Alzheimer's disease is characterized by the accumulation of two abnormal proteins: extracellular Aβ protein and intracellular tau protein (34). Amyloid and tau deposition progress spatiotemporally in a predictive manner. Amyloid first accumulates in the basal part of the frontal, temporal, and occipital lobes, and subsequently spreads to the entorhinal cortex, hippocampus, amygdala, insular cortex, and cingulate cortex, sparing the primary visual and sensorimotor cortices. Conversely, neurofibrillary tangle deposition progresses in the following order: transentorhinal cortex, entorhinal cortex, hippocampus, temporal cortex, association cortices, and finally the primary sensorimotor and visual cortices (5).

The Aβ hypothesis, the dominant theory of AD, suggests that overproduction or inadequate clearance of Aβ is a causative factor for AD; AD begins with the abnormal metabolism of the transmembrane amyloid precursor protein (APP). β- and γ-secretases cleave APPs to form several Aβ peptide fragments (1). Of these, the most important is Aβ₄₂, which is highly prone to aggregation and resultant plaque formation (6). Although amyloid deposits are typically observed in the extracellular space, Aβ is also found within neurons, and this may be related to the aggregation of other cellular proteins such as tau protein in AD (7). Subsequently, abnormal phosphorylation of the microtubule-associated tau protein in neurons and the formation of neurofibrillary tangles are thought to result in the disruption of normal neuronal function (1). Oxidative and inflammatory stresses from Aβ also contribute to the loss of synaptic and neuronal integrity, and finally, neuronal loss and brain atrophy. This downstream pathological cascade has been re-interpreted by the hypothetical model of Jack et al. (8) (Fig. 1). Conversely, Braak and Del Tredici (9) observed hyperphosphorylated tau protein in the absence of Aβ deposition in the medial temporal limbic isocortex of young individuals. Furthermore, recent evidence suggests that tau deposition is a requisite for amyloid toxicity in vivo (10). These findings raise questions regarding the role of Aβ as an initiator of the AD pathophysiological cascade.

The time course of AD is approximately 20–30 years from preclinical (cognitive normal) to prodromal (mild cognitive impairment [MCI] or pre-dementia) to overt AD. The prodromal phase of AD is commonly referred to as MCI, which is characterized by the onset of cognitive symptoms (e.g., memory or other cognitive dysfunctions) that do not meet the criteria for dementia (811). The prevalence of MCI in non-dementia individuals over the age of 70 years is approximately 15% with a 2:1 ratio of amnestic to non-amnestic types (1213). Previous studies have indicated that approximately 10–15% of MCI patients progress to AD yearly, and predictors of this conversion include the APOE4 allele of the apolipoprotein E gene (APOE), clinical severity, brain atrophy, certain cerebrospinal fluid (CSF) biomarkers, changes in cerebral glucose metabolism, and Aβ deposition (14). The National Institute on Aging-Alzheimer's Association recently developed new diagnostic criteria for AD, MCI due to AD, and preclinical AD that integrate biomarker evidence into the diagnostic frame (1516). In these criteria, concomitant observation of amyloid deposition (PET or CSF) and neuronal injury (tau, FDG-PET, MRI) indicates a high likelihood of MCI due to AD, while the presence of only one of these factors indicates an intermediate likelihood of MCI due to AD (16). Recent International Working Group-2 criteria further simplified the diagnosis of AD based on the requirement of only the presence of an appropriate clinical phenotype at any stage and presence of amyloid biomarker (e.g., decreased Aβ in CSF, or increased tracer retention on amyloid PET) (17).

Structural MR imaging is recognized as an important diagnostic tool for AD, as it can both differentiate AD from non-AD dementia and potentially estimate preclinical or prognostic tissue damage in vulnerable regions such as the hippocampus and entorhinal cortex (18). Structural MR imaging is also important for the exclusion of meningioma, glioma, subdural hematoma, vascular malformation, and normal pressure hydrocephalus (19).

Specific patterns of cortical atrophy and vascular pathology distinguish typical AD from other neurodegenerative dementias including frontotemporal dementia (FTD), corticobasal degeneration (CBD), progressive supranuclear palsy (PSP), and vascular dementia (19). Thus, identification and classification of cortical atrophy patterns and vascular pathology are of utmost importance for accurate and effective usage of MR imaging in patients with cognitive impairment.

White matter hyperintensity (WMH) has emerged as a strong correlate of cognitive aging and AD (596061). In 1987, Fazekas et al. (62) first described WMH as a periventricular halo, punctate, or early confluent hyperintensity observed in AD. WMH later became known by other terms such as leukoaraiosis, white matter lesion, leukoencephalopathy, and white matter disease, with various definitions that have hampered the integration of previous research on risk factors, pathophysiology, pathological correlations, and clinical implications (63). Recently, the Standards for Reporting Vascular Changes on Neuroimaging (STRIVE) defined WMH as a presumed vascular-origin signal abnormality of variable size in the white matter, showing hyperintensity on T2-weighted images without cavitation, exclusive of subcortical grey matter or brainstem lesions (63).

