Introduction

Dementia represents a syndrome characterized by multidomain acquired chronic cognitive impairment that has a profound impact on activities of daily living (ADL) [1]. Approximately 35.6 million people worldwide live with dementia, and the number is expected to triple by 2050 [2]. As life expectancy increases, the prevalence of dementia and the economic and social burden on healthcare systems is increasing worldwide [3]. Dementia is only the final stage of the disease, preceded by mild cognitive impairment (MCI). MCI is a stage of subtle cognitive changes that do not impact ADL and represents an important area of interest for drug development researchers [4]. People with MCI can still function independently, whereas the therapy that slows or stops its progression can improve ADL [5, 6]. Given the dementia consequences, there is an urgent need to treat, slow, or halt its progression through the development of improved diagnostic tools and disease-modifying drugs. Magnetic resonance imaging (MRI) has been indispensable in dementia follow-up, diagnosis, and treatment assessment, especially since the introduction of 3T MRI and its modern techniques. MRI may play a role in patient selection for new amyloid-reducing therapies such as aducanumab and also in diagnosing amyloid-related imaging abnormalities (ARIA) associated with amyloid modifying therapies [7]. Accordingly, the purpose of this work is to review the existing literature and describe distinct MRI changes in order to better utilize its results in the diagnosis and prognosis of already-developed dementia and its earliest stages.

Etiology

Based on the main pathophysiological mechanism, dementia can be classified into neurodegenerative (or proteinopathies) and nondegenerative dementia with numerous entities (Table 1) [1, 8]. Nondegenerative dementia is caused by structural brain pathologies or systemic diseases. The most common nondegenerative dementia is vascular dementia (VaD) [1]. By identifying causes other than degeneration, specific treatments can be initiated, and, in some cases, the cognitive decline can be stabilized or reversed. Neurodegenerative dementias represent the dominant group. Alzheimer’s disease (AD) as the leading phenotype occurs in 60%–80% of late-onset dementias, with a typical presentation that includes progressive memory loss. Another type is Lewy body dementia (LBD), which includes parkinsonism, fluctuating cognition, rapid eye movement sleep behavior disorder, and visual hallucinations. Substantial visuospatial, executive, and attention disorders in the absence of major memory deficits distinguish LBD from other types [1, 5, 9]. Next most common form is frontotemporal dementia (FTD), including three types: the behavioral variant FTD (bvFTD), the nonfluent variant primary progressive aphasia, (nfvPPA), and the semantic variant primary progressive aphasia (svPPA) [1, 9, 10]. In early-onset dementia, under the age of 65, the frequency of occurrence is variable, with AD still predominant, but, with a higher proportion of FTD and LBD. In the elderly, mixed pathology, most commonly consisting of AD and VaD, accounts for up to 30%–40% of all cases [9].

Pathophysiology

Neurodegenerative dementias are characterized by a slowly progressive course and the appearance of brain deposits of misfolded proteins [5]. This type of dementia could be caused by more than one abnormally folded protein, and each proteinopathy might have a unique clinical phenotype [11]. These proteins are considered to be the result of an imbalance in production and elimination. They activate brain immune cells, referred to as microglia, which initiate a “defence and repair” mechanism that ensures they are deactivated once their task is complete. This acute inflammatory process can, in an unknown way, progress to chronic inflammation if misfolded proteins accumulate [12]. Advances in the detection of molecular pathology have shown a crossover between phenotypes. There are tau aggregates in AD, but it is considered a secondary tauopathy because β-amyloid (Aβ) is more prevalent [11, 13].

Certain risk factors are included in this pathophysiological process, some of which are modifiable while others are not. The main nonmodifiable risk factors for dementia are older age and genetics [5]. The most commonly described genetic risks for AD include single nucleotide polymorphisms in the apolipoprotein E (APOE) region, with carriers of two copies of the APOE-ɛ4 alleles at increased risk compared with noncarriers [14, 15]. Another genetic risk involves some rare variants of the triggering receptor expressed on the myeloid cells gene (TREM), whose carriers have an increased risk for AD. Depending on the present TREM variant, microglia may play a protective or deleterious role. Rare variants, such as R47H and R62H, enable microglia deactivation [12]. Examples of modifiable risk factors include lack of physical activity, smoking, low education, social and mental inactivity, high blood pressure, diabetes, and inappropriate diet [3, 16].

