Clinical applications of neuroimaging in patients with Alzheimer's disease: a review from the Fourth Canadian Consensus Conference on the Diagnosis and Treatment of Dementia 2012

In May 2012, the Fourth Canadian Consensus Conference on the Diagnosis and Treatment of Dementia brought together in Montreal experts from around Canada to update Canadian recommendations for the diagnosis and management of patients with neurodegenerative conditions associated with deterioration of cognition. Multiple topics were discussed. The present paper is a highly condensed version of those recommendations that were produced to support discussions in the field of neuroimaging for clinical diagnosis of those conditions.


General considerations on the imaging group from the Fourth Canadian Consensus Conference on Diagnosis and Treatment of Dementia
Functional and structural neuroimaging in dementia was thoroughly reviewed in two papers from the last Canadian Consensus Conference on Diagnosis and Treatment of Dementia (CCCDTD) conference, published in 2007 [1,2]. Th e fi eld has since grown rapidly through refi nements of techniques previously discussed and through deployment of new strategies such as in vivo amyloid imaging. Th is article, a much abridged version of the thorough review papers prepared for discussion at the CCCDTD4 conference, covers the most signifi cant of those developments. Th e original articles can be found online [3].
Th e primary goal of the CCCDTD meetings is to elaborate practical recommendations for Canadian clinicians dealing with patients presenting with cognitive impairment. Other groups have of course come up with such recommendations. However, the present paper diff ers from those others in many ways. First, this paper certainly has the advantage of having reviewed a literature that is more recent (in a fi eld in rapid fl ux) than that on which publications from the previous CCCDTD meetings were based, as well as those from organizations such as the American Academy of Neurology [4], the UK National Institute for Health and Clinical Excellence [5] or the European Federation of Neurological Societies [6]. Th e paper also had access to the most recent eff ort of the same nature, from the National Institute on Aging-Alzheimer's Association [7,8], which is tailored to a very diff erent medical practice environment.
Th e CCCDTD4 conference was attended by a group of individuals selected based on their known clinical expertise as decided by the conference organizers, who themselves are pre-eminent Canadian experts in the fi eld of dementia, and who also organized previous meetings in that series. For the imaging section, a mix of imagers and clinicians, as well as basic scientists, were tasked with the preparation of the document to be discussed. Th e general attendance at the conference also included clinicians from the specialized (neurologists, geriatri cians) and general medicine communities, physicians with expertise in methodological evaluation, as well as people repre senting the public and the pharmaceutical industry (those latter two groups did not take part in the voting for approval/rejection of recommendations).
Th e search of the literature for this specifi c paper was based on PubMed, and covered the period from the last CCCDTD publication (CCCDTD3, January 2006) until January 2012, so as not to duplicate the reviews already performed in the previous meetings. Th e current work is not a meta-analysis and therefore the search was not defi ned by statistical parameters, but rather by a series of decisions by the authors to include papers or not based on the pertinence of the information they contain. In fact, this resulted in some selected papers being included from the period preceding the interval mentioned above, if they were deemed necessary for a better understanding of the recommendations being made.
Keyword sequences started with ' Alzheimer 's disease' (AD) or 'mild cognitive impairment' (MCI), and then added the modalities being covered ('X-ray computerized tomography' (CT), 'CT' , 'head CT' , 'magnetic resonance imaging' (MRI), 'MRI' , 'PET-FDG' , 'SPECT-blood fl ow' , and so forth). ' ADNI' was also used as a search term because articles related to that project constitute a rather unique sample of data analysis papers on well-charac terized cases that have generally been submitted to multiple diagnostic modalities.
Th e offi cial recommendations from the CCCDTD4 meeting for clinical use of neuroimaging in patients with cognitive decline can be found in Table 1.

