Role of emerging neuroimaging modalities in patients with cognitive impairment: a review from the Canadian Consensus Conference on the Diagnosis and Treatment of Dementia 2012

The Fourth Canadian Consensus Conference on the Diagnosis and Treatment of Dementia (CCCDTD4) was held 3 to 4 May 2012 in Montreal, Quebec, Canada. A group of neuroimaging experts were assigned the task of reviewing and summarizing the literature on clinical and research applications of different neuroimaging modalities in cognitive disorders. This paper summarizes the literature and recommendations made to the conference regarding the role of several emerging neuroimaging modalities in cognitive disorders. Functional magnetic resonance imaging (MRI), magnetic resonance spectroscopy, and diffusion tensor imaging are discussed in detail within this paper. Other emergent neuroimaging modalities such as positron emission tomography with novel ligands, high-field MRI, arterial spin labeling MRI and noncerebral blood flow single-photon emission computerized tomography are only discussed briefly. Neuroimaging modalities that were recommended at the CCCDTD4 for both clinical and research applications such as amyloid and flurodeoxyglucose positron emission tomography, computerized tomography and structural MRI are discussed in a separate paper by the same authors. A literature search was conducted using the PubMed database including articles in English that involved human subjects and covered the period from the last CCCDTD publication (CCCDTD3; January 2006) until April 2012. Search terms included the name of the specific modality, dementia, Alzheimer's disease, and mild cognitive impairment. A separate search used the same parameters but was restricted to review articles to identify recent evidence-based reviews. Case studies and small case series were not included. Papers representing current evidence were selected, reviewed, and summarized, and the results were presented at the CCCDTD4 meeting with recommendations regarding the utility of various neuroimaging modalities in cognitive disorders. The evidence was graded according to the Oxford Centre for Evidence Based Medicine guidelines. Due to the limitations of current evidence, the neuroimaging modalities discussed in this paper were not recommended for clinical investigation of patients presenting with cognitive impairment. However, in the research setting, each modality provides a unique contribution to the understanding of basic mechanisms and neuropathological markers of cognitive disorders, to the identification of markers for early detection and for the risk of conversion to dementia in the at-risk populations, to the differentiation between different types of cognitive disorders, and to the identification of treatment targets and indicators of treatment response. In conclusion, for all of the neuroimaging modalities discussed in this paper, further studies are needed to establish diagnostic utility such as validity, reliability, and predictive and prognostic value. More multicenter studies are therefore needed with standardized image acquisition, experimental protocols, definition of the clinical population studied, larger numbers of participants, and longer duration of follow-up to allow generalizability of the results to the individual patient.


