Neuroimaging Biomarkers in Alzheimer’s Disease

Abstract

Alzheimer’s disease accounts for approximately 50% to 80% of all causes of dementia. Co-existence of AD with other diseases causing dementia poses a diagnostic challenge, as we are still far from diagnosing AD accurately in order to manage it appropriately. Neuroimaging techniques, not only help diagnose AD but also consistently feature in diagnostic and research criteria for AD as biomarkers. Molecular biomarkers including positron emission tomography (PET) and single-photon emission computed tomography (SPECT), and structural biomarkers including magnetic resonance imaging (MRI), have been used in various therapeutic and prognostic studies in AD. This review highlights the recent advances in neuroimaging biomarkers, including molecular biomarkers (PET and SPECT tracers) and structural biomarkers (MRI), for AD. For the purpose of this review, molecular biomarkers have been further subcategorized into non-specific radiotracers (FDG-PET and blood flow SPECT) and specific amyloid- and tau-related radiotracers. The aim of this review is to discuss the recent advances and evidence of molecular and structural biomarkers of AD.

1. Introduction

Alzheimer’s dementia (AD) is the most common cause of dementia accounting for approximately 50–80% of all causes of dementia [1]. AD is characterised by cognitive decline especially memory impairment followed by impairment in other cognitive domains, with histopathological features of amyloid deposition and intra-neuronal tau deposition in senile plaques and neurofibrillary tangles, respectively [2]. The incidence of AD increases with age and doubles every five years after the age of 60 [3]. With increasing life expectancy worldwide, the prevalence of AD is expected to increase exponentially, increasing the urgency of accurate diagnosis in order to initiate management strategies early to help prevent progression of cognitive decline and reduce morbidity [4,5]. AD often co-exist with other diseases that cause dementia, and despite advances in diagnostic techniques, we are still far from accurately diagnosing and managing AD appropriately. Neuroimaging not only supports diagnosis of AD as the cause of dementia but also consistently features in the diagnostic criteria for probable AD [6].

The currently available therapeutic strategies can alleviate behavioural disturbances in AD, but drugs to halt disease progression comprehensively are still elusive [7]. Recently, drugs targeting Aβ protein have emerged with the promise of slowing cognitive impairment in AD, but the clinical relevance and long-term side effects require further longitudinal observation [8]. The complex aetiology and pathology of AD may be responsible for unsuccessful attempts at disease-modifying therapy [9]. Imaging biomarkers play an important role not only in visualising the typical abnormalities of AD for diagnosis and monitoring disease progression but also in therapeutics of AD by predicting response to novel AD drugs [10] (Figure 1). The National Institute of Ageing and Alzheimer’s Association (NIA-AA) in 2011 incorporated biomarkers in the diagnostic criteria of AD [11,12]. Further, the International Working Group (IWG) in 2014 categorised biomarkers for AD into diagnostic and prognostic biomarkers, focussing on MRI and PET biomarkers [13].

Recently, NIA-AA, has updated the recommendations for pre-clinical, mild cognitive impairment (MCI), and AD diagnosis, and categorised neuroimaging and cerebrospinal fluid (CSF) biomarkers into AT(N) groups. The A denotes amyloid deposition (A), T denotes pathological fibrillary tau, and N signifies neurodegeneration [6,14].

The recent shift towards different imaging biomarkers in AD research has made it necessary for us to review imaging biomarkers in AD, which not only include molecular biomarkers but also MRI biomarkers. This review highlights the recent advances and evidence in the imaging biomarkers, including molecular biomarkers (PET and SPECT tracers) and structural biomarkers (MRI), for AD.

2. Molecular Biomarkers

The molecular biomarkers which are commonly used for diagnostic, prognostic, or therapeutic purposes in AD can be further subcategorized into non-specific radiotracers, such as FDG-PET, blood flow SPECT, and dopamine transporter scan (DaT scan), and specific radiotracers including amyloid and tau radiotracers. Recent PET imaging approaches targeting neuroinflammation include mitochondrial translocator protein (TSPO), which provides evidence of microglial involvement in AD pathogenesis, and synaptic vesicle glycoprotein 2A (SV2A), a glycoprotein located on the membrane of synaptic vesicles, which may contribute to understanding AD pathophysiology and developing interventional strategies targeting inflammatory pathways [15]. However, PET imaging tracers targeting neuroinflammation are outwith the scope of this review.

