Keywords

1 Introduction

The 2018 Research Framework defined the beta-amyloid, tau, and neurodegeneration or AT[N] stages, determined by imaging biomarkers, as critical for Alzheimer’s disease (AD) research [1]. Imaging biomarkers for AD were used mostly in research until June 2021, when the US Food and Drug Administration (FDA) approved aducanumab for clinical use; since then, imaging biomarkers have been used in the clinic as well as in research because amyloid deposition in the brain needs to be documented before aducanumab may be used. Under the commercial name Aduhelm™, aducanumab is a human monoclonal antibody that selectively reacts with beta-amyloid brain aggregates, including soluble oligomers and insoluble fibrils [2]. Beta-amyloid (amyloid for brevity) is a protein that begins to accumulate in the brain of people who eventually will develop AD 5–15 years before the onset of clinical symptoms [3]. While amyloid imaging showed that aducanumab reduced brain amyloid in clinical trials including patients with mild cognitive impairment (MCI) and mild AD [4], the reduction in clinical worsening associated with AD was minimally or not affected by aducanumab at the clinical stages included in these trials [4, 5]. Nonetheless, physicians began prescribing this medication, which logically required demonstrating amyloid brain accumulation in a potential candidate for aducanumab therapy. While abnormal brain amyloid can be predicted by measuring amyloid and tau in cerebrospinal fluid [6], people prefer amyloid imaging when availability or cost does not preclude its use. Although the use of plasma biomarkers of amyloid deposition in the brain is promising and gaining in accuracy [7], at the time of this writing either CSF or imaging is still needed to document brain amyloid deposition in the clinic [8]. Brain amyloid removal by monoclonal antibodies occurs largely through the walls of small vessels, which become more permeable, giving rise to brain edema or microhemorrhages in about 25% of the treated patients [9]. The occurrence of both events can be monitored with MRI using the FLAIR sequence for edema and gradient echo or susceptibility-weighted sequences to monitor blood deposition in the brain [9].

The clinical use of imaging biomarkers was further encouraged by the January 2023 FDA approval of lecanemab (Leqembi™), another humanized monoclonal antibody, this one targeting soluble amyloid protofibrils and causing not only a reduction of brain amyloid but also a slowing of the clinical worsening as well [10]. That the clinical effect was modest could be explained by the stage of AD at which this medication was used. In mice, two broad stages can be observed, an amyloid-dependent stage and an amyloid-independent stage [11]. When excess amyloid is removed from the brain at the amyloid-dependent stage, the animals do not go on to develop Alzheimer’s disease. However, if excess amyloid is removed at the amyloid-independent stage, the animals continue to worsen relentlessly until death. At the amyloid-dependent stage, there is no abnormal tau in the brain of the animals, but at the amyloid-dependent stage, abnormal, phosphorylated tau has begun to be detectable in the brain [11]. Similar changes can be observed in humans using imaging biomarkers [12]. The brain deposition of amyloid alone is not associated with cognitive impairment [3]. However, when tau is detected by imaging outside the entorhinal cortex, people are already symptomatic, with the degree and type of clinical symptoms correlating closely with the degree of tau deposition and its location in the brain [13]. Areas with high tau are typically hypometabolic on FDG PET, such that there is a ying-yang relationship between these two imaging biomarkers: where tau is high, metabolism is low (Fig. 1). As both aducanumab and lecanemab were in clinical trials of symptomatic subjects, who were likely at the amyloid-independent stage, even the modest clinical effect is encouraging. From the foregoing, determining tau build up in the brain of a potential candidate for one of these therapies could be very helpful to predict benefit: people with more tau are less likely to benefit from anti-amyloid antibodies [14].

Fig. 1
12 MRI images of the brain with lesions diagnosed with Alzheimer's disease. Normal metabolism is observed in the primary sensor motor areas. The contrast MRI represents the deposition of tau that results in low metabolism. Medial occipital region and normal metabolism with high amyloid load is observed.