Diagnosis

Dementia remains mainly a clinical diagnosis, with laboratory diagnostics and neuroimaging providing only supportive evidence. Patient assessment begins with a medical history [1]. According to the 2011 NIA-AA revised core clinical criteria and Diagnostic and Manual of Mental Disorders V, dementia can be diagnosed if behavioral or cognitive symptoms: 1) interfere with ADL; 2) represent a decline from prior ADL; and 3) are not caused by a mental illness. Cognitive or behavioral deficits involve at least one domain (memory, executive function, language, visuospatial abilities, and personality/behavior) [6, 17]. The examination should include a detailed mental state analysis. A combination of history-taking and objective cognitive testing is used to identify cognitive impairment. There is a wide range of tests, such as the Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA). Neuropsychological testing represents a gold standard in the assessment of cognitive impairment [1, 17]. The presence and severity of ADL decline are established by conducting a structured or unstructured functional evaluation survey [18].

Additional examinations (biochemistry and vitamins and hormone levels) should be performed to rule out systemic processes. The following tests should be performed in selected patients based on clinical indication and affordability: immunologic and microbiologic blood tests, cerebrospinal fluid (CSF) tests such as Aβ 1 to 40, Aβ 1 to 42, total tau and phospho-tau, electroencephalogram (EEG), and genetic analysis [1, 15, 16].

Neuroimaging includes structural and functional imaging: MRI, computed tomography (CT), and cerebral positron emission tomography (PET) scanning. MRI is preferred over CT due to higher sensitivity for vascular lesions and various forms of dementia. CT may be used for ruling out secondary causes of dementia in the initial diagnosis and also in cases when MRI is contraindicated, or if no previously obtained MRI scans are available for comparison with the present findings [1]. PET diagnostics is also important in the evaluation of patients with cognitive impairment. 18F FDG fluorodeoxyglucose-PET detects hypometabolism patterns in different types of dementia. Amyloid and tau PET with radiotracers that have high affinity for Aβ and tau proteins are less frequently used due to limited availability [1, 18].

MRI dementia protocol

A 3T field strength scanner should be preferred over 1.5T, whereas a single field strength should be utilized for all patients involved in research to allow for comparability [16]. An advantage of 3T is increased image resolution. The specific MRI receive head coil used should also be taken into account. There is a significant difference between a low and higher number of channels, depending on the manufacturer’s recommendation. For example, Siemens recommends 64 channel head coil for anatomical studies [19]. Researchers should be aware that different head coils can alter the results and thus lead to misinterpretation while comparing the findings [20]. According to a European-wide survey, over 90% of 193 institutions perform MRI dementia protocol [1, 16, 21] which is in agreement with the protocol used at our center (Table 2).

Microbleeds

Cerebral microbleeds (CMBs) represent persistent deposits of blood breakdown products, especially hemosiderin. On T2, they are detected as focused spots of signal loss, less than 10 mm in diameter, that grow in size on T2*-weighted gradient echo images (“blooming” effect) (Figure 1) [33, 34]. An alternative, more sensitive sequence is susceptibility-weighted angiography (SWAN), with greater “blooming” with more apparent but possibly unevenly formed CMBs. Filtered phase SWAN is used to distinguish CMBs from calcifications. On SWAN, both of these can be seen as hypointense. On the other hand, in the filtered phase, calcifications (diamagnetic) can be seen as hyperintense, whereas CMBs (paramagnetic) as hypointense lesions (Figure 1). However, depending on whether the filtered phase SWAN is used in a right- or left-handed coordinate system, CMBs may appear hyperintense on a phase image [35]. They have been associated with several pathologies, such as cerebral small vessel disease (cSVD), stroke, brain injury, radiation, and VaD. Although CMBs might be asymptomatic, early detection is critical in risk assessment of late cerebrovascular illness and cognitive deterioration [36].

The classification of CMBs includes three categories: deep (BG, thalamus, internal and external capsule, and corpus callosum) and periventricular white matter (WM), lobar (frontal, temporal, parietal, and occipital), and infratentorial (brainstem and cerebellum). Localization of CMBs contributes to the differential diagnosis between cerebral amyloid angiopathy (CAA) and hypertensive disease [35]. Lobar localization is more common in CAA diagnosis, where a preferential parietooccipital distribution has been reported. The deep category is more affected by the hypertensive disease of the small penetrating arteries [35, 37]. A solitary CMB in any brain region cannot be regarded as clinically significant. However, several lobar CMBs in a person over 55 years of age without additional explanations indicate a probable CAA. Multiple lobar CMBs are frequently detected in patients who meet the AD clinical criteria due to common CAA and AD co-occurrence. If superficial siderosis is also present, the diagnosis of CAA is strengthened. Cerebellar CMBs can occur in both CAA and hypertension and thus do not significantly contribute to diagnosis [9].