2.1. Molecular Biomarkers with Non-Specific Radiotracers

Radioactive tracers such as 18F fluoro-deoxy glucose and 99m Tc-labelled tracer HMPAO (d, 1, hexamethylpropylene amine oxime) are used for brain FDG-PET and cerebral blood flow SPECT, respectively. These techniques are useful for differentiating dementia subtypes in clinical and research settings. In research settings, these techniques provide details of molecular and pathophysiology of the diseases causing dementia. Details of non-specific radiotracers in AD are provided below.

2.1.1. FDG-PET

FDG-PET is a well-established and extensively used neuroimaging tool to aid in clinical diagnosis of neurodegenerative conditions, including AD, as hypo-metabolism in certain brain regions corresponds to neuronal loss and functional impairment [16,17,18]. Reduced cerebral blood glucose uptake on FDG-PET in the areas of neurodegeneration, can help distinguish AD from other causes of dementia [19,20,21] and is considered a molecular imaging biomarker for AD [16,22]. Hypo-metabolism on FDG-PET precedes the detectible atrophy on structural MRI in AD and is a more sensitive technique for detecting early neurodegenerative disease [23,24,25]. Moreover, characteristic temporo-parietal hypo-metabolism can be seen on FDG-PET even before structural changes in people with AD [26,27] (Figure 2). With advancement in AD pathology, FDG-PET can detect extension of cortical hypo-metabolism to the frontal lobes, especially the prefrontal association cortex and anterior cingulate gyrus [28]. Moreover, it has been established that hippocampal, posterior cingulate cortex, and precuneus hypo-metabolism is seen in the early stage of AD [29,30]. Other neocortical areas are involved via brain networks, and the disease progresses in a stereotypical pattern [31].

Sensitivity and specificity of 99% and 98%, respectively, have been reported for FDG-PET to distinguish people with AD from normal subjects [32]. However, the specificity of FDG-PET to distinguish AD from other causes of dementia, including dementia with Lewy bodies (DLB) and frontotemporal dementia (FTD) were reported to be 71% and 65%, respectively, while sensitivity remains at 99% [32,33]. Recently, FDG-PET has been included as an independent biomarker in the ATN framework for AD diagnosis, as hypo-metabolism on FDG-PET predicts greater atrophy, poorer cognitive function, and a higher conversion to AD from mild cognitive impairment (MCI) [16]. Notably, conversion to AD from MCI can be predicted from a similar glucose uptake pattern as AD on FDG-PET [30]. Moreover, a higher rate of progression and associated cognitive decline can be predicted from the severity and progression of hypo-metabolism on FDG-PET in MCI patients. Thus, FDG-PET not only has diagnostic value but also prognostic value in people with AD [34,35,36,37].

Various clinical trials have used FDG-PET for treatment response evaluation in people with AD [38,39] and as an imaging biomarker for measurement of outcome and participant selection in AD and other dementia-related therapeutic trials [40,41].

2.1.2. SPECT

SPECT is a molecular imaging modality using gamma-emitting radiotracers; their distribution in the body is detected using a gamma camera [42]. SPECT highlights blood flow in different brain regions to evaluate brain and neurological conditions. The radiotracer ¹³³Xe was the first to be used for brain perfusion in 1990, followed by the use of various other radiotracers such as HMPAO (hexamethyl propylene amine oxime) and ECD (ethylcysteinate dimer), which are static radiotracers in contrast to ¹³³Xe, which is a diffusible radiotracer. However, SPECT should only be used as an aid to the diagnosis of a dementia subtype. The exceptions are Parkinson’s disease dementia and dementia with Lewy bodies, where dopamine transporter imaging (DaT) with DaTscan^TM (Ioflupane 123-I SPECT) has been found to be useful for confirming an underlying pre-synaptic dopaminergic deficit [43]. These radiotracers are manuctured by GEHealthcare and Siemens.