General considerations on clinical neuroimaging in dementia
Because the limited pharmacological therapeutic options available for patients with cognitive decline linked to neurodegenerative diseases are not, of course, universally eff ective across the diff erent entities found in that population, are costly, and can have signifi cant side eff ects [9,10], their rational use requires high accuracy of patient classifi cation. Th is classifi cation is what any type of neuroimaging technique is ultimately used for in such populations. Moreover, precise categorization has signifi cant implications in terms of prognosis and even counseling; for instance, frontotemporal dementias have a signifi cant probability (up to 40%) of occurring in a hereditary context [11], which is information many patients will value. Th ose reasons alone make it useful to discuss the role of neuroimaging in cases of cognitive decline.
Another important point to be considered is that neuro imaging of neurodegenerative diseases involves recognition on diff erent types of imaging techniques of patterns deemed to be associated with a specifi c condition. However, neurodegenerative diseases are associated with ageing, and the probability of fi nding any of them in isolation in a patient decreases with increasing age [12]. Finding a pure pattern therefore becomes less and less probable as patients advance in age. Moreover, the relationship between dementia and parameters evaluated by imaging probably vary with age: amyloid load may not be as specifi c for cognitive impairment in very old patients as compared with younger ones, whereas indices of neuronal loss (regional volumes, metabolic activity or absolute blood fl ow) might show a more stable relationship to dementia across ages [13]. Brain reserve also will infl uence the results from those studies. Recommendations on the use of imaging techniques have to be interpreted in light of such factors, whatever the technique used. We have tried throughout the present text to make sure to take those factors into account.

Structural neuroimaging -computed tomography and magnetic resonance imaging
Th e search approach described above identifi ed 416 papers on X-ray CT and MRI. A separate search using the same parameters with a fi lter for review articles yielded 63 papers. A search for review papers on head CT yielded only six papers.

General considerations
CCCDTD3 off ered two recommendations for structural imaging pertaining to the general clinical setting [1]. On the one hand, these partially exclusionary recommen dations supported the selective use of neuroimaging imaging procedures based on the First Canadian Con sensus Conference on the Assessment of Dementia published in 1991 On the other hand, for the fi rst time the recommendations supported the use of routine imaging to rulein asymptomatic cerebrovascular disease. No specifi c choice of imaging modality was recommended. Th e follow ing is a brief review of recent evidence pertaining to the clinical use of CT scans and MRI in the assessment of cognitive impairment in older people.

X-ray computerized tomography of the brain
For the initial assessment of patients presenting with cognitive diffi culties/symptoms of dementia, guidelines from several countries indicate that structural neuroimaging with CT or MRI is appropriate [4,14]. From a practical perspective, one should note that none of the guidelines diff erentiate between the types of technique (head CT or MRI) that should be used in the context of whether the request comes from a family physician practice or a specialist memory clinic.
One recent study found that no space-occupying lesion would have been missed in a memory clinic setting using the 1991 Canadian Consensus Criteria [15]. Th e study confi rmed that the systematic use of neuroimaging allows for modifi cations of the etiological diagnosis of dementia by revealing 26 clinically unsuspected strokes among the 210 patients evaluated. Th e value of fi nding unsuspected cerebrovascular disease lies in the added benefi t of managing vascular risk factors; the relevance of ordering neuroimaging is determined by the clinician's judgment as to whether aggressive vascular risk reduction is warranted in a given patient.
Space-occupying lesions, usually neoplasms or subdural hematomas, can be detected and may present with progressive cognitive impairment. Th ese lesions are uncommon, with estimations at approximately 3% [15]. Th e Mini Mental State Examination score and physical functioning may improve in one-half of people with existing dementia after surgical drainage of a subdural hematoma [16].
A recent review reaffi rmed several advantages of the enduring role of CT scan over MRI [17] in the clinical diagnosis of dementia, including lower cost, shorter duration of data acquisition, increased availability, ability to image patients with metallic devices, such as pacemakers, or who suff er from claustrophobia severe enough to prevent acquisition of MRI data.
One must emphasize that most available knowledge on CT scans has been obtained using older scanners that lacked the spatial resolution of newer, more powerful scanners. A study performed in a memory clinic setting compared the visual rating of hippocampal atrophy (HA) on coronal images, cortical atrophy on axial images and the assessment of white matter changes using a visual rating scale on a 64-detector-row CT scan, and on a 1.5 T MRI scan [18]. In that report, visual analysis of cortical and HA on CT scans was comparable with that obtained on matched brain MRI studies. In addition, the detection of white matter changes was also comparable for the two imaging modalities. Th ere was no systematic diff erence between CT and MRI ratings on any of the scales that were used, including the estimation of HA.