White matter hyperintensity can be classified as smooth periventricular or irregular periventricular (located in the periventricular system) or deep white matter lesions (found in the subcortical white matter) (6465). Postmortem studies have reported that smooth periventricular WMH presents as a loosening of the fiber network with myelin loss around the tortuous venules, irregular periventricular WMH demonstrates confluent areas with fiber and myelin loss around the perivascular space with reactive gliosis, and deep WMH shows perivascular rarefaction of myelin with varying degrees of axonal loss, astrogliosis, and spongiosis (6667686970). These findings imply that smooth periventricular WMH is likely caused by altered periventricular fluid dynamics, while irregular periventricular WMH and deep WMH reflect ischemic tissue damage (677071). Recent studies have provided strong evidence that WMH plays a causative role in cognitive decline and dementia (72737475), and might lead to AD by contributing in an additive or synergistic manner to disease pathogenesis (597677).

White matter hyperintensity has been considered to be a primary pathology in subcortical ischemic vascular dementia. However, the histopathological features of WMH are also comparable with that of AD, suggesting that subcortical ischemic vascular dementia and AD may be part of a pathological continuum (7879). There are only few pure conditions of AD and subcortical ischemic vascular dementia that have been reported, and rather varying degrees of neurodegenerative and vascular pathologies comprise the spectrum of AD and vascular dementia (80). Furthermore, AD and subcortical ischemic vascular dementia share common risk factors such hypertension, atherosclerosis, diabetes, hypercholesterolemia, obesity, metabolic syndrome, and smoking (80).

The severity of WMH can be quantified semiquantitatively or quantitatively, but a gold standard has not been established until date. The Fazekas scale is most commonly used, which distinguishes periventricular versus deep WMH and scores them according to a 4-point scale (0–3) (Fig. 8) (62). Another well-known scale is the Age-Related White Matter Changes scale, which defines WMHs as ill-defined hyperintensities ≥ 5 mm and scores them according to a 4-point scale in five different regions of each hemisphere (0–30) (81). Fazekas et al. (82) also proposed a WMH grading scale based on type, location, number, and size (0–84). Another new rating system according to the ratio of periventricular versus deep WMHs has been recently proposed by the Clinical Research for Dementia of South Korea (83). These semiquantitative scales have shown good intra-observer and inter-observer agreement and close correlations with volumetric assessments; however, they were designed to measure WMH at a single time point and are not suitable for evaluating WMH progression (848586).

The term "lacune" is derived from the French word for a small fluid-filled cavity, and has historically represented the late stage of a small deep brain infarct (87). The STRIVE criteria define lacunes as a "round or ovoid, subcortical, fluid-filled (similar signal as CSF) cavity measuring between 3 mm and 15 mm in diameter, consistent with a previous acute small deep brain infarct or hemorrhage in the territory of one perforating arteriole" (37). In most cases, lacunes have a CSF-like hypointensity at the center with a surrounding rim of hyperintensity on fluid-attenuated inversion recovery images, which is helpful for differentiating them from enlarged perivascular space along size (< 3 mm) and location criteria (lower basal ganglia region) (888990). Lacunes are assumed to exist as remnants of symptomatic or silent small subcortical infarcts, appearing central to the most common cause of vascular dementia (37). Though there is no consensus regarding the number or location of lacunes required for diagnosis of vascular dementia, two or more lacunes outside the brain stem are considered to be sufficient to support the diagnosis (91).

Lacunes are known to be associated with stroke and gait disturbance in elderly individuals (9293). Furthermore, patients with AD are more likely to have lacunes on MRI than individuals without dementia (7794). Previous pathological studies have also revealed that patients with lacunes are more likely to have dementia for the reason that patients with infarcts require a lower burden of plaques and tangles for clinical diagnosis of AD (959697). Moreover, the location of lacunes is an important factor in dementia, as thalamic and basal ganglia lesions are associated with poor cognition (98). The underlying mechanism relating lacunes and cognitive impairment remains unclear, and may be resultant from the coexistence of other factors such as WMH, cortical microinfarcts, or other unidentified factors (99).

Microbleeds are well-known MRI expression of small vessel disease as well as lacunes and WMH. Cerebral microbleeds refer to small, round, or ovoid hypointensities, ranging 2–10 mm in diameter on T2^* gradient-recall echo (GRE) or susceptibility-weighted (SWI) MR sequences, which provide a high contrast-to-noise ratio between the brain parenchyma and paramagnetic materials (100101102). Histopathological-radiological correlation studies have shown that hypointensities found on MRI correlate with areas of hemosiderin deposition with impaired vessel integrity, such as in the case of lipofibrohyalinosis or vascular Aβ deposition in patients with dementia (103104). These hypointensities can be classified into two different patterns of distribution according to their etiology and location: deep or infratentorial location with the presence of vascular risk factors (e.g., hypertension), and lobar (cortico-subcortical) location associated with APOE ɛ4 carrier status or cerebral amyloid angiopathy (CAA) (Fig. 9) (103105106).