2.1.3. ECD and HMPAO-SPECT

SPECT is being widely used for cerebral blood flow (CBF) studies and helps identify functional abnormalities relevant to Alzheimer’s disease [44]. ^99mTc-ECD and ^99mTc-HMPAO are the two most commonly used static radiotracers [45], having several technical, economic, and logistical advantages over other radiotracers [46]. ^99mTc-ECD has in vitro chemical stability lasting several hours after reconstitution, faster blood clearance, and a high signal-to-noise ratio, making it an efficient technique to elucidate brain functions in various neurological disorders including cerebrovascular disease [47] and Alzheimer’s disease [48]. In addition, ^99mTc-HMPAO, in which ^99mTc is attached to the chelating agent HMPAO, is useful for the detection of regional blood flow in the brain and hence is used in diagnostic evaluation of dementias. It has greater specificity in comparison to the clinical criteria in differentiating AD from other dementias and can help in differential diagnosis of AD [49,50,51].

Notably, SPECT radiotracers are relatively less expensive and having a longer half-life than PET radiotracers, enabling more flexibility in tracer provision, administration, and imaging time. SPECT has been widely studied and may be used for the diagnostic workup of people with diseases causing dementia. Earlier, concerns were raised regarding SPECT use due to image quality, artefacts, and difficulty in quantifying blood flow, but emerging technology has addressed these shortcomings. The technological advancements include both the hardware and software techniques that help improve image quality and subsequent image analysis [52]. Data from large cross-sectional studies have been used to make inter-subject comparisons using t- or z-statistics, in order to eliminate individual variations in position, shape, and size [53]. The sensitivity and specificity of rCBF SPECT in differentiating AD from VaD are 74.5% and 72.4%, AD from FTD are 79.7% and 79.9%, AD from DLB are 70.2% and 76.2%, and AD from healthy controls are 76.1% and 85.4%, respectively [54].

Hypoperfusion on SPECT has been well correlated with the onset, the severity, clinical features, and prognosis of AD [55]. Parietal hypoperfusion is the most commonly seen in early-onset AD [56], while late-onset AD is characterised by medial temporal lobes hypoperfusion [57]. Some studies have found hypoperfusion in temporo-parietal regions to correlate with mini mental state examination (MMSE) scores in people with AD [58]. Notably, frontal lobe hypoperfusion has been negatively correlated with cognitive functions and positively correlated with rate of progression of the disease [59]. However, median survival time for people with AD has been negatively correlated with left temporal lobe hypoperfusion [60]. Thus, blood flow SPECT can be considered as a biomarker in AD [61,62], depicting the disease much earlier than the clinical manifestation [63].

2.1.4. Dopamine Transporter SPECT

SPECT for investigating the striatal dopamine transporter (DAT) status has been widely used in uncertain parkinsonian syndromes [64] (Figure 3). Despite no changes on ioflupane SPECT in AD in comparison to controls, it is helpful in differentiating AD from DLB. In DLB, there is 40–70% loss of striatal dopamine and dopaminergic cells, resulting in loss of dopamine transporter [65]. Dopamine transporter SPECT can thus be useful in differentiating neurodegenerative parkinsonian syndromes (PS) from non-dopamine-deficiency aetiologies, including drug-induced parkinsonism and essential tremor [66,67], with sensitivity and specificity of over 90% [68].

2.2. Molecular Biomarkers with Specific Radiotracers

A growing class of radiotracers targeting specific protein aggregates, amyloid-β (Aβ), and tau, are providing new avenues for research in AD diagnosis, as these radiotracers directly label the underlying disease pathologies. Amongst the two specific type of tracers, tau tracers are emerging as preferable to researchers for therapeutic and prognostic evaluation in AD clinical trials, due to the failure of agents targeting amyloid pathology. Thus, in this review, we provide a brief summary of amyloid PET and greater detail about tau PET.