Measurement of atrophy
Th e characteristic neuropathological progression described by Braak and Braak [19] typically results in associated atrophy of the anterior parahippocampal gyrus, Table 1. Recommendations from CCCDTD4 for the clinical use of imaging procedures X-ray computerized tomography and magnetic resonance imaging • We recommend a head MRI when a radiologist/neuroradiologist and/or a cognitive specialist (neurologist, geriatrician, or geriatric psychiatrist) can interpret patterns of atrophy and other features that may provide added diagnostic and predictive value as deemed appropriate by the specialist (Grade 2B) • Standardization of clinical acquisition of core MRI dementia sequences is recommended in Canadian centers that have radiologists and cognitive specialists with expertise in assessing cognitive disorders, particularly when repeat MRI scans can provide additional diagnostic, prognostic and safety information (Grade 2B) • In addition to previously listed indications for structural imaging, a CT or MRI scan should be undertaken in the assessment of a person with cognitive impairment if the presence of unsuspected cerebrovascular disease would change the clinical management • The practical message is that structural imaging is not required in all (although will be indicated in most) persons with cognitive impairment. Although more costly and less available, MRI is preferable to CT • When available in the clinic, we recommend that cognition specialists use the computer images of the brain to educate persons with cognitive impairment about changes in the brain. This knowledge may reinforce adherence to vascular risk factors management and to life style modifi cations to improve brain health (Grade 3C) • We recommend against the use of functional MRI for the clinical investigation of patients presenting with cognitive complaint (Grade 1B) • Magnetic resonance spectroscopy is not currently recommended for clinical use to make or diff erentiate a diagnosis of dementia in people presenting with mild cognitive impairment (Grade 2C).
FDG-PET and SPECT regional cerebral blood fl ow imaging • For a patient with a diagnosis of dementia who has undergone the recommended baseline clinical and structural brain imaging evaluation and who has been evaluated by a dementia specialist but whose underlying pathological process is still unclear, preventing adequate clinical management, we recommend that the specialist obtains an 18 F-FDG PET scan for diff erential diagnosis purposes (Grade 1B) • If such a patient cannot be practically referred for a FDG-PET scan, we recommend that a SPECT rCBF study be performed for diff erential diagnosis purposes (Grade 2C) • There was only partial consensus for the proposition that for a patient with MCI evaluated by a dementia specialist and in whom clinical management would be infl uenced by evidence of an underlying neurodegenerative process, an 18  hippocampus and amygdala, followed by atrophy in multimodal association cortices. Th e structural imaging parameter with the strongest association in the literature to a diagnosis of AD is HA. Th e HA parameter can be estimated either with complex, more or less automated volumetric MRI techniques, or with simple visual rating scales, using equivalent likelihood ratios [20]. MRI-defi ned HA has been incorporated in both the Dubois Criteria [21] and the new National Institute of Aging-Alzheimer's Association Criteria [7] for AD. Medial temporal lobe atrophy predicts progression from MCI to AD [22][23][24][25].
However, this diagnostic test has not yet been adopted widely in clinical settings. One of the main limitations of studies on the usefulness of HA measurements for evaluating the presence of AD pathology is that almost all of them come from referral cognitive disorder clinics, and the usefulness of HA in a more general setting has not been established [20]. Progressive atrophy measured on MRI is accepted as a biomarker of the progression of MCI to AD [26]. Rates of atrophy have been measured in the hippocampus and whole brain in a number of studies and have been reviewed recently [27]. Longitudinal studies of structural MRI aim to characterize the earliest regions aff ected by AD and their subsequent neuropathological progression and rate of change, as well as changes in regions of the neocortex as they progressively become involved [28][29][30][31][32][33][34]. A longitudinal study using MRI automatic segmentation of hippocampal volume, at three time points over 10 years in over 500 normal controls, showed that declining hippocampal volume predicted onset of clinical dementia [31]. Still, although longitudinal data are interest ing, the predictive power of baseline MRI information is most useful to clinicians [25].
As a result of the marked contrast between cerebrospinal fl uid and the surrounding tissues [35,36], ventricu lar enlargement -a consequence of atrophy -can be easily measured on MRI T1-weighted images and is another potential marker of progressing atrophy.
Manual segmentation to measure the hippocampal volume is recognized as the gold standard [37,38]. However, an initial survey of the 12 most cited manual segmentation protocols revealed a 2.5-fold volume measurement diff erence [39]. Automated validated measures of hippocampal volume will help increase reproducibility of results. Th is increased reproducibility will not only increase the measures' diagnostic usefulness, but will also make it possible to use them as inclusion/exclusion criteria to enrich samples in trials testing therapeutic effi cacy of diff erent agents, allowing reduction of sample sizes [40,41]. Various automated methods to classify people with AD and MCI using structural MRI T1weighted images have been proposed and have been reviewed [32]. Th e authors concluded that most of the techniques accurately classi fi ed normal controls and subjects with AD. However, these methods had lower sensitivity in diagnosing pro dromal AD.
Several recent review papers on structural MRI in AD have been published [26,38,42].