Introduction
Neuroimaging plays a central role in the clinical research of cognitive disorders. A group of Canadian neuroimaging experts were asked to review and summarize the literature, and to provide recommendations regarding the clinical and research applications of diff erent neuroimaging modalities in patients with cognitive impairment to the Fourth Canadian Consensus Conference on the Diagnosis and Treatment of Dementia (CCCDTD4) held 3 to 4 May 2012 in Montreal, Quebec. Other groups have published similar guidelines and recommendations, most recently the European Federation of Neurological Societies guidelines on the use of neuroimaging in the diagnosis of dementia [1].
Th e group built on other published guidelines, updated the evidence, and presented a rationale for the recommendations to CCCDTD4 in two papers. One paper by the same authors covered neuroimaging modalities currently indicated for clinical use in addition to research in cognitive disorders including [ 18 F]fl urodeoxyglucose and [ 11 C]-labeled Pittsburgh compound-B amyloid positron emission tomography (PET), computerized tomography and structural magnetic resonance imaging (MRI). Th e current paper covers promising neuroimaging modali ties that remain in the research realm but are not yet recommended for the clinical investigation of cognitive disorders.
For any medical test to have a diagnostic value in the individual patient, it must have established validity (having being tested across the spectrum of severity with appropriate randomization and compared with a golden diagnostic standard by an independent blinded rater), and to have established sensitivity, specifi city, predictive value, and test-retest and inter-rater reliability (Appraisal of Diagnostic Test 2010 [2]). Most of these elements are still lacking for the neuroimaging modalities discussed in this paper. On the other hand, these modalities have important roles in addressing several research questions, such as: identifi cation of the basic mechanisms of cognitive impairment and basic neuropathological markers of cognitive disorders; identifi cation of markers for early detection and diff erentiation between normal aging, risk state, and dementia; identifi cation of markers for diff erentiating between diff erent types of cognitive disorders; under standing the neurobiological course of cognitive dis orders (for example, changes in neuronal networks as a result of neurodegeneration and brain injury); identifi cation of therapeutic targets and evaluation of the mechanism of action of diff erent therapies in cognitive disorders; and enriching patient selection for therapeutic trials by identifying reliable neuropathological markers for diff erent types of cognitive disorders.
Th ere is signifi cant variability in individuals presenting with cognitive disorders. Th e variability could be specifi c to the neuroimaging modality used but it could also be a general variability aff ecting most test results whether it is clinical, biological or from neuroimaging. In addition to the expected general variability (for example, demographic characteristics, stage of illness, and co-morbidity), there is emerging evidence regarding the impact of cognitive reserve on various test results. Th is eff ect was demonstrated in recent neuroimaging studies where regional cerebral blood fl ow measured by PET scanning was inversely related to the level of education and occupational attainment, indicating higher resiliency in those with higher cognitive reserve in the face of Alzheimer's pathology [3]. Issues related to sources of variability such as cognitive reserve have not been considered adequately in neuroimaging studies at this stage.
Th e modalities discussed in greatest detail in this expert consensus review are: functional magnetic resonance imaging (fMRI), magnetic resonance spectroscopy (MRS), and diff usion tensor imaging (DTI). Other modalities such as PET imaging with novel ligands, arterial spinlabeling perfusion MRI, high-fi eld MRI, and noncerebral blood fl ow single-photon emission com puterized tomography (SPECT) are only discussed briefl y due to the limited literature.

Methodology
A literature search of the PubMed database was performed and covered the period from the last CCCDTD publication (CCCDTD3; January 2006) until April 2012, so as not to duplicate the reviews already performed in the previous meetings [4,5]. Search terms included the name of the specifi c neuroimaging modality (fMRI, MRS, or DTI) combined with the terms 'dementia' , ' Alzheimer's disease' (AD), or 'mild cognitive impairment' (MCI). Th e search was limited to 'English' and 'Humans' . A separate search used the same parameters but restricted it to review articles to identify recent evidence-based reviews. Th e purpose of this process was to summarize the evidence for the consensus meeting and make recommendations regard ing the role of emerging neuroimaging modalities in cognitive disorders. Th is review was therefore focused and selective rather than a systematic review, and the search was not defi ned by statistical parameters but rather by a series of decisions by the authors to include or not include papers based on the pertinence of the information they contained. Th is approach 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.
Th e evidence was summarized, and recommendations were submitted to the CCCDTD4 meeting for deliberation towards the goal of reaching a fi nal expert consensus recommendation regarding the role of each selected emerging neuroimaging modality in cognitive disorders based on a majority vote on each of the recommendations by the attending members.
We graded the evidence based on the Oxford CEBM level of evidence grading guidelines (Levels of Evidence, March 2009 [2]). Th e evidence for the three main neuroimaging modalities discussed in this paper (fMRI, MRS and DTI) is mainly driven from well-designed casecontrolled studies with signifi cant heterogeneity of participants and methodo logy. Th erefore, the evidence for these modalities could be graded as 3b.
Functional magnetic resonance imaging fMRI techniques identify brain activity level based on endogenous blood oxygen level-dependent (BOLD) image contrast. Th e BOLD signal is the product of the integrated synaptic activity of neurons and is mainly detected by MRI based on hemodynamic change in the oxyhemoglobin/deoxyhemoglobin ratio. An increased BOLD signal is usually interpreted as activation and a negative BOLD response as deactivation of the underlying brain areas [6].
Th ere are mainly two types of fMRI studies used in cognitive disorders: those evaluating brain activation during episodic memory encoding and retrieval, and those evaluating the default mode network.

