thus define an atrophy rate, but cross-sectional studies have to make an assumption of comparability with the normal control group when referring to atrophy [6].

A second issue related to atrophy is uncertainty in the apportionment of structural (e.g., volume, thickness or shape) differences between: (1) longstanding effects of brain development or prior injury, (2) ongoing alterations due to aging or to disease processes that are not primary to the study, and (3) the focus of the study, in the current instance risk of clinical AD. By clinical AD is meant the expression of the underlying pathophysiology of AD in the form of memory or other symptoms. Developmental endowments, life experiences or specific activities such as exercise could both protect the individual from clinical AD and subtly influence brain structures, e.g., by increasing cortical thickness, thus decreasing risk of AD even though they have no direct connection to AD pathophysiology [7]. The essence of the problem is whether one is “finding the beginning or predicting the future” [8]. However, the issue of apportionment is often not explicitly considered in the literature.

Finally, the general term “risk of AD” needs to be qualified by being specific about the time interval over which risk is defined. It is important to recognize the observation period of individual imaging studies because if they are short, there is little time for the underlying biology of AD to express itself, whereas if it is long, there is greater likelihood of cognitive decline or AD expression but also greater effects of competing risks, such as co-morbid cardiovascular disease. Likewise with the term “prediction”, but that term is avoided herein.

This review will attempt to be mindful of the general terms CN, atrophy and AD risk as describe above, while addressing the following questions: (1) Are there structural imaging alterations in CN that precede MCI and AD?, (2) Is there a regional pattern of such alterations and does this pattern provide insight into AD pathophysiology?, and (3) How long before symptoms of AD appear do structural alterations occur? A context for addressing these questions is provided first in terms of a simple model and a brief description of brain structural alterations in aging.

2. Conceptual Model for AD Pathology & Expression

A basic framework for this article is illustrated in Figure 1. In this simple model, the basic assumptions are that: (1) underlying AD neuropathology accumulates in many CN persons years before any cognitive symptoms or even measurable declines in cognitive testing occur, (2) the brain has intrinsic neural mechanisms that compensate for injury due to AD neuropathology (ADN) at all stages, e.g. by synaptic remodeling, and (3) the human brain has inborn, developmental and learned sources of cerebral reserve that may offset the cognitive effects of injury, or, inversely, specific liabilities to the effects of ADN.

Figure 1.

Simple Model of the AD Timecourse (Modified from [17]). A. Illustration of a statistical cohort of 100,000 persons beginning at age 25 years, with bar height representing the expected remaining population at each 5-year interval to age 85 based on the US census [88, 89]. Each bar divided into proportionate parts representing the expected number of persons with AD (black) or MCI (white), or who remain normal (grey), based on US population model estimates [89, 90]. The incidence of MCI and AD, and overall death rate accelerate after age 55. B. Percentage of Braak stages over the same age range based on Braak [89, 91, 92]. C. Conceptual diagram of the appearance of different image features by modality and age in a nominal cohort of persons who develop symptoms of AD at age 75 years, based on results in the current literature [17]. Dotted lines represent the variable appearance in individuals of the first detectible image changes within each modality, some of which may precede symptoms by several years. Solid lines represent the more certain appearance of alterations near the time symptoms appear (within a few years). The character and distribution of image alterations may change over time, e.g., in functional MRI activation, a subtlety not captured in this simplified diagram. D. In the nominal cohort, memory decline within the normal range of scores becomes detectible by sensitive tests and longitudinal assessment, later associated with a complaint noticed by patient and outside observers, but without functional decline. Differences in rate of decline within the normal range may be dissected out using imaging and other biomarkers of AD neuropathology. These differences could result in a re-definition of CN.

To clarify the framework, more needs to be said about “underlying AD pathology”. Here the essence of AD pathology is taken to be the coupling between amyloid and tau. In this view beta-amyloid is not necessarily in the form of an accumulation (as in plaques) particularly in the earliest stages [9], but it must be present and coupled to abnormal tau for there to be AD pathology. This coupling could be mediated through, for example, prion protein [10, 11]. Accumulations of both amyloid and tau are the evidence this coupling has occurred, and indicate the severity of the process and likely the duration. Atrophy is taken as evidence of synaptic and neuronal loss or shrinkage. These changes are accompanied by widespread tau aggregates within the affected region. The amount of amyloid in a region does not always correspond with atrophy – in this framework that is determined by the strength of the coupling with tau (the signature of an underlying process of synaptic and neuronal degeneration).