Structural neuroimaging in memory clinics
Th e added value of neuroimaging in a memory disorders clinic has been evaluated in a retrospective study of 193 consecutively referred patients with cognitive impairment. Neuroimaging confi rmed, clarifi ed or contradicted the initial clinical diagnosis in more than 80% of patients. In 60% of the subjects, indicators of cerebrovascular disease were present [43].
When faced with the possibility of ordering brain imaging in the work-up of cognitive impairment, the clinician in a universal healthcare system such as that in Canada is faced with a dilemma: the values of social justice and allocation of resources must be balanced against the value of individual psychological benefi ts to patients and their families of investigating the symptoms in a thorough manner. Th is subjective benefi t has not, to our knowledge, been assessed formally, but clinical experience suggests that patients with cognitive disorders expect some kind of brain imaging when they seek medical attention in this day and age.
Th ere are no compelling reasons to withhold neuroimaging in the routine clinical assessment of cognitive impairment. Th e most obvious exception is situations where clinical management of vascular risk factors would not be modifi ed if silent cerebrovascular lesions were revealed [15]. Th ere is now documentation of the comparable value of the latest CT scanners compared with MRI for the routine work-up of dementia [18]. Head CT scanning is more widely available in Canada with typically shorter waiting lists than MRI and is suffi cient for routine investigation of cognitive impairment (see CCCDTD4 recommendations in Table 1). MRI off ers the advantage, how ever, of higher sensitivity to detect cerebrovascular lesions when compared with most currently used CT scanners, including in the detection of micro-bleeds on gradient-echo sequences [44], and this advantage will probably become stronger with the increasing availability of 3 T systems. Automated quantitative volumetrics and deformation-based protocols are also an advantage of MRI, but for the time being are mostly restricted to research settings [45,46]. Th e absence of ionizing radia tion is another advantage of MRI. Th e higher resolution of MRI is preferred to detect specifi c patterns of cortical and subcortical atrophy in non-Alzheimer's dementia, such as fronto-temporal dementias [47,48], multi-system atrophy or progressive supranuclear palsy [49]. For rarer disorders, diff erent MRI specialized sequences can assist the diagnosis in human prion-linked dementia [50].
Again, the diagnostic value of specifi c HA measurements has been well established in referral clinic populations, but its diagnostic value has not been demonstrated in unselected primary care patients, and this will remain a challenge for the foreseeable future.

Positron emission tomography fl uorine-18fl uorodeoxyglucose and single-photon emission computed tomography cerebral blood fl ow studies
Using the search strategy described above yielded 208 articles for positron emission tomography (PET) and 98 articles for single-photon emission computed tomography (SPECT).

Positron emission tomography fl uorine-18-fl uorodeoxyglucose versus single-photon emission tomography cerebral blood fl ow: a brief reminder
PET imaging of the fl uorine-18-fl uorodeoxyglucose ( 18 F-FDG) distribution is a reliable way of evaluating regional cerebral metabolic rates of glucose utilization, whereas SPECT imaging with either Technetium-99m-hexa methylene propyleneamine oxime or Technetium-99m-ethylene cisteinate dimer, the most frequently used agents, allows evaluation of regional cerebral blood fl ow (rCBF). Under most circumstances, rCBF and regional cerebral metabolic rate of glucose utilization are tightly coupled and represent the regional intensity of gluta matergic transmission [51]. Physiologically, therefore, these techniques can be considered generally equivalent, and the distribution patterns seen in neurodegenerative disorders as established over almost 30 years of their use are largely the same [52,53]. Th is equivalence also means that both approaches are subject to similar eff ects linked to age (see above) and to the infl uence of cerebral reserve [54], which must be factored in during clinical interpretation.
However, these two approaches are technically quite diff erent. PET imaging has a strong advantage over SPECT in terms of spatial resolution. Th is translates into the ability to pick up earlier disease-linked alterations and is probably the main reason why PET displays a 15 to 20% better accuracy [55][56][57]. Resolution performances are still evolving -in fact with fewer possibilities for improvement with PET than with SPECT, which is closing in on the performances typically associated with PET, and SPECT is considered a valid diagnostic option [58]. At the same time, the major advantages of SPECT over PET -that is, its greater availability and lower costare being eroded by the signifi cantly increased availability of PET scanning.
Finally, remember that nuclear medicine techniques are minimally invasive. Th eir track record is remarkable, with 0.9 recorded acute, limited side-eff ects per 100,000 administrations [59]. Th e long-term risk of radiation exposure is that of developing cancer, a limited concern given the age of populations evaluated for dementia -the mean latent period for most solid tumors secondary to exposure to acute radiation exposure is reported at between 20 and 30 years, with an average risk of dying from such a tumor of 0.028% [60].