Imaging findings in patient with AD (logopenic aphasia). Metabolism, amyloid, and tau imaging from a 57-year-old woman with the logopenic aphasia variety of Alzheimer’s disease. The primary sensory-motor areas (asterisks), as well as the primary visual (striatal cortex) and auditory (Heschl’s gyrus) regions (arrowheads), have normal metabolism and no tau deposition. By contrast, areas with a high tau deposition (e.g., inferior parietal lobule, arrows) tend to have decreased metabolism. In some areas, a high amyloid deposition corresponds to a low metabolism and an increased tau (e.g., the precuneus). However, there are areas with high amyloid load and normal metabolism, such as the medial occipital region. Uptake in the region of the substantia nigra does not correspond to tau deposition. (From [13] with permission)

Full size image

At the time of this writing, several monoclonal antibodies targeting brain amyloid are being studied at the pre-symptomatic, amyloid-dependent, stage. These studies are made possible by the availability of amyloid imaging to detect excess brain amyloid in people who are cognitively unimpaired. Furthermore, although neuropsychological scores are used as outcome measures, brain tau provides a measure with less day-to-day variability than neuropsychological testing, and it is beginning to be used as an outcome measure [15]. This chapter will review the imaging biomarkers mentioned in this introduction and others most extensively used in dementia, leaving for future reviews potentially useful biomarkers, for instance, cortical mean diffusivity [16].

2 Neuroimaging Biomarkers

The imaging modalities used to study AD include MRI and PET. Single photon emission computed tomography (SPECT) is also being used to study brain perfusion, but its use has been largely replaced by the use of an MRI sequence, arterial spin labelling, that allows for the study of brain perfusion.

2.1 MRI Regional Brain Volume

Neurodegeneration causes progressive loss of brain volume, which cannot be appreciated on MRI images nearly as well as other brain lesions, such as tumors or infarction. Although brain volume loss, widely known as atrophy, can be rated visually [17], automated methods are less time-consuming and more precise and facilitate longitudinal follow-up. Classical automated methods follow one of the two approaches: voxel-based morphometry (VBM) [18] or surface-based morphometry (SBM) [19,20,21]. These methods are based on the automatic segmentation of the brain cortex, deep nuclei, white matter, and ventricles, based on the different intensities of these structures, mostly on T1-weighted images. In VBM, through sophisticated deformation techniques, beyond the scope of this chapter, the brain of a given individual is placed in a standard brain template, thus facilitating the statistical comparison of the various brain regions of this individual with similar regions of a control sample, typically healthy people of similar age and sex as the individual of interest [22, 23]. In SBM, similar procedures are used for segmentation of the various components of the brain, but the boundary between the cortical gray and underlying white matter is obtained. This boundary, together with surface coordinates of the brain of the individual of interest, as well as its deep structures, is ingeniously compared to standard brains and atlases that contain the typical anatomical regions [24, 25]. The first SBM software, FreeSurfer [26, 27], which is available through an open access license, is perhaps the volumetric software most commonly used in research. For clinical use, several VBM or SBM commercial packages are available for seamless integration with clinical PACS systems. In dementia MRI, the accuracy of software that classifies clinically appropriate cases has been compared favorably with the accuracy of trained readers [28]. Interestingly, even among image specialists, those with more experience in reading brain images obtain the best clinical results in dementia patients from automated MRI volume methods [29]. More recently, machine learning and neural network computing are revolutionizing the use of MRI and other imaging datasets for the longitudinal assessment of brain changes in AD/ADRD [30,31,32,33]. Since the steps of data processing are not as clear as with VBM or SBM, the reliability of these techniques can be best evaluated by researchers with an extensive knowledge of brain anatomy and function and by comparison with other quantitative techniques [34].

Volume loss in the medial temporal regions was the first reliable neuroimaging finding detected in AD [35] and still thought to be the most robust on MRI. The name of neurodegenerative pattern has been assigned to the pattern of atrophy most often observed in AD [36] (Fig. 2). Indeed, regional atrophy in a set of mostly post-Rolandic structures is a strong predictor of AD on MRI [19]. In cognitively unimpaired people, the presence of a neurodegenerative MRI pattern is predictive of the development of mild cognitive impairment later in life, particularly when associated with amyloid deposition [37].

Fig. 2
5 MRI templates of the brain's frontal, sagittal, and horizontal plane represent Alzheimer's disease. The regions highlighted in yellow represent the thicker cortex in the higher region and red represents the lower cortex.