Neuropathologically amyloid alone is not sufficient. Most radionucleide-label imaging only assesses amyloid, so the coupling to tau is only a presumption. There is a strong correlation between tau measured in CSF and (decreased) CSF amyloid and (increased) brain amyloid. So while not perfect, the amyloid-tau relationship is strong. However, combined amyloid-tau labels such as FDDNP (2-(1-{6-[(2-[F-18]fluoroethyl)(methyl)amino]-2-naphthyl}ethylidene)malononitrile ([(18)F]FDDNP)) may in future prove better markers of AD pathology than amyloid-only agents [12]. Tau alone may be present in other conditions, but it is the coupling with beta-amyloid that is distinctive of AD pathology. Whether a decoupling might occur later in the process wherein further neurodegeneration becomes autocatalyzing is unknown.

Within this framework the mere presence of AD-like brain pathology by itself doesn’t provide enough information to predict later dementia. It could be that the rate of accumulation is low, at least low enough to be successfully compensated, or that factors leading to an acceleration of the AD neuropathologic process are absent. Additional evidence that detected AD pathology is causing brain dysfunction may be needed in early cases, e.g., subtle cognitive impairment or at least decline, or brain network disruption detected by functional studies. However, it is reasonable to assume that the more pathology that is present, the more likely it is to be predictive of later AD, and that the absence of pathology is prognostically favorable [13].

Alzheimer’s disease occurs in late life and thus in an environment of multiple competing causes of death (Figure 1A). The earliest Alzheimer’s disease neuropathology (Figure 1B; nominal age range 55-65 years) occurs in middle adulthood, accompanied by “compensatory” imaging functional changes (Figure 1C) but unaccompanied by brain volume loss on imaging or by diminished cognitive performance (Figure 1D; [14]). As age and ADN increase (nominal age range 65-75 years), compensatory functional changes accompany early regional brain atrophy associated with increasing global amyloid and regional tau accumulation. This stage represents the borderland between CN and early cognitive changes (memory decline +/− MCI). Subsequent symptoms of cognitive impairment represent a critical point beyond which functional alterations, brain atrophy, and cognitive performance continually worsen (nominal age in this diagram 75 years). Functional losses in addition to compensations, increasing atrophy, and diminishing cognitive performance ensue in diagnosed clinical AD [15-17].

3. Structural Imaging in Aging

Well-characterized, longitudinally followed cognitively normal persons undergo brain volume losses with age, emphasizing the need for carefully selected age-matched controls in cross-sectional structural studies [18-20]. Only longitudinal studies can confirm baseline normal individuals who develop cognitive decline, MCI or AD at some future time. Retrospectively removing from analysis baseline structural scans from CN subjects who later develop MCI or AD could reduce effects of subtle disease-related volume losses, on the assumption these scans could skew regional volume distributions for CN [21]. Even when this step is taken, age-associated volume reductions are still observed. Grey matter volume reductions with age are generalized, without the relatively selective regional medial temporal, lateral temporal and parietal reductions characteristic of AD. Selective volume reduction with age also occurs in the deep frontal white matter and genu of the corpus callosum [21, 22]. In one recent study cortical thickness was reduced predominantly in primary sensorimotor regions, but did not correlate with cognitive performance, whereas white matter integrity (assessed by fractional anisotropy) was reduced in genu of corpus callosum and in frontal white matter, and correlated with deficits of “cognitive control” in 38 CN elders [23].

Grey matter volume is not a static property and is potentially modifiable in aging. Erickson et al found that vigorous levels of baseline physical activity (operationalized as blocks walked per week) were associated with increased cortical volume, particularly in premotor and prefrontal regions, nine years later [24]. Grey matter volume increases are seen in aging even after short exercise interventions [25]. The main caveat to these studies is the possibility that disease associated with brain volume losses biases toward non-participation in exercise.

Studies of hippocampal function (using cerebral blood pool volume measurements, CBV) suggest that losses with aging, including memory decline, may occur in the subiculum and dentate nucleus, whereas entorhinal functional changes are more specifically associated with disease [26], particularly AD. Diabetes may be associated specifically with dentate gyrus functional decreases, hippocampal ischemia with CA1 decreases, and more generalized ischemia both with CA1 and subiculum sector decreases [27]. A structural study of hippocampal formation subfields in subjects carefully screened for vascular risk and diabetes demonstrated an age-related reduction in subiculum volume commensurate with global grey matter reductions, but relative preservation of other hippocampal segments (CA1 and dentate subfields; [28]). These results are consistent with the pattern of subfield changes associated with specific pathologies and with aging seen in CBV-based studies.