Single-photon emission computed tomography regional cerebral blood fl ow studies
Review articles since the last CCCDTD continue to report on nonattenuation corrected studies and come up with numbers for diagnostic accuracy that are of the same order as those cited in 2007: the pooled sensitivity and specifi city are reported at around 75% and 90%, respectively [2,57].
Recent papers have confi rmed the ability of SPECT rCBF studies to effi ciently diff erentiate AD from other causes of dementia/cognitive impairment, such as vascular dementia and fronto-temporal dementia [61,62]. Th e diff erential diagnosis between AD and depression with cognitive impairment appears less optimal, in fact raising the possibility of a direct link between neurodegeneration and depression [63]. Diff erentiating dementia with Lewy bodies (DLB) from AD on the basis of hypoperfusion of the visual cortex also appears established [64], although the sensitivity and specifi city (65% and 80%, respectively) might be suboptimal [58].
Diff erentiation of DLB from AD with SPECT agents targeting the dopamine cell membrane transporter is available in Europe and the USA but not in Canada. A complete review on the subject has recently confi rmed the usefulness of this approach [65]; cases showing unclear results with that agent are likely to represent combined AD/DLB pathology.
Th e ability of SPECT rCBF to predict clinical evolution is discussed by diff erent authors, with variable conclusions. In AD, rapidity of progression can be inferred from the initial rCBF pattern [66]. However, prediction of conversion from MCI to dementia of the Alzheimer type with rCBF studies is not clearly established at this time, the results being discordant [67,68]. Still, diff erences of rCBF patterns are seen in MCI subtypes with diff erent risk of progressing to AD [69], suggesting that brain perfusion studies can probably predict progression of cognitive deterioration.
Computer-assisted image analysis is an important issue impacting on the clinical usefulness of rCBF SPECT imaging. Multiple techniques have been proposed to facilitate interpretation. It is virtually impossible to determine from currently available data whether one approach is better than another, as head-to-head comparisons are not generally available. When they are, diff erences are limited and of uncertain cause [70]. Ultimately, the most important question is whether such approaches are superior to visual evaluation. Invariably, the impact on accuracy is more signifi cant for less experienced interpre ters than for those regularly reporting such studies [71].