With regard to dementia, microbleeds have drawn research attention as potential mediators of cognitive impairment. Microbleeds are more commonly found in patients with MCI, AD, and vascular dementia (107108109110). The mechanism of cognitive impairment due to microbleeds is not fully understood, but is hypothesized to be due to focal dysfunction or damage to the surrounding tissue, or due to more generalized processes such as small vessel disease or extensive CAA (103111). According to the Boston Criteria, for patients older than 55 years who present with lobar intracerebral hemorrhage, the imaging diagnosis of CAA can be made based on either MRI or CT findings: 1) probable CAA: multiple hemorrhage restricted to lobar, cortical, or cortico-subcortical regions; and 2) possible CCA: single lobar, cortical, cortico-subcortical hemorrhage (112).

Microbleeds are also frequently encountered in the context of Aβ immunotherapy for AD (amyloid-related imaging abnormalities) as an important side effect (113). For this particular immunotherapy, Food and Drug Administration guidelines recommend exclusion of patients with more than five microbleeds, which can be evidence of underlying CAA.

Reports of inter-observer agreement on the presence of microbleeds on conventional T2^* GRE MR imaging are varied (κ = 0.33–0.88), likely due to detection difficulties (114). Calcification, cavernous malformations, cross-sectional pial vessels, hemorrhagic metastases, and diffusion axonal injury can mimic the appearance of microbleeds on MR imaging, such that a diagnosis should be made in the context of a patient's clinical history and with consideration of other MR sequence findings (63115). Furthermore, T2^* GRE versus SWI parameters, resolution, field strength, and post-processing technique can affect the observed number of microbleeds (116117); SWI MR imaging is well-known for detection of a larger number of microbleeds than T2^* GRE MR imaging as it offers higher resolution and an increased susceptibility effect (117118119). Therefore, future research should focus on the development and validation of controlled standards for the diagnosis of microbleeds in order to strengthen our understanding of their utility as risk and prognostic factors.

Emerging evidences shows that cerebral microinfarcts are associated with cardiovascular risk factors as well as cognitive decline and dementia, particularly if they are multiple and cortical (120121). Moreover, cerebral microinfarcts can occur in cognitively normal elderly people, but the prevalence is higher in AD patients (122).

Cortical microinfarcts are defined as sharply defined microscopic regions of cellular death or tissue necrosis, ranging from 50 um to 5 mm as per histopathology studies. Recently, several studies aimed to detect microinfarcts on in vivo 7T MRI (123124). According to a previous study, a cut-off value of five microinfarcts on 7T MRI could differentiate AD and controls with sensitivity of 64.3% and specificity of 94.4% (124). When using 3T MRI, only 25% of cortical microinfarcts identified on 7T were detectable, probably due to limited spatial resolution (123). Consequently, it is necessitated to improve detection of cortical microinfarcts lesion on widely used 3T MRI by protocol optimization to reflect exact burden of microinfarcts in clinical practice.

The first thing that must be determined during potential imaging diagnosis of dementia is whether the patient has any focal lesions such as those due to stroke, tumor, or vascular malformation, or if the patient has hydrocephalus (19125). Second, the presence of MTA must be determined, as MTA indicates a high probability of AD. Third, the degree of WMH and the presence of microbleeds should be assessed in order to verify AD or classify the type of dementia (vascular dementia, CAA, and other dementias). If a patient is categorized as having an "other dementia," the spatial pattern of brain atrophy must be more carefully evaluated. FTLA is indicative of FTD. If the FTLA has a right predilection, behavioral variant FTD is indicated, whereas a left predilection indicates SD. Peri-insular atrophy with a left predilection or less commonly, with a right predilection, indicates progressive non-fluent aphasia. Alternatively, presence of parieto-occipital atrophy suggests neurodegenerative diseases such as PCA, CBD, or dementia with Lewy bodies. Infratentorial atrophy with relative sparing pons and prominent atrophied midbrain is indicative of PSP (Fig. 10A) (126). The presence of brain atrophy with markedly disproportionate ventriculomegaly suggests normal pressure hydrocephalus (Fig. 10B) (127). When the cortex and basal ganglia show hyperintensity on diffusion weighted imaging (DWI), Creutzfeldt-Jakob disease is indicated (Fig. 10C) (128). A flowchart of the above approach to the imaging diagnosis of cognitively impaired patients is provided in Figure 11.

For the reason that most patients with suspected dementia are elderly and represent a distinct population, a tailored MRI protocol is mandatory. First, imaging markers that reflect important neurohistopathological features of AD should be included. Second, imaging markers that can be used to demonstrate functional changes related to AD should also be incorporated. Third, imaging markers must be identified and utilized to monitor the therapeutic response of AD pathology. Fourth, contrast administration should be avoided unless there are other suspicious imaging findings and MRI should be performed within proper acquisition time. Finally, considering its association with subcortical ischemic vascular dementia, the necessity of MR angiography should be revisited in elderly patients. For MR examinations of patients with suspected dementia, consistency of imaging parameters across subjects and across time are also important in terms of longitudinal follow-up and multicenter clinical trials (146147). We have provided such an MR protocol, which was initially devised for the Korea-Alzheimer Dementia Neuroimaging Initiative (K-ADNI) study by modification of ADNI protocols (Table 4) (148).