Studies evaluating brain activation during episodic memory encoding and retrieval in Alzheimer's disease and mild cognitive impairment patients
Typically fMRI experiments compare the MRI signal during one cognitive condition with a control condition such as visual fi xation [7].
Episodic memory encoding function is most commonly investigated because of its early and consistent involvement in AD. Successful memory formation and retrieval requires coordinated and reciprocal activation of medial temporal lobe (MTL) structures and deactivation in areas connected to MTL areas such as the posterior cingulate cortex and parietal cortex [8].
During episodic memory encoding, AD patients consis tently show lower activation in MTL structures, particu larly the hippocampus [9][10][11], failure of the normal deactivation in posteromedial cortical areas such as the posterior cingulate and medial parietal cortex [12,13], and increased activation in the prefrontal cortex, probably as a compensation mechanism [14,15]. Similar pattern emerges during episodic memory retrieval (especially the lower activation of MTL).
In patients with MCI, which is a risk state for AD, the fi ndings are mixed. During episodic memory encoding, some studies reported decreased activity [9,16,17] while others reported increased activity in the MTL area. Th is discrepancy is probably due to several factors such as diff erences in clinical defi nition of MCI, severity of cognitive impairment, and level of task performance [18].
Few fMRI studies followed MCI patients longitudinally. In one such study, baseline hyperactivation in MTL structures during episodic encoding correlated with future cognitive decline [19]. Another study showed that this activation declines over time, creating a U-curve pattern from hyperactivation to pseudo-normalization and fi nally to hypoactivation [20]. Table 1 summarizes the key fi nd ings of fMRI studies in AD and MCI patients during episodic memory encoding and retrieval based on published meta-analysis and evidence-based reviews [21][22][23].

Studies evaluating the default mode network in Alzheimer's disease and mild cognitive impairment patients
Th e brain network referred to as the default mode network includes several cortical areas that are particularly active at rest and deactivate during cognitive tasks.
Th is network includes the medial prefrontal cortex, posterior cinculate cortex, precuneus, anterior cingulate cortex and parietal cortex. Th e hippocampus is functionally connected to this network. Recently, the correlation between the intrinsic oscillations or time courses of activity in the hippocampus and diff erent brain regions of the default mode network in cognitive disorders has received signifi cant attention [24]. Th is technique is feasible in patients with cognitive impairment because of the ease of acquisition and the lack of task performance confounding [25].
A signifi cant alteration in the intrinsic functional connect ivity between the hippocampus and areas in the default mode network at rest and during cognitive tasks in patients with MCI and AD has been reported [26][27][28]. Table 2 summarizes the fi ndings of default mode network imaging studies in the resting state and deactivation during cognitive tasks based on a recent published review [29].

Evaluation of the eff ect of pharmacological treatment using functional MRI
Cholinergic therapy has been the mainstay for the treatment of cognitive disorders for over two decades. fMRI was used to evaluate the underlying mechanism of cholinergic therapy in healthy subjects and in patients with cognitive disorders. Th ese studies evaluated the changes in the fMRI signal after a single dose and after short-term or long-term exposure to cholinergic agents. Th ere is evidence for changes in the brain activation pattern in key memory, and especially attentional networks in response to cholinergic agents in healthy controls and in patients with MCI and AD during cognitive tasks. On the other hand, a signifi cant variability in the results is observed -most probably due to heterogeneity of treatment response, variability in the pharmaco kinetics and pharmacodynamics of these agents, and the inherent test-retest variability in fMRI studies [30]. More recently, a study by Li and colleagues demonstrated enhanced resting state functional connectivity and cerebral blood fl ow in patients with AD in response to long-term cholinergic treatment [31].

Functional MRI in asymptomatic apolipoprotein E ε4 carriers
fMRI studies in asymptomatic genetically at-risk subjects carrying apolipoprotein E ε4 documented diff erences in BOLD signal during diff erent cognitive tasks compared with noncarriers, although the direction of this diff erence and the factors aff ecting it are quite variable [32]. Th ere are diff erences in resting state functional connectivity between apolipoprotein E ε2 and ε4 carriers on one end and apolipoprotein E ε3 carriers on the other -the implications of this fi nding for cognitive disorders remains unclear [33] Functional MRI in diff erentiating Alzheimer's disease from non-Alzheimer's disease dementia Few fMRI studies compared AD with other forms of dementia. A diff erential pattern of resting state functional connectivity and task-induced BOLD signal has been identifi ed between patients with AD patients compared with those with Lewy body dementia [34,35]. Other studies identifi ed diff erential BOLD signal during cognitive tasks between AD patients and those with subcortical vascular cognitive impairment [36] and those with fronto-temporal dementia [37].