Positron emission tomography fl uorine-18-fl uorodeoxy glucose
Numerous reviews of the application of PET 18 F-FDG imaging to the evaluation of dementia have been published, reporting on large numbers of subjects, sometimes expressing wonderment at the ongoing questioning the technique still faces [72].
A recent review covers the period from 2000 to the latter half of 2011 [73]. Th e methodology employed is similar to that proposed here, and the conclusions are close to those we arrived at when covering, along with the previous CCCDTD report, the same material. Eleven studies meeting standard meta-analysis criteria were identifi ed. Th ose studies report on hundreds of subjects (the exact number is diffi cult to specify as some cohorts come from the same centers and probably contain many of the same patients) who have been classifi ed, based on diff erent diagnostic criteria (autopsy results, clinical follow-up, clinical assessment), as AD, healthy controls, or bearers of other dementia-associated conditions.
From the fi ve case-control studies reviewed, all at American Academy of Neurology diagnostic evidence level III, the pooled sensitivity, specifi city and accuracy for the diagnosis of AD were respectively 96%, 90% and 93%. From the two longitudinal studies, sensitivity results were 78% and 91.7% for the presence of dementia, with specifi cities of 81% and 88.9%. For the two pathologycontrolled studies looking at the presence or absence of AD, sensitivities were 94% and 84%, and specifi cities were 73% and 74%. 18 F-FDG imaging diff erentiated AD from other dementia-causing conditions with accuracies ranging from 79 to 93%.
Such percentages increase physicians' confi dence in the diagnosis given to a patient above the level associated with purely clinical approaches. Diagnosis is helped signifi cantly in over 70% of cases in a memory clinic setting, particularly in cases of atypical/unclear demen tias [74].
Predicting clinical evolution from 18 F-FDG studies in patients with cognitive disorders remains an important clinical objective. Clinical identifi cation of nondemented patients at risk of evolving to AD in general involves recognizing the presence of amnestic MCI, but not all amnestic MCI evolve to AD [75]. A normal PET study predicts with very high accuracy a low risk of cognitive deterioration for a number of years [35,76,77]. Subjects at high risk for AD, members of families with familial AD [78,79] and nondemented subjects with evidence of brain accumulation of β-amyloid [80] can show changes in 18 F-FDG studies comparable with those seen in AD subjects years before cognitive deterioration, as do many amnestic MCI sub jects (at high risk of evolving to AD) [81], suggesting that such a pattern might indeed carry prognostic information. Clinical follow-up of amnestic MCI/pre-amnestic MCI subjects shows that the initial 18 F-FDG study is a strong predictor of conversion to AD or cognitive decline [82][83][84][85]. 18 F-FDG might in fact be one of the better AD biomarkers in terms of its ability to follow cognitive decline throughout the course of the disease [86]. Improved specifi city in that domain might come from supplemental information provided by βamyloid PET imaging [87][88][89].
A practical issue arising from the previous discussion is the value of repeated imaging when an initial PET study is interpreted as normal or inconclusive for the presence of a dementing process. Very little information is available in that area but, should a second study be performed, results from one group suggest that absence progression over 1 year goes against the presence of an active neurodegenerative process [90].
Other diseases can present with cognitive impairment aside from AD, vascular dementia, fronto-temporal dementia and DLB -in particular, those that present with movement disorders often mimicking early Parkinson's disease (multisystem atrophy, cortical basal ganglia degeneration, progressive supranuclear palsy). Patients aff ected by those diseases are sometimes referred for evaluation of accompanying cognitive decline. Although the literature is much more limited on that topic than it is for other conditions aff ecting cognition, there is evidence that specifi c brain metabolism patterns can be seen with PET 18 F-FDG imaging [91,92].
Just as with SPECT, multiple computer-assisted interpre tation schemes have been proposed for 18 F-FDG studies, some having achieved signifi cant and well deserved acceptance [93]. Numerous technical issues remain to be settled in that fi eld, including the nature of the reference population being used for comparison purposes [94], pointing to a pressing need for standardization. In general, the same principles seem to apply here as were already discussed for SPECT; that is, such approaches seem to be particularly clinically helpful when used by observers with limited experience (clinical visual interpretation of those tests being subject to signifi cant variability [95]) or in research settings [96].

Introduction
PET imaging with amyloid ligands allows in vivo detection of amyloid plaques, a core pathologic feature of AD [97]. Although the 20-minute physical half-life of carbon-11-labelled Pittsburgh Compound B (PiB) has limited its use to research centers equipped with a cyclotron, a second generation of amyloid agents labeled with 18 F (110-minute half-life) has recently been developed, making it feasible to produce and distribute tracers for clinical use. In April 2012 18 F-fl orbetapir became the fi rst US Food and Drug Administration-approved 18 F amyloid tracer, and was approved by European authorities in early 2013. Higher non specifi c white matter binding has been reported with 18 F amyloid agents (fl utemetamol, fl orbetapir, and fl orbeta ben) but not with AZD4694 [98]. All have shown perfor mances comparable with those of PiB in clinical popula tions, and high correlation with postmortem measures of fi brillar amyloid-beta has been found [99].