2.2.1. Amyloid PET

Pittsburg Compound B (PIB), a 11-C labelled PET tracer was first to be used as an amyloid-specific radiotracer [69], and demonstrated that amyloid deposition precedes clinical manifestation of AD by 15 years [70,71]. However, this has been recently refuted, with proposals that cognitive decline precedes amyloid deposition [72], thus opening a new avenue in research to look beyond our understanding of amyloid in AD. The Amyloid Imaging Task Force [73] has clearly stated that amyloid PET should only be used if there is diagnostic ambiguity and that tracers can improve diagnostic accuracy. However, recently, it has been reported that initiation and discontinuation of medications in the management of AD can be influenced by the use of amyloid PET imaging in clinical practice [39,74].

PiB binds specifically with insoluble fibrillary Aβ, and cortical binding of PiB has been seen in more than 90% of people with AD [75]. Following PIB, several other 18F-labelled PET tracers such as [¹⁸F]Flutemetamol, [¹⁸F]Florbetaben, and [¹⁸F]Florbetapir have emerged as potential tools to aid in AD diagnosis and are widely used for research purposes [76,77]. The 18F-labelled PET tracers have a half-life of 110 min, overcoming the disadvantage of PIB, as PIB has a shorter half-life of 20 min, resulting in practical difficulties for clinical and research use. The U.S. FDA has provided its approval for use of [¹⁸F]Florbetapir in 2012 for exploring AD or other causes of cognitive impairment, followed by approval of [¹⁸F]Flutemetamol and [¹⁸F]Florbetaben in 2013 and 2014, respectively [77]. Amyloid PET has been found to be more sensitive but less specific than FDG-PET in distinguishing AD from other causes of dementia such as FTD [78]. Despite the association of amyloid deposition with AD, it has also been consistently found in cognitively healthy people [79], raising the question of whether amyloid deposition is a pathognomonic or necessary feature for the diagnosis of AD.

Although, it is difficult to ascertain the timing of amyloid deposition, it is believed to begin at the preclinical stage [71]. Amyloid PET has shown cortical uptake of amyloid tracers in the cingulum, precuneus, and frontal, parietal, and lateral temporal cortices in people with AD [80,81]. Moreover, amyloid PET has prognostic value, as people with MCI with positive findings on amyloid PET have a higher chance of conversion to AD, while those who are negative on amyloid PET have higher negative predictive value for conversion to AD from MCI [82,83,84]. It has also been noted that the rate of cognitive decline is directly proportional to amyloid deposition in people with MCI [85], and amyloid PET may help identify people at risk of cognitive decline and disease progression [86].

Amyloid tracers have also been used recently to monitor the therapeutic response of the amyloid-based drugs. Primary endpoints in these trials are usually measures of cognitive function such as the Alzheimer’s Disease Assessment Scale—Cognitive Subscale (ADAS-cog) [87] or the Clinical Dementia Rating Scale (CDR) [88]. However, due to cognitive reserve and ceiling and floor effects of psychometric tests, clinical presentations usually differ between people with AD [89,90]. Recently, 18-Florbetapir PET has been used in the PRIME study to evaluate treatment response of human monoclonal antibody on participants with prodromal or mild AD participants, which depicted a dose- and time-dependent reduction of Aβ plaques [91]. For the last few decades, most research has slavishly focused on Aβ as a target for the treatment of AD. However, most of the novel therapeutic agents targeting Aβ have failed to improve cognitive function or slow cognitive decline [92,93].

2.2.2. Tau PET

Tau protein helps stabilise the microtubules, which are essential for intracellular transport and cytoskeletal support. In its pathological state, tau becomes hyper-phosphorylated and accumulates intra-neuronally as neuro-fibrillary tangles (NFTs). Tau propagates intra-neuronally by releasing intracellular pathological tau into extracellular space, where it is taken up by recipient neurons and subsequently forms intracellular aggregates [94] (Figure 4).