Cortical thickness in Alzheimer’s disease. On MRI templates of the brain, in color are areas of the brain of a patient with Alzheimer’s disease where the cortex is thinner at a higher (yellow) or lower (red) statistical level as compared to a group of controls of the same age and sex

Full size image

Most cortical thickness studies have assumed a linear volume loss in the AD process, starting at the pre-symptomatic stages. However, a biphasic pattern is more likely, with increased thickness at some point of the very early process, followed by subsequent progressive thinning [38, 39]. This pattern would agree with early inflammatory changes resulting in cortical swelling that would be compensated for and surpassed by the volume loss caused by later progressive neuronal loss. This pattern would explain why atrophy has not been found by every study to predate the onset of cognitive impairment in familial autosomal dominant AD [40,41,42,43]. It would also explain the paradoxical increased “atrophy” in patients treated with monoclonal antibodies targeting amyloid [44]. A reduction in cortical amyloid has been documented neuropathologically to reduce inflammation in the cortex [45]. These data suggest that MRI volumetry is not a reliable marker of neurodegeneration in therapeutic trials.

By correlating postmortem findings with the pattern of atrophy on MRI, three distinct atrophy patterns have been found in patients with typical AD neuropathology, including amyloid deposition: typical AD (about 70% of cases), limbic-predominant AD (20%) and hippocampal-sparing AD (10%) [46]. Most patients with typical and limbic-predominant AD initially present with an amnestic syndrome, but only about 40% of those with hippocampal-sparing AD do. Medial temporal atrophy is most severe in patients with limbic-predominant AD, followed closely by typical AD, and milder in those with hippocampal-sparing AD. Conversely, the most severe cortical atrophy was noted in patients with hippocampal-sparing AD, followed by those with typical disease, and then limbic-predominant AD. The ratio of hippocampal to cortical volumes allowed the best discrimination between subtypes [46]. In addition, some AD patients, particularly younger ones, present with a disorder of visual perception, including one or several components of Balint’s syndrome, alexia, and even field defects on confrontation testing, caused by posterior cortical atrophy [47,48,49] (Fig. 3).

Fig. 3
4 illustrations of cortical thickness in MRI scan for a patient diagnosed with Alzheimer’s disease. A posterior cortical atrophy variant with decreased cortical thickness is exhibited. A radiograph with multiple views of the brain represents decreased cortical thickness in the posterior region.

Cortical thickness, amyloid, and tau in a patient with the posterior cortical atrophy variant of Alzheimer’s disease. Areas of decreased cortical thickness are indicated as in Fig 2. Areas with increased amyloid or tau are in red. Note the similar topography of these changes, most pronounced in the posterior portion of the brain

Full size image

The pattern of atrophy in AD resembles that of dementia with Lewy bodies (DLB) [50], but in DLB there is more atrophy in the fusiform gyrus and paracentral cortex [51]. The imaging similarity between the two diseases can be explained at least in part by the frequent coexistence of AD and alpha-synuclein neuropathologies [50, 52, 53]. Patients with pure alpha-synuclein pathology have little atrophy, such that the lack of hippocampal atrophy associated with memory loss in MCI is indicative of DLB [54]. Atrophy in AD, which tends to affect the posterior brain regions, differs from atrophy in FTD, which tends to affect the anterior portion of the brain [55]. Hippocampal volume alone poorly differentiates AD from FTD; hippocampal sclerosis associated with FTD could explain the overlap [56].

2.2 Metabolism

Regional brain metabolism is currently used as a biomarker of neurodegeneration, for instance, to document the “N,” neurodegeneration, in the AT[N] system [1]. Metabolism is measured with 18F-FDG PET [57,58,59]. Metabolism may be closely linked to the pathophysiology of AD; as in older people, the regional brain expression of AD-risk genes correlates with regional metabolism [60]. The most typical metabolic pattern found in early AD is decreased metabolism bilaterally in the parietotemporal association cortex and posterior cingulate gyrus [61] (Fig. 4). Metabolism reflects synaptic activity and therefore is most affected early in the regions to which medial temporal neurons project [62, 63] and may reflect impaired connectivity even in pre-symptomatic subjects [40, 64]. As atrophy corresponds to neuronal loss, it is no surprise that the regions most affected on volumetric MRI and metabolic PET do not coincide early in AD [65], but they partially overlap as the disease progresses [66]. As AD progresses, some areas of the frontal association cortex become hypometabolic, while the paracentral cortex (primary motor-sensory areas) remains preserved (Fig. 1). The specificity and sensitivity of these findings continue to be debated. In studies of AD with neuropathological confirmation, the sensitivity (84–95%) has been higher than the specificity (71–74%), that is, a normal study is seldom associated with AD [59, 67]. Using consensus diagnosis, in an area under the receiver operating characteristic (ROC) analysis for three automated approaches to mild AD diagnosis, the specificity approximates 85% when the sensitivity is pegged at 80% [68]. Depending on the approach and the sample studied, the accuracy for predicting the evolution of MCI to AD varies from 0.774 to 0.983 [68]. Among persons with MCI, those most likely to progress to AD have metabolic findings similar to AD [69]. 18F-FDG PET may predict better than structural MRI or SPECT the worsening from MCI to AD [70].