Amyloid positron emission tomography in various clinical populations
Cognitively normal older individuals show elevated PiB binding in 10 to 34% of cases, a rate similar to that observed for amyloid pathology in autopsy studies [100]. Although the signifi cance of a positive amyloid scan in cognitively normal individuals remains uncertain, crosssectional studies have shown AD-like brain changes (hippocampal and temporal-parietal atrophy) while early longitudinal data have strengthened the notion that many such subjects are in a preclinical phase of AD. At present, there is no clinical indication for amyloid imaging in cognitively normal individuals or for the initial investigation of cognitive complaints.
Current data in MCI patients indicate that amyloid imaging provides prognostic information, presumably by identifying patients with underlying AD pathology [101]. As a group, MCI cases show elevated PiB binding in 52 to 87% of cases, with a regional distribution similar to that found in AD. In the largest longitudinal eff ort to date [102], the conversion rate was 82% in those with increased PiB uptake, but only 7% in PiB-negative subjects.
Most studies in AD have found very high (≥90%) PiB-PET sensitivity, with a pattern that closely mirrors the distribution of plaques found at autopsy [103]. Much like FDG-PET [75], amyloid imaging will probably not add value to the diagnostic work-up of patients with straightforward clinical AD, but is likely to be useful in patients with atypical complex presentations or early ageof-onset dementia. Considering that frontal temporal lobar degeneration and AD are the leading causes of dementia in young patients, that distinguishing the two during life can be clinically challenging, and that amyloidbeta plaques are not part of the frontal temporal lobar degeneration pathologic spectrum, authors have shown a valuable role for amyloid imaging in the diff erential diagnosis of these conditions, providing information beyond that obtained with 18 FDG-PET [104]. Other clinical conditions studied with amyloid PET include vascular dementia, cerebral amyloid angiopathy, Parkinson's disease dementia, and DLB, but a detailed discussion of results in those conditions is beyond the scope of this article (see [105] for a more detailed review).

Amyloid positron emission tomography in clinical practice: unresolved questions and recommendations
Amyloid imaging represents a promising technique in the evaluation of dementia but many ill-defi ned factors can probably impact its diagnostic validity and utility [106]. Currently, the only country where amyloid imaging is approved by government authorities is the USA. Clinicians from nearby countries such as Canada may therefore be called upon to interpret results from amyloid tests performed on their patients. Physicians should be very cautious in their interpretation because, used in isolation, this test cannot diagnose AD or MCI, or diff erentiate normal from abnormal aging. Recommendations are that the physicians consult with a dementia specialist familiar with this technique when confronted with amyloid imaging reports for studies performed in a region where they are available. Should amyloid imaging become more widely available, it is unlikely to become a routine test. Rather, it will probably be part of a comprehensive evaluation for complex and atypical cases referred to tertiary-care memory clinics when a more accurate clinical diagnosis is needed.
Although surveys suggest that rates of cholinesterase prescription in people with AD range from about 10 to 50% depending on the country, it is our clinical impression that inhibitors are prescribed to a large number of patients with non-AD dementia unlikely to benefi t from such therapy, while certain populations that could indeed benefi t remain untreated (for example, MCI due to AD). A decision to treat might be arrived at in a more rational fashion if amyloid PET was applied in the right circumstances, such as in atypical cognitive disorders and dementias [106]. Th is application could possibly result in signifi cant savings, but further cost-eff ectiveness studies are required. Th e more immediate impact of amyloid imag ing, however, will probably be improving clinical trial design by enrolling patients based on biological, rather than clinical, phenotype. A positive amyloid scan could well become the primary inclusion criterion for a study focused on prevention of AD progression.
Since the CCCDTD4 meeting, an important consensus paper has been published by the Amyloid Imaging Taskforce to provide guidance to dementia care practitioners, patients, and caregivers on appropriate use of amyloid PET [107]. A set of criteria were agreed upon that defi ne the types of patients and clinical circumstances in which amyloid PET could be used. Both appropriate and inappropriate uses are considered and discussed. Future research directions are also outlined, including diagnostic utility and patient-centered outcomes. Th is consensus paper is generally in line with the CCCDTD4 recommendations, but one should remember that no amyloidimaging tracer is approved in Canada at this time, or in fact being considered for approval.

Conclusion
Neuroimaging is now a major diagnostic player in the evaluation of subjects reporting cognitive decline, in large part because the technique no longer is called upon solely to rule out non-neurodegenerative causes, but rather to positively identify neurodegeneration by probing for a series of anatomical and functional modifi cations it infl icts on the brain. Although stronger confi rmation is always desirable, enough data are available on the value of HA assessment with MRI, or of metabolic/blood fl ow disturbances with PET and SPECT, to diff erentiate AD from other processes and to recommend using such techniques when diagnosis remains doubtful and clinical management is therefore undefi ned in selected cases. Amyloid imaging will also contribute to that process, although its optimal use remains to be clarifi ed. All of those techniques will also benefi t the clinical evaluation of potentially disease-modifying therapies, as well as help to develop a better clinical understanding of neurodegenerative processes as such. Th e recommendations from the CCCDTD4 meeting, which were based on the review presented above and on some additional texts found at the website [3], are, once again, summarized in Table 1.

Competing interests
The authors declare that they have no competing interests.