Involutional age-related erosion of brain reserve and mild cognitive decline appears to be a fact of life in brain aging [16]. However, more sensitive methods for evaluating grey and white matter structure, and finer resolution of these methods to the level of individual grey matter nuclei and subfields, are now demonstrating subtle structural alterations that can be apportioned to specific co-morbid pathologies. As discussed further on, risk factors for pathologies, e.g., maternal family history of AD, may be associated with alterations in grey matter in CN aged subjects. Thus, apportionment of structural alterations between aging involution and underlying pathologies may soon be placed on a better scientific footing. Finally, effects of intervention such as exercise show that the aging brain may demonstrate plastic re-growth detectible as structural change.

4. Structural Imaging in MCI

MCI is defined as a mild but definite decline from previous cognitive ability, confirmed by a reliable observer, and substantiated by deficits on neurocognitive testing [29-31]. Generally structural MRI studies have demonstrated decreased size of the hippocampal formation in MCI, by rating techniques [32] and in both region-of-interest [32-36] and voxel-based volumetric studies [32, 37-41]. Studies are mixed regarding the presence of more global grey matter reductions in MCI compared to CN, e.g., in frontal and parietal regions [32, 42, 43]. Older adults who have memory complaints but do not meet formal criteria for MCI may nonetheless show similar grey matter reductions [32, 44].

A recent study based on subjects from the landmark Alzheimer’s disease Neuroimaging Initiative (ADNI) demonstrated that structural imaging data added significantly to the prediction of AD in patients with amnestic MCI [45]. The study used a modified version of freesurfer software [46] to compute cortical thickness and volumes from salient regions in subjects comparing CN and AD at baseline and over time using discriminant analysis. The discriminant functions were then used to compute risk scores for the MCI subjects. Subjects with high percentile scores were much more likely to convert to AD (odds ratios of 7.2 for highest vs. lowest percentiles for baseline scans, increasing to 12.0 when one-year longitudinal change was included in the analysis).

5. Structural Imaging in CN Persons at Risk of AD

5.1. Overview

This section refers to imaging studies characterizing the brain in cognitively normal (also denoted by CN in context) subjects with the goal of characterizing risk for MCI or AD (cf. Figure 1). Studies have operationalized CN by using expected cognitive test scores as they must, as discussed in the introduction. But this approach may not answer the question of whether the subjects “truly” have CN from an evolving theoretical point of view based on neurobiology of AD, because there is always a better cognitive test, clinical interview, or longitudinal follow-up method for assessment of cognitive decline that might yet detect a subtle difference. This is a practical problem only if this potential difference would alter treatment or prevention, an important unresolved issue. Despite uncertainty, if risk stratification in defined CN subjects could be accomplished with reasonable accuracy using structural imaging, then interventions might be made to slow the progress of ADN or to enhance intrinsic brain compensatory mechanisms before symptoms, perhaps entirely preventing or at least delaying these symptoms [47, 48]. This consideration motivates the search for structural, and other, biomarkers of AD in CN persons.

5.2. Brain Structure and Amyloid

The issue of defining CN mentioned in the introduction has a specific application to presumed pre-state Alzheimer’s disease (AD), because it has been found that brain amyloid accumulation is correlated negatively with memory scores in CN, although the scores are in the normal range [49]. Nonetheless, cognitive differences in general were not found between the 20-30% of CN with positive amyloid scans and the remainder with amyloid negative scans [13, 50]. In a group of 95 CN subjects, 32 were found to have elevated 11C-PIB amyloid brain labeling (PIB-positive; 34%). Again there were no cognitive differences or mean hippocampal volume differences between PIB-positive and PIB-negative CN, but there was a negative correlation between hippocampal volume and inferior temporal PIB retention in the former group [51]. Other recent studies have demonstrated decreased volume [52] or cortical thickness [53] in regions associated with early AD pathology (hippocampus, temporal neocortex, anterior and posterior cingulate) in CN subjects with increased amyloid burden. It was structural loss and not amyloid burden that correlated with cognitive performance in these studies, a finding that can be interpreted within our framework as the result of amyloid coupling to synaptic and neuronal degeneration associated with abnormal tau.