Moreover, the clinical severity of dementia is closely associated with tau deposition in comparison to amyloid deposition [95,96]. Notably, Braak’s pathologic staging of AD corresponds with the distribution of NFTs, which correlates with cognitive impairment [97]. The distribution of NFTs and its correlation with brain atrophy further justify close association of tau protein with neurodegenerative conditions like AD [98]. Furthermore, clinico-pathological studies reveals a close association of tau pathology with loss of neurons and synaptic activity, the key correlates of cognition [99]. Recently, tau pathology was reported to be present much earlier than amyloid pathology in younger AD subjects, pointing to a role of tau in the initial stages of AD and in the initiation of the disease process [100,101]. After the onset of cognitive impairment in AD, tau pathology progresses at a faster rate [97]; thus, PET imaging could have an important role in evaluating prognosis of the disease.

Following the failure of anti-amyloid therapies, a paradigm shift has been observed in the last decade toward tau as a therapeutic target, with the development of several PET tracers to target tau pathology [100]. A variety of PET ligands have the ability to detect tau aggregates formed in AD with high affinity [102,103]. The retention pattern of tau PET overlaps with regions of brain atrophy [104] and predicts cognitive performance [105]. 2-(1-{6-[(2-[¹⁸F]fluoroethyl)(methyl)amino]-2-naphthyl}ethylidene)malononitrile ([¹⁸F]FDDNP) was the first PET tau tracer to accomplish imaging in humans, and a high density of NFTs was observed in the hippocampus [106,107]. However, FDDNP was reported to bind senile plaques and NFTs equally [107], leading to a search for a more selective tau tracer. A tracer which specifically binds tau should have high affinity towards paired helical filaments (PHF) to produce enough signal to be detected in vivo. Further, the tracer should be more selective towards tau rather than Aβ, as tau is found to be less frequently distributed in the neocortex in comparison to Aβ [108]. Furthermore, (18)F-labelled THK-523 or ((18)F-6-(2-fluoroethoxy)-2-(4-aminophenyl)quinoline) was the earliest tau-specific tracer to target tau in vivo, but its use in research or clinical trials was precluded due to high retention rate in the white matter [109,110]. Several other ¹⁸F-labelled arylquinoline derivative tau PET radiotracers, such as (18)F-THK-5105 and (18)F-THK-5117, were developed [111,112]. but the more recently developed ¹⁸F-THK-5351 has better imaging characteristics, with higher tau binding affinity and lower white matter retention [113].

Furthermore, radiotracers ¹⁸F-AV-1451 (also known as T807 and ¹⁸F-flortaucipir) and 18-F-AV-680 (also known as T808), both benzimidazole pyrimidine derivatives, were developed [37,114] with higher affinity for tau aggregates and lower white matter retention. Recently, it has been concluded that ¹⁸F-flortaucipir has better sensitivity than amyloid PET and can help detect preclinical cognitive changes in people with AD [105]. Further, region-based analysis of ¹⁸F-flortaucipir depicted that the spread of tau pathology is consistent with Braak staging [115,116] (Figure 5). Recently, FDA has approved the first tau PET tracer, ¹⁸F-AV-1451(¹⁸F-flortaucipir), for the estimation of NFTs in people who are being evaluated for AD [117].

In addition, tau PET tracers, such as the pyridinyl-butadienyl-benzothiazole derivative 11-C-PBB3, show higher sensitivity for tau [119] and MK-6240, a pyrrolo-pyridine-isoquinolone amine, has significant tau binding capacity without affinity for normal brain tissue [120]. Notably, 11-C-PBB3 can be quantified using cortical grey matter as a reference region for AD and other non-AD tauopathies to enhance its use in research [121]. Further, newer tau radiotracers such as [¹⁸F]RO-948 and [¹⁸F]JNJ64349311 [103,122,123,124] have been used to detect tau pathology, and it has been established that the tau pathology is deposited independently of Aβ deposition in the brain and is related to age [125,126,127]. Moreover, [¹⁸F]RO-948 is also claimed to be promising radiotracer for quantitative imaging in people with AD [128].

A recent study suggests that tau PET is better than amyloid PET and MRI in predicting cognitive change and has a role as a prognostic biomarker in preclinical and prodromal stages of AD [129]. However, tau PET tracer’s propensity for off-target binding to amyloid deposits and monoamine oxidase B may hamper their specificity in detecting tau pathology [130].