Lacunes and dilated perivascular spaces

A lacune is a lesion 2–20 mm in diameter, typically affecting the subcortical WM and the deep GM. It is a fluid-filled gliotic cavity that occurs prior to infarcts of the deep perforating arteries. Lacunes appear with CSF signal intensity on all sequences, bordered by a T2 hyperintensity rim. They should not be confused with dilated perivascular spaces (PVS), i.e., spaces filled with CSF that surround perforating arterioles and venules as they pass through the parenchyma out from the subarachnoid space. Dilated PVS is typically formed by volume loss in the surrounding tissue, with a predisposition for the BG [9, 42, 43]. Both are considered cSVD manifestations. Areas smaller than 2 mm, presented in the lower third of the corpus striatum and showing intensified CSF, can help in differentiating dilated PVS from lacunes [34]. There is further indication that lacunes may be linked to cognitive decline, death, and poor clinical outcomes. Lacunes were independently associated with cognitive function and strongly predictive of rapid global functional decline in a sample of older persons living independently [40].

Strategic infarcts

We have previously described the effects of lacunar infarcts on cognition. However, another location-dependent type of infarction can also be found in patients with dementia, especially VaD (Figure 2). Acute cognitive deterioration may be presented with a large cortico-subcortical ischemic lesion. However, dementia progression due to a stroke may be driven by infarcts in “strategic” regions that are important for normal cognitive functioning. Research on the relationship between brain lesion location and cognition found that the degree of cognitive impairment following stroke may be dependent on its features, such as type, size, number, location, and severity [9, 44]. Interestingly, a multivariate lesion-symptom mapping study showed that strategic infarction correlated with compromised cognition was of smaller size [45]. Strategic infarct localization includes the angular gyrus, bilateral anterior cerebral artery, paramedian thalamic, inferior MTL, parieto-temporal and temporo-occipital association areas, and angular gyrus, superior frontal, and parietal watershed areas in the dominant hemisphere. Also, HC or bilateral/unilateral thalamic infarctions have been found to cause dementia [9, 44]. In the previously mentioned mapping study, the angular gyrus, BG, and the WM surrounding BG, predominantly on the left side, had the highest correlation with global decreased cognition [45].

Standardized assessment

As qualitative evaluation could result in increased ambiguity and interobserver variability, several efficient, repeatable, and cost-effective semiquantitative scales have been proposed [13]. They have been created to provide reproducible semiquantitative volume loss evaluation. The scales can be useful in everyday imaging and may increase diagnostic precision [46]. A variety of scales have been developed, not only for atrophy assessment but also for CMBs and WMHs. We will mention only a few that are the most common in the literature. It is also important to include advanced techniques, such as arterial spin labeling (ASL), which has become increasingly important in the evaluation of dementia and volumetry analysis.

A radiological report should contain [46]:

• Scales, at least: MTA scale; Global cortical atrophy (GCA) scale; Koedam score; Fazekas scale;

• Extent and location of infarcts;

• CMBs;

• WMHs;

• Other alterations (tumor, hydrocephalus, subdural hematoma, and other possible causes);

• Comparison with prior assessments.

MTA scale

The degree of MTA can be assessed using the Scheltens scale (Table 3) on coronal planes perpendicular to the long axis of the HC [42]. Figure 3 shows the Scheltens scale, which focuses on three MTL features: the choroid fissure, the temporal horn width, and the HC height. Both sides are evaluated independently [47, 48]. Over time, the Scheltens scale has been modified. For example, Galton et al. [49] expanded it to include non-HC structures. The scale has been divided into two sections, the first utilizing the Scheltens scale and the second designed to grade the anterior, non-hippocampal medial (the parahippocampal gyrus), and lateral temporal regions of each hemisphere independently [47, 49]. Urs et al. [47, 50] and Duara et al. [47, 51] have independently created a grading system focused on a single identifiable landmark slice at the level of the mamillary bodies.

Fazekas scale

Fazekas scale assessment is the most often used rating scale for determining the severity of WMHs. It also represents a four-step scale from 0 to 3 (Table 6). The whole WMH assessment can be seen in Figure 6. It plays a role in evaluating post-stroke cognitive impairment since grades 2 and 3 have been associated with disability 90 days and one year after ischemic stroke [9]. Also, The Radboud University Nijmegen Diffusion Tensor and Magnetic Imaging Cohort study demonstrated that a greater grade indicates a greater pathologic burden, corresponding to a lower MMSE score and a decline in cognitive functioning [43].