These findings suggest that in its early stages amyloid accumulation may not be associated with cognitive alterations but rather with subtle asymptomatic structural volume loss, implying ongoing brain injury that is somehow compensated. Evidence that structural alterations are associated with cognitive performance in CN subjects with increased amyloid burden, and amyloid burden with subsequent cognitive decline, suggests this state of compensation is not long-lived [52]. The reverse conclusion, that amyloid accumulation is the consequence of volume loss, is less likely because there was no correlation of between brain volumes and amyloid burden in the frontal lobes, for example [51]. Nonetheless, amyloid accumulation correlating with hippocampal volume was located exclusively in local inferior temporal regions immediately upstream of the entorhinal cortex and hippocampus, not in other neocortical regions (Figure 2).

Figure 2.

Voxelwise negative correlation between amyloid burden in normal subjects with elevated global ¹¹C-PIB binding and hippocampal volume (r = −0.6, p = 0.02 using inferior temporal region). This relationship was not seen in subjects with MCI or AD, or in other brain regions outside of the inferomedial temporal lobe and parieto-occipital junction. With permission from [51]. Views: a- sagittal; b- coronal; c- axial.

Amyloid positive subjects are CN in one practical sense but not necessarily in a theoretical sense, because there is evidence of an underlying disease process, a process excluded from the concept of aging, which might cause the observed disease biomarker correlations with memory scores or with brain volumes. In this grey area lies the borderland between very early mild cognitive impairment (MCI) and CN aging.

5.3. Longitudinal Structural Studies

Annual rates of ventricular expansion have shown an increase in baseline CN subjects who received a clinical diagnosis of MCI within an average of 2.3 years, with an apparent acceleration of ventricular volume expansion rate in the two-year lead-up to MCI diagnosis [51, 54]. In a second study with 3.4 years follow-up from the last structural scan, ventricular expansion rate was associated with subsequent MCI diagnosis [55]. Volume loss rate in whole brain, hippocampus and entorhinal cortex was not associated with MCI conversion in this study, but regional medial temporal boundary shift atrophy rate has been associated with this conversion over 6 years observation [56].

Because the earliest ADN involves entorhinal cortex, amygdala, hippocampus and other medial temporal structures, early atrophy in these regions could predict normal subjects who will later develop MCI or AD [57-61]. Medial temporal (particularly entorhinal) baseline volume also predicted decline on tests of delayed memory in CN subjects imaged at baseline and followed for 3.8 years [61, 62].

An unexpected and significant early finding in the ADNI study was that temporal lobe atrophy predicted subsequent cognitive decline even in the baseline cognitively normal participants in that study [41, 61]. It has since been found that rate of atrophy in medial temporal structures in these subjects was associated with cognitive decline assessed by logical memory and AVLT tests [61, 63]. These studies have been supplemented by analyses demonstrating cortical thinning can be detected over one year in normal subjects, particularly in temporal and prefrontal regions [64]. These reductions were not confined to subjects with low CSF beta-amyloid levels, a marker of potential AD neuropathology, but in subjects with levels below a threshold, a correlation with cortical thinning was evident [65].

A longitudinal general community study of 511 nondemented subjects demonstrated that reduced global, hippocampal and amygdalar volumes at baseline were predictive of dementia within average 6 years [66], with rate of volume change serving as an independent predictor [66, 67]. Another recent study on 136 longitudinally-followed normal subjects undergoing a baseline structural scan demonstrated that volume in a combined region identified from these scans using VBM (anteromedial temporal and left angular gyrus grey matter; Figure 3A) predicted MCI within five years with 76% accuracy. Predictive accuracy increased to 87% by combining volume in the combined region with a cognitive measure, raw Wechsler Memory Scale score [66, 68, 69]. A study using a subset of these subjects for manual medial temporal tracing further supported these results [70].

Figure 3.

Brain structural alterations in cognitively normal persons at risk of MCI or AD. A. Volume within a template identified by retrospective classification of baseline scans in CN subjects demonstrating a region of significantly reduced grey matter volume at baseline in those who later developed MCI within 5 years (yellow-white color scale). Views: a,d- sagittal; b- coronal; c- axial. Prediction accuracy of volume within this brain region alone was 76%, and 87% when combined with psychometric data (With permission from [21]). B. Decreasing rates of conversion from CN to AD over the observation period by average baseline cortical thickness, averaged across nine regions of interest, classified as low (1 SD below the entire cohort mean), average (within 1 SD of the mean), and high (1 SD above the mean – the rate was zero in this group). Regions of interest were selected by finding significant differences in cortical thickness between AD and CN in a prior study (inset left, nine color-coded regions projected on partially inflated brain surface image). Retrospective classification of baseline scans in CN subjects allowed calculation of thickness differences between baseline CN who developed AD vs. those who remained CN (With permission from [71]).