Fig. 4
6 FDG PET scans of the brain in coronal, horizontal, and sagittal views for the diagnosis of Alzheimer’s disease. The highlighted region represents low metabolism in the early stages of Alzheimer’s disease.

FDG PET group findings in Alzheimer’s disease. Projected on a rendered MRI and shown in red are areas with a low metabolism in a group of 28 patients with early Alzheimer’s disease, compared with 28 healthy controls. Note sparing of the paracentral (primary motor-sensory) cortex

Full size image

The AD metabolic pattern can also be found with DLB, in part because the two brain pathologies often coexist [52, 53]. However, while AD tends to render hypometabolic the posterior cingulate gyrus, this structure is often spared in DLB, giving rise to the “posterior-cingulate island sign” on FDG PET [71] (Fig. 5). Unlike AD, which tends to affect posterior brain regions, the frontal and anterior portions of the temporal lobes are usually hypometabolic in FTD [58]. Patients with progranulin mutations, however, often have parietal involvement [72].

Fig. 5
2 sagittal MRIs with positron emission tomography for the diagnosis of Lewy body dementia. The highlighted region represents decreased perfusion. The metabolism is coupled with Alzheimer’s disease.

“Island sign” in Lewy body dementia (LBD). On MRI templates of the medial aspect of the brain, areas of decreased metabolism (18F-FDG PET) in AD (A) and decreased perfusion (H215O-PET) in LBD (B). Metabolism and perfusion are coupled in AD and LBD. Note involvement of the posterior cingulate gyrus in AD but sparing of this region (arrow) in LBD. (Modified from [153])

Full size image

2.3 Perfusion Imaging

In the absence of associated vascular disease [73], perfusion is typically coupled to metabolism in neurodegenerative disorders. In current clinical practice, brain perfusion is most often studied with MRI arterial spin labelling (ASL), a sequence that can be obtained together with more conventional MRI sequences. As expected, cerebral blood flow (CBF) obtained with ASL tends to correlate topographically with metabolism, particularly in the more advanced AD stages [74,75,76,77,78]. However, FDG PET slightly outperforms ASL in separating AD and, particularly, MCI patients from controls, both in visual readings and using automated procedures [74, 77, 79, 80].

Brain perfusion can also be assessed with SPECT, using Tc-99m HMPAO (hexamethyl propylamine oxime, Ceretec™), a lipid soluble macrocyclic amine, or Tc-99m ECD (ethyl cysteinate dimer, Neurolite™). A head-to-head comparison of perfusion SPECT with metabolism PET has shown better sensitivity and specificity of PET over SPECT in AD and diffuse Lewy body disease [57].

2.4 Amyloid Imaging

Brain amyloid was initially imaged with “Pittsburgh compound B” (11C-PIB) [81]. PIB is available bound to 11C, a positron-emitting isotope with a half-life of 20.4 min, requiring an on-site cyclotron. However, since 2012 there are amyloid-imaging compounds bound to 18F, with a half-life of 109.8 min. The longer half-life allows for the radiotracer to be synthesized at a facility with a cyclotron and then shipped to institutions with PET cameras, more widely available. Good concordance with histologically measured amyloid load has been shown not only for PIB [82, 83] but also for three 18F amyloid PET tracers, 18F-florbetapir [84], 18F-flutemetamol [85], and 18F-florbetaben [86], which are approved by the FDA for use in the clinical setting. At the time of this writing, 2023, a fourth amyloid PET tracer, 18F-flutafuranol, also known as 18F-NAV4694, is only used in research, but it has much less white-matter binding than other 18F tracers, thus providing cleaner images, similar to those obtained with PIB [87].