3. Structural Biomarkers

Computed tomography (CT) and MRI are two widely used structural imaging techniques, and CT is less expensive, faster, and more widely available than MRI, even in underprivileged nations [131]. However, CT is mostly used to rule out other reversible causes of dementia, such as tumour or subdural haematomas, and to help elucidate cortical atrophy and ventricular enlargement in AD [132]. Cortical atrophy and ventricular enlargement detected on CT scans are very late structural changes in AD and hence limit the use of CT as a biomarker for AD. MRI, another structural imaging technique, can readily detect characteristic structural changes in AD, such as medial temporal lobe in brain, quite early and hence qualifies as a neuroimaging biomarker for AD [133,134,135].

MRI Biomarkers in AD

It is imperative to discuss MRI biomarkers amongst the neuroimaging biomarkers, as structural MRI is an accessible neuroimaging technique for the evaluation of people with AD and is recommended in various diagnostic guidelines [136]. Hippocampus and medial temporal lobe volume measures on MRI are the most validated MRI biomarkers for AD, but structural and functional connectivity analyses using diffusion tensor imaging and resting state fMRI, respectively, can depict network disruption in AD.

For more than a decade, a characteristic atrophy pattern on 3DT1weighted MRI has been used to distinguish AD from other causes of dementia [137] (Figure 6 and Figure 7). MRI is not only used for visual- and voxel-wise analysis of atrophy pattern to predict different causes of dementia from AD [138], but also the location of WMHs in deep and periventricular brain regions is characteristic of cerebral small vessel disease (CSVD), secondary to hypertension. Microbleeds in cortical area are characteristic of Aβ pathology [139] and also of amyloid-related imaging abnormalities seen as a complication of anti-amyloid therapy.

With advancement in voxel-based morphometry to extract intracranial volumes, including brain volume, hippocampal volume, ventricular volume, cortical thickness, and cortical surface area, these have been recognised as biomarkers in AD. Shape and volume analysis of substructural changes in the hippocampus can distinguish AD from other causes of dementia [140]. Furthermore, for targeting pre-dementia stages in drug trials, hippocampal volume has been validated as a neuroimaging biomarker [141]. The pattern of cortical thinning can detect presymptomatic AD subjects, for example, in those at genetic risk of AD, and can predict severity of symptoms [142,143]. Furthermore, ventricular volume can also help differentiate AD from healthy controls [144].

Resting-state fMRI has been assessed as a biomarker to distinguish AD from other causes of dementia [145]. However, task-based fMRI did not show consistent findings [146,147]. Resting-state fMRI uses measures based on graph theory [148] or functional connectivity [149]. Resting-state fMRI in AD shows changes in the default mode network for interacting brain regions such as hippocampal co-activation [150] and increased activation in the medial temporal lobe [151].

AD has traditionally been considered as a cortical dementia, but some have demonstrated white matter (WM) abnormalities in AD, such as gliosis and demyelination, and considered WM damage to be independent of cortical damage [152,153]. WM damage and axonal deposition of tau may precede the cortical damage [154,155,156], suggesting not only a link between tau pathology and white matter damage but also the implication of tau in the initiation of AD pathogenesis. This WM damage can be quantified using diffusion tensor imaging (DTI), an MRI-based neuroimaging technique, which maps the WM microstructure changes in the brain [157]. Recently, diffuse microstructural changes have been observed in AD on DTI [152]. A summary of biomarkers used in AD has been provided in

Table 1. Biomarkers Used in Alzheimer’s Disease: Relevance, Strengths, and Limitations.