A second potential structural signature of AD was found in a collection of cortical thickness measures in two independent groups of CN subjects using a retrospective comparison between subjects who developed AD versus groups of 25 control subjects who remained normal over the same observation periods of 7.1 (7 converters) and 11.1 (8 converters) years [71]. The sample sizes are relatively small and thus findings are prudently viewed as preliminary. The regions for cortical thickness comparison were established from previous studies of AD. Although medial temporal lobe (MTL), temporal pole, and superior frontal gyrus thickness was significantly less in CN to AD converters, averaging MTL with six additional region thickness values appeared to add further predictive power. The additional regions were inferior temporal, angular gyrus, supramarginal gyrus, superior parietal, precuneus, and middle frontal gyrus regions segmented in freesurfer. High cortical thickness in some of these regions could have been protective for conversion to AD, because the conversion rate in the18% of subjects with this signature was zero (Figure 3B).

5.4. Structural Studies in CN Involving Genetics

Normal subjects with a maternal history of AD (+FHm) have demonstrated decreased cortical volumes compared to subjects with a paternal AD history or no family AD history, mainly in frontal regions [72, 73]. Grey matter loss over two years in +FHm, however, displayed a somewhat different pattern, incorporating the dorsal frontal lobes but also the precuneus and inferotemporal cortex. A maternal inheritance influence on both earlier cortical development and later degenerative change is one explanation for these findings.

Apolipoprotein E (APOE) is a cholesterol carrier linked with cardivascular disease and late-onset AD [74], particularly though the influence of the epsilon-4 allele on age of AD onset. Cognitive decline in CN APOE epsilon-4 carriers diverges from that of non-carriers before age 60 years according to one model based on longitudinal observation of approximately five years [75]. APOE allele status was not found to add to risk of future cognitive impairment in CN subjects over an observation period of six years when medial temporal volume, walking speed and logical memory II baseline scores were included in the analysis [76]. Increased amyloid burden has been associated in a dose-dependent manner with the epsilon-4 allele [77] in CN subjects. Reduced grey matter density in the right medial temporal lobe and in frontotemporal cortex has been reported in of CN carriers of a singe epsilon-4 allele [78]. These findings suggest that APOE risk in CN subjects has proxies through amyloid accumulation and tissue loss in grey matter areas known to harbor neuropathology early in AD.

Familial Alzheimer’s disease (FAD) involving single autosomal dominant gene mutations with early-onset dementia has been studied to provide insights into the far more common sporadic, polygenetic disease late-onset Alzheimer’s disease (LOAD). Mutations in the amyloid preursor protein (APP) gene on chromosome 21 are quite rare. One study of presymtomatic APP V717G mutation carriers demonstrated slight baseline left hippocampal volume difference compared to controls with a rapid hippocampal volume loss (4.8% per year, n=3) as symptoms appeared over two years [79]. These subjects retrospectively had lower than expected global or memory scores despite normal day-to-day function and absence of reported memory decline. For comparison a later study in the same family using a global brain volume technique, median global annualized volume change was 0.2 %/year in 26 matched controls, 0.1%/year in 23 at-risk carriers who remained well, and 1.5 %/per year in 5 who became symptomatic within two years of baseline scan [80]. These findings suggest the timeline for structural alterations in FAD due to APP mutations may be compressed compared to LOAD, with accelerated brain losses occurring in parallel with asymptomatic cognitive decline on testing, and the appearance of symptoms at an early age.

Of the three known FAD genes, the presenillin-1 (PSEN1) gene on chromosome 14 harbors the most common mutations. A study of six CN PSEN1 mutation carriers gathered from four families found increased cortical thickness in precuneus and parietotemporal regions approximately 10 years before expected symptoms, whereas symptomatic subjects demonstrated diffuse cortical thinning, including areas of increase found in asymptomatic subjects [81]. Other studies in FAD demonstrate cortical thinning in the precuneus at 4.1 years prior to symptoms, and at 1.8 years in posterior cingulate in mutation carriers [82]. Thus the full trajectory of structural change in CN persons at risk of FAD is incomplete, and might include a very early phase of cortical swelling or enlargement.