Advantages and limitations of functional MRI in cognitive disorders
Th ere are several advantages for fMRI in cognitive disorders as a noninvasive and radioactivity-free modality allowing safe repeated scanning, with relatively high spatial and temporal resolution, and measurement of brain activity during specifi c behavior (like memory encoding).
On the other hand, fMRI is limited by sensitivity to head motion, eff ect of task performance, brain atrophy, and transient brain and body states at the time of scanning (for example, arousal, attention, and eff ort). fMRI is highly dependent on neurovascular coupling, resulting from neuronal and glia activity. Th is dependency makes fMRI especially vulnerable to the eff ect of conditions that aff ect vascular coupling, such as vascular insuffi ciency, change in blood gas levels, and exposure to substances. Furthermore, a change in activation in one area during a task might be related to a pathological change in other functionally connected areas [38]. Finally, test-retest reliability data for fMRI are just beginning to emerge in cognitive disorders [39].

Conclusion/recommendations
Owing to several limitations, such as variability in imageacquisition protocols and experimental paradigms used, heterogeneity of the clinical population studied, limited data in other forms of dementia, small sample size, limited longitudinal data, limited test-retest reliability data, and inverted U-curve activation patter in the MTL in the people with MCI, fMRI is not currently recommended for the clinical investigation of patients presenting with cognitive impairment (level of evidence 3b). On the other hand, fMRI holds promise in several areas such as early detection of dementia and predicting conversion of MCI to AD, distinguishing between AD and non-AD dementia, identifying treatment target and changes in brain activation in response to intervention, and evaluating neuropsychiatric and behavioral symptoms in the context of preclinical and clinical dementia. Standardization of image-acquisition protocols and experi mental paradigms and establishing validity and reliability data with large number of participants and longer follow-up periods are needed. Resting state methodology is a feasible and promising methodology and can be easily integrated into current large multicenter neuroimaging studies.

[ 1 H]Magnetic resonance spectroscopy [ 1 H]Magnetic resonance spectroscopy in Alzheimer disease
[ 1 H]MRS is a noninvasive technique used to measure the concentration of low molecular weight metabolites in vivo, with a detection threshold of approximately 1 mmol/l (1 mM). Although MRS data are acquired using a MRI scanner, the method is limited by a low signal to noise ratio resulting in volumes of interest typically ranging between 1 and 8 cm 3 . Data analysis also involves removal of line-shape distortions, removal of macromolecule/lipid contributions if needed, and fi tting of the data to mathematical model functions to increase the accuracy of metabolite-level estimates [40]. Th e most commonly reported in vivo brain metabolites include Nacetylaspartate (NAA), glutamate, glutamine, cholinecontaining compounds (Cho), creatine (Cr) compounds and myo-inositol. Altered levels of NAA or NAA/Cr are the most common fi nding reported in subjects with AD [41][42][43][44][45][46][47][48][49] and MCI, although alterations in other metabolites including myo-inositol [50], scyllo-inositol [51], and glutamate [52] have been reported. Decreased NAA has been documented in subjects with AD in the occipital lobe [53], temporal lobe [54], parietal lobe [55], and frontal lobe [56].
NAA is an amino acid located primarily within neurons in the central nervous system. Th e concentration of NAA is among the highest of the free amino acids in the brain (8 to 10 mM in brain tissue) and it normally produces the largest peak in the magnetic resonance spectrum [57]. NAA is considered a marker of neuronal density and/or viability. Reduced NAA may therefore imply neuronal death, or it may indicate neurometabolic impairment based on the known correlation between the rate of mitochondrial activity and NAA synthesis [58]. NAA and myo-inositol change has been noted in the hippocampus in AD, and this change has been associated with cognitive measures [59]. In fact, the NAA/myo-inositol ratio has been shown in several studies to be the most robust marker for discriminating AD patients from age-matched normal older controls [60,61]. NAA/myo-inositol ratios have also been previously shown to correlate with Mini-Mental State Examination scores [60]. Interested readers are directed to the review by Griffi th and colleagues for a detailed description of MRS fi ndings in the dementias [48].