As another biomarker of AD, decreased CSF amyloid 42 [88], amyloid brain deposition begins in the preclinical stages of AD, increases during the MCI stage, and, by the time of the AD diagnosis, remains relatively stable as the disease progresses [3, 89]. Thus, amyloid deposition is a marker of the pre-symptomatic stages of the disease and correlates with the degree of cognitive impairment only in the preclinical stages and MCI, not during AD [3, 90], while atrophy and synaptic dysfunction continue to increase and spread as clinical AD worsens and cognition deteriorates [89].

In asymptomatic individuals of similar age, amyloid deposition has been found more often among APOE4 carriers [91], but this genotype may not have an effect on the risk of cognitive worsening once its effect on amyloid deposition is accounted for [92, 93]. Lifetime cognitive engagement has been found to protect from preclinical amyloid deposition [94], but this effect, like the protective effect of physical exercise, may be restricted to APOE4 carriers [95]. Impaired sleep has been associated with an increased amyloid burden [96].

Amyloid deposition is the strongest and earliest neuroimaging predictor of future cognitive impairment in healthy elderly and of worsening from MCI to AD, increasing the risk between three- and sevenfold [92, 97, 98]. The effect of amyloid deposition on cognitive impairment in the early stages of the AD continuum may be modulated by some common genetic variants. For instance, healthy APOE4 carriers have not only a greater amyloid deposition but also worse memory and visuospatial skills for the same amount of 11C-PIB binding [99]. This finding may reflect a longer period of time with amyloid deposition in the APOE4 carriers. Healthy, amyloid-positive carriers of the Met genotype of the brain-derived neurotrophic factor (BDNF) Val66Met allele have a greater worsening on follow-up in episodic memory, language, and executive function than the Val homozygotes despite similar amyloid PET binding in both groups [100].

Amyloid imaging is also a powerful tool to separate the dementias characterized by amyloid deposition, such as AD and diffuse DLB, often associated with AD [53], from the FTD, which course without amyloid deposition. Separating patient samples of AD and FTD validated clinically, areas under the ROC curve for 11C-PIB (0.888) and 18F-FDG (0.910) were similar [101]. 11C-PIB slightly outperformed 18F-FDG in patients with known histopathology [101]. A confounder is the presence of amyloid deposition in some older people with FTD because the prevalence of amyloid positivity increases with age [37, 102]. Although the diagnosis of AD is predicated on the presence of amyloid plaques in the brain [103], a few cases with AD have a tau PET typical for AD, with intense uptake in the cortex of AD regions, but a negative amyloid PET [104, 105]. Neuropathology is still lacking, but it is possible that these patients have diffuse or cotton-wool plaques or some other type of amyloid burden not well imaged with the current amyloid PET tracers [106, 107]. These patients should not be confused with patients who have a negative amyloid PET, but a positive signal, although typically weaker than in AD, in FTD-typical areas with one of the tau tracers. In these cases, the signal is often greatest in white matter, which on neuropathology contains a lower density of known abnormal protein aggregates, such as tau or TDP-43, than the cortex [108].

Patients with an AD clinical phenotype may have a negative amyloid PET scan. In a clinical trial of early AD, 14% had negative amyloid scans among 214 with AD symptomatology [109]. This proportion parallels the 14% amyloid-negative in a population sample of 154 amnesic MCI patients and 16% of 58 MCI patients from ADNI [110] and may rise to 30% when the patients studied are older than 82 years [111]. It may reflect the smaller subset of patients with dementia who do not have elevated amyloid or tau at autopsy [112]. These imaging findings could reflect the rather mixed pathology found in the oldest-old [113]. However, even with a careful neuropathological exclusion of other etiologies, clinical and neuropathological findings are occasionally dissociated: individuals with marked amyloid and neurofibrillary pathology may be cognitively intact [112]. In these individuals there is less amyloid deposition in the form of fibrillar plaques and intimately related oligomeric amyloid assemblies, less hyperphosphorylated soluble tau species localized in synapses, and less glial activation [114].

In early AD, amyloid deposition is highest in the default network and, thus, in fronto-parieto-temporal association cortex, including the precuneus, but sparing the paracentral regions and primary visual and auditory sensory cortex (Fig. 1). The caudate nucleus is often affected as well.

Longitudinal amyloid imaging allows for the evaluation of the natural history of amyloid deposition among at-risk genotypes [91], and it is being used as a marker of effectiveness in clinical trials carried out during the preclinical stage of AD, because it has helped elucidate brain changes during AD therapy [10, 115, 116].