Biomarkers	Relevance to AD	Strengths	Limitations
Molecular biomarkers (Non-specific Tracers)
1. FDG-PET	Characteristic temporo-parietal hypo-metabolism can be seen on FDG-PET even before structural changes in people with AD.	Reduced cerebral blood glucose uptake on FDG-PET in the areas of neurodegeneration, can help distinguishing AD from other causes of dementia.	Despite higher sensitivity of 99%, the specificity of FDG-PET to distinguish AD from other causes of dementia, including DLB and FTD were reported to be 71% and 65%, respectively.
2. SPECT	SPECT is being widely used for cerebral blood flow (CBF) studies and help identify functional abnormalities relevant to AD.	It has greater specificity in comparison to the clinical criteria in differentiating AD from other dementias and can help in differential diagnosis of AD.	The sensitivity and specificity of SPECT in differentiating AD from other conditions is a bit low. Sensitivity and specificity of differentiating AD from VaD are 74.5% and 72.4%, AD from FTD are 79.7% and 79.9%, and AD from DLB are 70.2% and 76.2%, respectively.
Molecular biomarkers (Specific Tracers)
1. Amyloid PET	Amyloid PET has a prognostic value as people with MCI having positive findings on amyloid PET have higher chances of conversion to AD, while those who are negative on amyloid PET have higher negative predictive value for conversion to AD from MCI.	Amyloid PET has been found to be more sensitive than FDG-PET in distinguishing AD from other causes of dementia such as FTD.	It has also been consistently found in cognitively healthy people, raising the question of whether amyloid deposition is a pathognomonic or necessary feature for the diagnosis of AD and Amyloid PET can be used as the biomarker for AD.
2. Tau PET	A variety of PET ligands have ability to detect tau aggregates formed in AD with a high affinity.	Tau PET is considered to be better than amyloid PET and MRI in predicting cognitive change and has a role as a prognostic biomarker in preclinical and prodromal stages of AD.	Propensity of the off-target binding of tau PET tracers to amyloid deposits and monoamine oxidase B hampers the specificity of these tracers to detect tau pathology.
Structural Biomarkers
1. CT	It helps elucidate cortical atrophy and ventricular enlargement in AD.	CT is less expensive, faster, and more widely available, even in underprivileged nations than other imaging modalities.	Cortical atrophy and ventricular enlargement detected on CT scan are very late structural changes in AD and hence limit the use of CT as a biomarker for AD.
2. MRI	Advancement in voxel-based morphometry to extract brain volume, has led to recognition of various biomarkers based on MRI in AD.	Shape and volume analysis of sub-structural changes in the hippocampus can distinguish AD from other causes of dementia.	MRI may not be able to detect subtle functional changes or molecular abnormalities that occur in early AD.

.

4. Future Directions

Machine learning techniques offer opportunities to analyse the value of imaging and other biomarkers in AD [158].With regard to the algorithms used for identifying AD from healthy controls, research has been conducted using machine learning, deep learning, or a combination of both approaches, and some studies even combine other modalities such as PET and cerebrospinal fluid (CSF) with MRI to differentiate AD from healthy controls and other type of dementias [159]. Machine learning approaches can classify AD from healthy controls using hippocampal volume, with area under tAddedhe curve (AUC) of 0.912 [160]. Exploring areas beyond the hippocampus, researchers have achieved an AUC of 0.97 using temporal lobe structures, including hippocampus, para-hippocampal gyri, entorhinal cortex, and perirhinal cortex [161]; hippocampal features had the highest contribution in classification decisions (approx. 25–45%), temporal features stood second in classification decision (approx. 13%), followed by cingulate and frontal regions (approx. 8–13% each) [162]. In order to distinguish AD from other dementias, machine learning approach achieved an AUC of 0.948 and 0.731 with DLB for testing and training datasets, respectively [163]. Moreover, an AUC of 0.93 has been achieved in differentiating AD from FTD by developing frontotemporal dementia index, which is a ratio of the weighted sum of volumetric indexes of FTD-dominant and AD-dominant structures on MRI [164]. Moreover, machine learning algorithms can not only help identify the best treatment options but also predict the effectiveness of specific interventions in AD [165]. Hence, it can be concluded that machine learning seems to be widely used for classification of AD to differentiate it from either healthy controls or other causes of dementia.

5. Conclusions

Imaging biomarkers are of proven utility in supporting the clinical diagnosis of AD in a person with dementia, in selecting those to participate in clinical trials, in monitoring progression, and in measuring responses to treatment. The combination of biomarkers along with other personalised variables and genetic variables in convolutional neural network artificial intelligence algorithms, trained and validated using real-world data, is likely to be the most valuable in defining who, when, and with what to treat in the future.