Longitudinal studies in subjects with mild cognitive impairment
Th e value of MRS as a prognostic indicator of impending dementia in subjects with MCI can only be answered by longitudinal studies. Since 2005 there have been nine longitudinal MRS studies performed in subjects with MCI. Subjects were typically followed for 1 to 3 years to identify a cohort that converted to dementia. Th e results from these studies are summarized in Table 3.
Th e most consistent fi nding reported in MCI subjects that convert to dementia compared with MCI subjects that remain stable is lower NAA or NAA/Cr. Lower levels of NAA/Cr have been noted in several brain regions, including the occipital cortex [62], paratrigonal white matter [63], temporoparietal lobe [64], posterior cingulate [65,66], and posteriomedial cortex [67]. In the study by Rami and colleagues, lower NAA/Cr was observed in subjects that were classifi ed as prodromal AD and later converted to AD [64]. Kantarci and colleagues demonstrated that NAA/Cr measured from the posterior cingulate added predictive value for conver sion to dementia when combined with hippocampal volume [66]. Furthermore, of these studies listed, three have included receiver operator curve analysis. Modrego and colleagues demonstrated that receiver operator curve analysis for NAA/Cr <1.61 predicted conversion with 100% sensitivity and 75% specifi city [62]. Th e area under the curve was 0.91 with a positive predictive value of 83% and a negative predictive value of 100%. Similarly, Fayed and colleagues showed that NAA/Cr <1.40 in the posterior cingulate predicted conversion of MCI to probable AD with sensitivity of 82% and specifi city of 72% and an area under the curve of 0.82 [65]. Finally, Modrego and colleagues showed that NAA/Cr <1.43 in the posteromedial parietal cortex predicted conversion to probable AD with 74% sensitivity and 84% specifi city and an area under the curve of 0.84 [67].
Despite the consistency of the studies listed above, two studies reported no baseline diff erences in [ 1 H]MRS between MCI stable patients and MCI converters in the bilateral posterior cingulate and inferior precuneus [68,69] or in the parietal white matter [69]. Another study showed no diff er ences in the medial temporal lobe in cognitively impaired not demented (CIND) stable patients and CIND converters in the medial temporal lobe [70].
Although current studies show an emerging trend of lower NAA/Cr in MCI subjects that convert to dementia compared with MCI subjects that remain stable, further study is needed. Variability in participant selection, the criteria for conversion to dementia, and methodological inconsistencies in the brain region studied, the spectroscopy acquisition protocol, and the spectroscopy analysis procedures limits the generalizability of the current studies.

Conclusion/recommendations
MRS is not currently recommended for clinical investigation of patients presenting with cognitive impairment due to several limitations discussed above (level of evidence 3b). On