A large family harboring the E280A PSEN1 mutation has been identified in Columbia, South America and characterized from the earliest, pre-symptomatic stages through dementia, serving as a model for sporadic AD [83-85]. Structural studies in this large kindred (449 identified mutation carriers) have not yet been reported but will be extremely important because these subjects are likely to participate in clinical prevention trials in the near future. Family members who are PSEN1 carriers have a cognitive phenotype that parallels LOAD, a decade time scale for evolution of symptoms from earliest cognitive alterations, and certainty of dementia. In contrast, LOAD has no model that can approach the predictive power of an autosomal dominant mutation in patient selection for prevention trials [82, 86]. Thus, in addition to the great medical significance of prevention in this particular family, there are many important implications of a successful prevention trial in this kindred for the worldwide family of the AD-at-risk.

6. Summary and Future Directions

In summary, responses to the three questions posed at the beginning of this review are:

(1) Are there structural imaging alterations in CN that precede MCI and AD?

Approximately half of normal persons harbor Braak stage I-II neurofibrillary pathology by age 55 years, approximately 20 years before nominal onset of AD at 75 years (Figure 1B). The neuropathologic burden increases progressively with age thereafter, at least in many persons. The exact manner of this progression is however unknown – whether it is linear, stuttering in bursts, or accelerating at some time. These unknowns become more important as we focus attention on AD neuropathology in CN subjects. Structural alterations in CN are observable well before dementia onset, and, when these alterations involve grey matter, may be associated with subtle cognitive decline in groups of clinically normal persons. Grey matter losses during this pre-state are also associated with fibrillary amyloid burden in the brain assessed by imaging or CSF studies, but these losses are more likely related to the amyloid coupling to abnormal tau than to amyloid per se. White matter diffusion changes are observed perhaps even earlier in interhemispheric and intrahemispheric long association tracts (cf. companion review on white matter neuroimaging in this issue).

(2) Is there a regional pattern of such alterations and does this pattern provide insight into AD pathophysiology?

The most consistent finding is a decrease in medial temporal volume well before clinically symptomatic stages (MCI or AD). However cortical thinning has been detected in other regions in CN subjects, particularly the temporal and parietal neocortex. Grey matter atrophy appears in areas where AD neurofibrillary pathology is known to be heaviest, and is correlated both globally and locally with amyloid burden. Under our framework this association is mediated by the coupling between amyloid and abnormal tau, the latter representing the marker for synaptic and neuronal shrinkage and loss. For example, CN individuals with increased global amyloid burden demonstrate strong correlation between regional amyloid burden in the inferolateral temporal lobe and hippocampal atrophy. Thus structural imaging in grey matter, by itself, only confirms gross shrinkage of neurons and neuropil in a pattern consistent with expected local AD neuropathology. When marked by the detection of amyloid globally and locally, the strength of the inference that underlying AD neuropathology associated with abnormal tau increases. Insight into how this coupling might come about is not forthcoming from these findings alone.

White matter alterations in association fibers appear to involve large stretches of neocortex, but a more detailed review appearing elsewhere in this volume qualifies this conclusion. Early work suggests that white matter alterations may be reporting on altered axonal-myelin signaling interactions, perhaps an earlier indication of underlying AD neuropathology than grey matter atrophy.

(3) How long before clinical symptoms of MCI or AD appear do structural alterations occur?

Approximately 5-10 years for grey matter atrophy, and perhaps longer for white matter diffusion alterations. This estimate suggests that interventions to prevent AD will need to begin early indeed, before the expected onset of even the earliest symptoms of cognitive impairment.

One understudied area of structural imaging is analysis of dynamic compensatory change (e.g., due to exercise or intensive educational activity) or components of cerebral reserve such as cortical thickness that may confer resistance to the effects of aging, AD, and co-morbid pathologies. Measures to enhance compensation for the presence of existing early AD neuropathology is an underexplored area of research in AD.

Many other questions regarding AD neuropathology in CN remain, some of which can be addressed with structural imaging. The answers are important for the rational consideration of AD prevention. The data presented here and elsewhere in this volume are sufficiently convincing that, for the purposes of including and excluding potential participants in clinical prevention trials, a combination of: (a) amyloid/FDDNP imaging [12, 52] or CSF [87] and (b) a structural MRI pattern signature [67, 68, 71] would enhance the efficiency and speed of these trials. In a long-term trial, even expense might be reduced.

*Highlights.

• ➢Review focuses on MRI imaging in normal subjects before symptoms of dementia
• ➢Structural MRI changes are found in normal persons at risk of Alzheimer’s Disease
• ➢These findings are important for prevention of dementia