2.5 Tau Imaging

In the healthy brain, the protein tau stabilizes neurotubules and is therefore essential for normal neural function [11]. However, in AD and other neurodegenerative disorders, tau becomes abnormally hyperphosphorylated, dysfunctional, and misfolded, constituting the tangles observed neuropathologically in AD and other tauopathies. PET tracers are available that bind strongly to the abnormally folded tau, using the folding properties of this protein for binding. These tracers do not bind to the healthy, native form of tau, but here we refer to hyperphosphorylated tau simply as “tau,” as has become common usage. PET tracers currently used to image tau include 18F-T807, most recently known as 18F-AV-1451 or 18F-flortaucipir [117,118,119], which was approved for clinical use by the FDA after a postmortem study proved that 18F-flortaucipir binds to tau tangles in AD [120]. 18F-Flortaucipir shows highly specific uptake in areas known neuropathologically to contain a large amount of tau in AD [13, 118, 121] (Fig. 1). It has little white matter binding, but there is uptake in the substantia nigra, explained by binding of 18F-flortaucipir to melanin [122, 123], and in the choroid plexus, possibly from binding to calcifications or even tau in this structure [124, 125]. In older individuals, even those cognitively intact, there is nonspecific binding in the lenticular nucleus, red nucleus, and subthalamic nucleus, possibly due to iron deposition [124], as well as in the upper portion of the cerebellum (Fig. 6).

Fig. 6
3 PET tomographs in the coronal, axial, and sagittal views for tau imaging. 18 F-flortaucipir images appear normal. Globus Palladius, substantia nigra, and cerebellum are the tau tracers.

“Nonspecific” uptake with the PET tau tracer 18F-flortaucipir. From left to right and projected on MRI, coronal, axial, and sagittal 18F-flortaucipir images from a cognitively normal 72-year-old man. Note uptake in the globus pallidus (coronal section, white arrows), substantia nigra (axial section, arrows), and superior portion of the cerebellum (sagittal section, arrow). None of these areas are known to harbor tau

Full size image

18F-Flortaucipir binds to tau in AD [126], which is associated with 3- and 4-repeat (3R and 4R) tau aggregates, but much less or not at all with 3R or 4R tau found in most varieties of tau-related FTD [122, 124]. The configuration of tau aggregates, which differs in various tauopathies [127], most likely determines binding. For instance, 18F-flortaucipir binds to patients harboring a p.R406W mutation in the MAPT gene, encoding tau [128]. This mutation results in 3R and 4R tau aggregates like those in AD [128]. 18F-Flortaucipir also binds weakly to the regions most affected in FTD cases and, particularly, in semantic dementia [129], but careful neuropathological evaluation has shown a lack of binding to 4R tau or to TDP-43 [124, 130, 131]. Furthermore, the signal in FTD involves the white matter, rather than the cortex, where the accumulation of misfolded proteins is greatest [108]. This binding has been postulated to correspond to MAO-B, abundantly expressed by astrocytes, but the 18F-flortaucipir signal has not been suppressed by blocking MAO-B [108].

Compared to 18F-flortaucipir, two commonly used newer tau PET tracers have less nonspecific binding to the lenticular nucleus, 18F-MK6240 and 18F-PI-2620. There is extensive experience with 18F-MK6240, which has less binding to choroid plexus than 18F-flortaucipir, thus allowing for a better quantification of tau deposition in medial temporal regions, including the entorhinal cortex [105]. A negative characteristic of 18F-MK6240 is the frequent intense binding to meningeal structures and to the skull (Fig. 7); various methods have been suggested to compensate for this binding [132]. Less experience exists with 18F-PI-2620, which also seems to bind to the meninges and skull [133]. 18F-PI-2620 has been postulated to bind not only to AD tau [134] but also to 4R tau as well and thus be useful in imaging corticobasal degeneration and progressive supranuclear palsy [133, 135].

Fig. 7
A PET MRI in the coronal, axial, and sagittal axis of the brain. 18F-MK6240 images are cognitively normal. Uptake is observed in the skull and meninges region while globus pallidus, and substantia nigra appear normal without any uptake.