Diff usion tensor imaging
Volumetric MRI measures of the MTL or hippocampus have been validated as markers of neuronal damage and can support the diagnosis of early AD [71]. As gray matter atrophy is presumably a relatively downstream event, eff orts to detect earlier changes at the microstructural level are worth pursuing as part of the general eff ort to defi ne the earliest measurable changes that lead to AD.
Diff usion-weighted imaging has been used clinically over the past 10 to 20 years, mainly to detect acute stroke. Th is imaging modality can provide a window into the structural integrity of cerebral tissue beyond what is visible on standard MRI. Molecular diff usion refers to the random movements of molecules, which can be described statistically given the size of the molecule and the temperature and nature of the medium through which it travels [72]. In pure water, molecules move in all directions with equal probability (isotropy). Diff usionweighted imaging uses information derived from the random movements of water molecules through biological tissues; white matter tracts with their neuronal membranes and myelin sheaths interfere with the random movements of water molecules, thus rendering the movements an-isotropic. Th e main measures obtained are mean diff usivity (MD) and fractional anisotropy (FA). Th ese measures provide indirect information on the microscopic structural properties of white matter fi bers [73]. With recent advances in diff usion data modeling, it is now possible to determine in vivo the location of white matter nerve bundles using algorithms that estimate the likelihood that two adjacent voxels are connected (DTI tractography). Th is type of analysis can provide exquisite detail of the white matter tracts that connect diff erent areas of the brain, including the large-scale neural networks that underlie all cognitive functions [74]. Analysis of DTI data becomes more complex at fi ber intersections, and estimates of white matter voxels contain ing crossing fi bers range from 63 to 90% depending on the estimation method used [75]. Th is implies that most fi rst-generation DTI studies are limited in their power in at least two-thirds of the voxels, enabling clear visualization of only the most prominent white matter tracts such as the cortico-spinal tract. With the emergence of new high angular resolution diff usion imaging methods, it has been possible to manage the problem of crossing fi bers and the complex confi gurations of bundles of white matter [76,77], enabling much more detailed visualization of smaller white fi ber bundles. Th ese tools are well suited to study large-scale brain anatomical connectivity and are particularly well suited for the study of complex brain disorders such as AD.
Several studies in the past decade have focused on the value of DTI measures in the diff erentiation of MCI and AD from controls; 55 studies comprising close to 2,800 subjects were included in a recent meta-analysis. Th is analysis compared the contribution of FA and MD to the better established imaging measures of MTL atrophy in the diagnosis of MCI and AD using the magnitude of eff ect sizes as the method of comparison. Studies using FA measures at the level of regions of interests showed signifi cant diff erences between AD and controls, with the largest eff ect sizes in the cingulum, splenium of the corpus callosum, uncinate fasciculus, superior longitudinal fasciculus and frontal lobes. For studies using MD, the regions of interest most useful to distinguish AD from controls were the hippocampus, splenium of the corpus callosum, parietal lobes and temporal lobes. In general, eff ect sizes were superior for the volumetric MTL measures compared with the DTI measures for the discrimination between controls and AD subjects. Predictably, the diff erences between MCI and controls were of less magnitude, with only regions of the cingulum showing moderate eff ect sizes using FA. However, in the comparison between MCI and controls, studies using hippocampal MD showed superior eff ect sizes compared with volumetric MTL measures [78].
Another recent review, using pooled analysis of regional mean FA and MD values, showed that MD values are diff erent in all white matter regions of the brain between controls and AD subjects, and that FA showed similar results except for the parietal lobe and internal capsule [79]. Furthermore, a few studies in healthy older subjects at risk for AD showed abnormalities in MD values in regions known to be aff ected in AD [80,81]. In addition to showing early alterations in MCI patients, DTI appears to correlate with cognitive performance independent of cortical atrophy, which suggests access to an upstream process in the neurodegenerative cascade [82].
Th e search for appropriate DTI and high angular resolution diff usion imaging parameters for the diagnosis of cognitive impairment is still a work in progress [83]. Various pa rameters behave in diff erent ways according to localization [84]. In addition to the choice of diff usion parameters, recent tractography studies illustrate the superiority of analysis methods that can manage crossing fi bers [85,86]. Th e analysis of large-scale networks involved in cognitive functions might enable more specifi c correlations between brain structure and function [87]. Table 4 summarizes DTI fi ndings in AD and MCI based on the two recently published meta-analyses [78,79].

Conclusion/recommendations
Th e DTI technique shows promise as a potential imaging biomarker in the early diagnosis of high-risk states for AD, and has similar accuracy to volumetric hippocampal measures for the diagnosis of established AD. However, due to signifi cant heterogeneity in studies in terms of patient characteristics and image-acquisition parameters, DTI is not currently recommended for the clinical investi gation of patients presenting with cognitive impairment (level of evidence 3b).
Potential clinical applications of tractography include support for the diagnosis of neurodegenerative diseases [88], but also identifi cation of amnestic MCI [82,89] and prediction of progress to AD in risk groups [90,91]. Further studies with standardized image-acquisition and analysis protocols, a standardized defi nition of participants, a larger number of participants across age and gender distributions, and a longer follow-up period will increase validity and reliability of the data and allow better generalization of results to the individual patient.

Other emerging neuroimaging modalities used in the evaluation of cognitive disorders
In this section we briefl y discuss neuroimaging modalities that hold promise as research tools in cognitive disorders but have less published literature, which limits conclusions and recommendations at this time.

Scanning with emerging PET ligands
PET imaging off ers a variety of techniques that have a signifi cant role in investigating patients with cognitive impairment. Amyloid imaging with [ 11 C]-labeled Pittsburgh compound-B amyloid and [ 18 F]fl urodeoxy glucose PET are covered elsewhere.
Several novel amyloid ligands are being developed to enhance sensitivity and specifi city of PET amyloid imaging and to increase feasibility of scanning by having a longer half-life (see Figure 1 for an example of amyloid binding with the novel ligand [ 18 F]fl orbetapir). Amyloidbinding SPECT agents are being considered, which will off er another feasible method of imaging amyloid [92].
A molecular probe with high affi nity to tubulin associated unit (TAU) fi brils and a low affi nity for synthetic amyloid-β 1-42 fi brils is in the early phase of development [93].
Th ere is evidence for regional neurotransmitter alteration in AD. Imaging the dopamine system with PET tracers can help to diff erentiate AD from Lewy Body Dementia (LBD) and LBD from Parkinson's disease. Several PET ligands are available to image the serotonin system; for example, 5-HT 1A imaging can reveal hippocampal neurodegeneration in AD, and reduction in 5-HT 2A binding has been demonstrated in AD and in MCI patients. Imaging the cholinergic system off ers better understanding of the basic mechanism of cognitive symptoms and response to cholinergic therapies in AD, Parkinson's disease dementia, LBD and other dementia. Cholinergic nicotinic declines have been measured with PET nicotine receptor ligands. Reduced acetyl cholinesterase activity has been shown in MCI converters, AD, LBD and Parkinson's disease dementia compared with healthy controls, and basal acetyl cholinesterase activity predicts therapeutic eff ects of cholinesterase inhibitors.
Neuroinfl ammation is a neuropathological feature of numerous neurodegenerative conditions. PET allows for in vivo quantifi cation of various aspects of neuroinfl ammatory responses, which may serve to monitor the eff ects of pharmacological interventions. Using PET ligands, high phospholipase enzymatic and high astrocytic activity has been described in the AD brain [94,95]. Finally, microglial activation detected via specifi c PET ligand binding has been described in several neurodegenerative conditions. Biomarkers of tissue pathology have potential applications for monitoring disease progression and evaluating the eff ects of new pharmacological interventions, accumulation abnormal proteins in the brain, neurodegeneration, neuroinfl ammation and neurotransmission. Table 5 presents common PET ligands used to delineate diff erent neuro biological processes in cognitive disorders. Interested readers are referred to recent published reviews for more details [96,97].

Arterial spin labeling perfusion magnetic resonance imaging
In this technique, the water content in arterial blood is labeled to detect change in perfusion. Th is method is Recent studies using this method have shown hypoperfusion in some brain areas in MCI and AD patients compared with controls, including the right inferior parietal, bilateral posterior cingulate gyri, and bilateral middle frontal gyri, a pattern of hypoperfusion that is similar the one seen with PET and SPECT scan studies in this population [98,99].

Noncerebral blood fl ow-based single-photon emission computerized tomography
Noncerebral blood fl ow studies can be performed to diff er entiate LBD from AD with SPECT agents targeting the dopamine cell membrane transporter. [ 123 I]-labeled iofl upane is useful in diff erentiating Lewy body dementia from AD with high specifi city over 90% [100].
Another noncerebral blood fl ow SPECT approach is based on imaging of the norepinephrine transporter on the plasma membrane of sympathic terminals in the heart, which are lost early in LBD and Parkinson's disease, but not in AD using the ligand iodine-131 metaiodobenzylguanidine. Th is tracer is approved for human use mostly for the diagnosis and treatment of paragangliomas. Results in small studies have demonstrated above 90% accuracy for the diagnosis of LBD [101,102].