“Nonspecific” uptake with the PET tau tracer 18F-MK6240. From left to right and projected on MRI, coronal, axial, and sagittal 18F-MK6240 images from a cognitively normal 73-year-old man. Please compare it with Fig. 6. Although there is no uptake in the globus pallidus (coronal section, white arrows), there is still uptake in the substantia nigra and uptake in the skull and meninges (red arrows)

Full size image

Tau accumulation measured with tau PET tracers correlates better with the degree of cognitive impairment than amyloid accumulation [136], a finding in agreement with prior neuropathological studies [137]. Furthermore, there is an inverse correlation between tau accumulation and brain metabolism: regions high in tau have uniformly depressed metabolism [13] (Fig. 1). This correlation is not as tight with amyloid accumulation (Fig. 1).

In amyloid-negative, clinically normal people older than 60, tau accumulation in the entorhinal cortex is associated with worse cognitive performance and greater tau in other brain regions [138].

2.6 Inflammation Imaging

Although brain inflammation is prominent in AD and related disorders, the use of inflammation imaging is not as widespread as that of previously described imaging biomarkers. Inflammation can be pathogenic, reflect scavenging of neurons and neuronal processes, or have a neuroprotective effect [139,140,141]. Animal models of tau-induced neuronal loss have shown earlier and more severe inflammation than models of increased amyloid [142], and both microglia and reactive astrocytes are found at autopsy to be increased in areas of the brain affected by neurodegenerative pathology. However, in vivo brain inflammation data in human neurodegeneration is scant. PET imaging allows in vivo quantification of neuroinflammation by measuring the density of the 18-kDa translocator protein (TSPO), which is expressed in microglia, astrocytes, and reactive endothelial cells. TSPO has been imaged with 11C-PK11195, a compound that in humans has a low affinity for the receptor [143] and a low ratio of specific-to-nonspecific binding [144]. The limitations of 11C-PK11195 prompted the development of second-generation radioligands for imaging activated microglia. 11C-PBR28 is a second-generation radioligand with a high affinity to TSPO, favorable in vivo kinetics, and greater signal-to-noise ratio than 11C-PK11195 in monkey brain [144]. Unfortunately, the affinity of this and other TSPO PET tracers is strongly determined by the rs6971 polymorphism on the TSPO gene, leading to high- and low-affinity groups, as well as an intermediate phenotype. More recently developed, 11C-ER176 has a higher affinity for TSPO and allows for imaging of people with the low-affinity rs6971 polymorphism of the TSPO gene [145, 146].

Using these tracers, increased brain inflammation has been documented even at pre-symptomatic stages of AD [147], with a good topographic correlation between inflammation and amyloid deposition (Fig. 8). At the MCI stage, many studies, for instance [148, 149], but not all [150] have shown neuroinflammation. The lack of consistency at the MCI stage may be related to a biphasic effect of inflammation, with earlier and later peaks [151], possibly neuroprotective at the early stages, but harmful at later stages. While this is still unclear, neuroinflammation seems to mediate tau spreading [152]. In dementing diseases more focal than AD, such as semantic dementia, inflammation has been shown to peak at the boundary between involved and healthy brain (Fig. 9), suggesting that inflammation plays an important role in the progression of neurodegeneration [108].

Fig. 8
A PET scan in coronal, sagittal, and horizontal view represents the diagnosis of pre-symptomatic Alzheimer’s disease. Amyloid deposition in the frontal lobe and precuneus with inflammation is highlighted in the tomography.

Amyloid and inflammation PET in a pre-symptomatic AD patient. Although this person was cognitively unimpaired, amyloid PET evidenced increased amyloid deposition in the frontal lobe and precuneus. Similar regions had inflammation on 11C-ER176 PET. This tracer binds to TSPO

Full size image
Fig. 9
2 MRI scans represent the inflammation in the cortex region with thickening in the core. The anterior region appears abnormal and the core is damaged.

Cortical thickness on MRI and inflammation PET in semantic dementia. On brain templates, in color are areas where a group of patients with semantic dementia differ from controls. Cortical thickness is most abnormal at the anterior portion of the left temporal lobe, in the core of the damage, while inflammation peaks at the periphery of the area with reduced cortical thickness. (From [108])

Full size image

In conclusion, the availability of imaging biomarkers for several of the major components of AD has greatly furthered the understanding of the development of this disease in humans. Furthermore, it has facilitated the performance of clinical trials that have recently yielded positive results. In terms of imaging, the development of tracers for alpha-synuclein and TDP-43, of great importance in LBD and FTD respectively, is being worked on. Furthermore, perfecting plasma biomarkers would greatly facilitate population screening, so that putative therapies could be applied to prevent or thwart the pathological processes causing irreparable neuronal loss in diseases leading to dementia.