Ultra-high-fi eld magnetic resonance imaging
Th e use of ultra-high-fi eld (>3 T) human MRI has increased over the past two decades. Th e advantages of ultra-high-fi eld MRI include greater signal to noise ratio, reduced scan times, increased image resolution, increased spectral dispersion, and novel contrast. However, there are technical and logistic challenges including increased magnetic fi eld distortions, decreased radiofrequency fi eld homogeneity, increased power deposition, and the unknown safety profi le of most surgical implants at fi eld strengths above 3 T. Th ese issues need to be addressed but initial studies have demonstrated the advantages of this technology for the study of the human brain, and AD in particular [103][104][105].

Discussion
In this expert consensus review, we summarize the literature regarding emerging neuroimaging modalities used to study patients with cognitive impairment that are not currently recommended for the clinical investigation orbetapir scan is characterized by a low cortical uptake and high uptake in the adjacent WM. An abnormal scan is characterized by a focal increase of cortical [ 18 F]fl orbetapir uptake. Increased cortical uptake is evidenced by the reduction between cortical to WM contrast in at least two brain areas. A negative scan indicates sparse to no neuritic plaques, and is inconsistent with a neuropathological diagnosis of AD at the time of image acquisition. A negative scan thus implies reduced likelihood of clinical AD. A positive [ 18 F]fl orbetapir scan indicates a neuropathology consistent with the presence of moderate to frequent amyloid neuritic plaques. However, the interpretation of a positive scan should take into consideration that amyloid neuritic plaques may also be found in patients with other types of neurologic conditions as well as older people with normal cognition. Images obtained from Alzheimer's Disease Neuroimaging Initiative (ADNI) and processed at PR-N's laboratory.
of this population. Th ese modalities have a role in identifying indicators of neuropathology, which has the potential to inform research by providing better markers for inclusion/exclusion in clinical and research protocols in the future. For these neuroimaging modalities to have diagnostic value in the individual patient, they must have established validity and to have established sensitivity, specifi city, predictive value, and test-retest and interrater reliability. Most of these elements are still lacking for the modalities discussed in this paper.
International multicenter neuroimaging initiatives are well underway and provide the opportunity for multimodal standardized protocols allowing combining neuroimaging modalities together with clinical and other biological markers (for example, genetics and cerebro spinal fl uid) [106]. Some of the neuroimaging modalities discussed in this paper -such as DTI, resting-state fMRI, and MRS -are quite feasible to acquire during MRI studies. Furthermore, post-acquisition processing and analy sis is getting more accessible with the advances in image processing and analysis software. Larger longitudinal studies with improved homogeneity of participants and methods, combining neuroimaging and other diagnostic data, will probably give modalities discussed in this paper clinical utility in the near future.

Conclusion
For neuroimaging modalities discussed in this paper, further studies are needed to establish diagnostic utility. Larger, multicenter, multimodal studies with better homo geneity of the clinical population and methodology are therefore needed to establish diagnostic value for the individual patient.

Competing interests
RB is co-founder and Chief Scientifi c Offi cer of Bioscape Imaging Solutions Inc. AMB received consultation fees from industry, including Pfi zer and Lundbeck Canada. PR-N has ongoing scientifi c collaborations with NAVIDEA biopharmaceuticals (nonfi nancial competing interests). The remaining authors declare that they have no competing interests.

Declarations
This article has been published as part of Alzheimer' s Research & Therapy Volume 5 Supplement 1, 2013: Background documents to the 4th Canadian Consensus Conference on the Diagnosis and Treatment of Dementia (CCCDTD4). The full contents of the supplement are available online at http://alzres.com/supplements/5/S1. Publication charges for the supplement were funded by the Canadian Consensus Conference on the Diagnosis and Treatment of Dementia (CCCDTD). Although residual conference funds used include contributions from pharmaceutical companies, no commercial organization has been involved in the selection of participants, choice of topics, preparation of background papers or recommendations. In kind support was also provided by the Canadian Dementia Knowledge Translation Network, and the offi ces of Drs Serge Gauthier (McGill University), Christopher Patterson (McMaster University) and Howard Chertkow (McGill University), whose role as Guest Editors involved the coordination of the project without involvement in the journal's standard peer review process which applied for all articles.