Table 6: Divided Fazekas scale for assessment of the deep and periventricular white matter

Fazekas	0	1	2	3
Whole WMHs	Sporadic or nonpunctate WMHs	Multiple punctate WMHs	Confluent lesions bridging the punctate WMHs	Extensive confluent WMHs
PWMHs	Absent WMHs	Thin WMHs running along ventricular horns (caps and bands)	More extensive WMHs (smooth hallo)	Irregular periventricular signal extending into deep WM
DWMHs	Absent WMHs	Punctate foci	Foci that started to confluent	Large confluent areas of WMHs

WMHs: White matter hyperintensities; PWMHs: Periventricular white matter hyperintensities; DWMHs: Deep white matter hyperintensities.

There is a division of the Fazekas scale into DWMHs and PWMHs scales (Figure 6) [58]. Some authors use this divided form, while others use only the overall Fazekas score for the whole WM. As shown in Table 6, there are some differences between the whole and divided scale. The whole scale considers DWMHs to a greater extent, whereas the first two grades of PWMHs are almost considered normal in the whole WM form of the Fazekas scale. Research has found that DWMHs have greater and more correlative diagnostic value to clinical presentation value (see: White matter hyperintensities); however, specific guidelines on which scale to use have not yet been established, and in both scales, there is a possibility to underdiagnose or overdiagnose patients [59].

Volumetry

Specialized programs that could provide volumetric evaluation are also available. Volumetry is not only used for evaluating the GM and WM volume but also for CMBs, WMHs, and ischemic lesions [13]. The volBrain is an example of volumetry software, and its segmentation process is presented in Figure 7 [63]. Another commonly used volumetric tool is FreeSurfer [64]. NeuroQuant, the FDA-approved software for routine dementia assessment, is also worth noting. It contains a standardized assessment, with the possibility of comparing MRI scans to national databases with comprehensive patient information [65]. Pemberton et al. [66] found that using quantitative reporting in addition to visual evaluations enhances sensitivity and precision for identifying volume loss compared with visual assessments alone. Progression of MTA, quantified through the HC volume, has been investigated in both control and preclinical, prodromal, and probable AD patients [67]. Volumetry and MTA comparisons have revealed correlations ranging from acceptable to excellent. Hippocampal volumetry can help differentiate HC sclerosis in epilepsy and dementia, especially FTD and AD. Volumetric analysis of HC subfields showed that sclerosis develops in different parts of the HC [68]. Another important differential diagnosis includes some other causes, such as normal pressure hydrocephalus. Volumetry has improved diagnostic accuracy of differentiating normal pressure hydrocephalus from AD and Parkinson’s disease but it could also be used to predict outcomes after shunting [69]. Based on the European-wide survey, methods such as volumetry are currently mostly applied in research settings, and only 5.7% of all centers perform it regularly, even though these programs were introduced years ago [21].

Normal aging

In older persons, the cognitive decline spectrum ranges from normal age-related cognitive decline to subjective cognitive impairment (cognitive complaint with normal cognitive screening test) to MCI and dementia [17]. Hence, the question is, what specific degree of the abovementioned MRI changes should be considered normal? Brain atrophy and cSVD imaging markers were the most prevalent imaging findings in aging that coincide with those reported in dementia [9]. The basic rule for distinguishing age-related atrophy from neurodegeneration is that age-related atrophy is typically symmetric and widespread, whereas dementia exhibits unique patterns and asymmetries. Semiquantitative visual rating scales could assist in distinguishing between “normal” and “abnormal.” In addition, finding a solitary CMB is not considered relevant; however, multiple ones should arise suspicion [35]. On the contrary, cerebrovascular alterations are prevalent in elderly. Some criteria to diagnose VaD proposed by the National Institute of Neurological Disorders and Stroke and Association Internationale pour la Recherché et l’Enseignement en Neurosciences are: a) confluent WMHs affecting 25% of the total WM; b) lacunar infarcts affecting the frontal WM, multiple BG, and both thalami; or c) large-vessel infarcts affecting strategic regions [9]. Nevertheless, every change should be noted and interpreted along with the clinical findings.

From mild cognitive impairment to dementia

MCI diagnostic criteria include impaired cognitive performance in one or more domain; however, with proper ADL [17]. The yearly rate of progression to dementia is approximately 15% [70]. But how can we recognize a patient transitioning from MCI to dementia on neuroimaging? According to quantitative MRI studies, HC atrophy is detectable before dementia starts and worsens with conversion to clinically manifest disease. A longitudinal study including MCI patients has shown that memory decline over time was associated with bilateral HC atrophy, while left HC atrophy predicted a visuospatial decline [71]. Pyun et al. [72] reported that PA in amyloid-positive MCI is substantially linked with dementia development, regardless of the presence of MTA. This demonstrates the predictive usefulness of PA for progression in amyloid-positive MCI patients and suggests that PA can be used as a progression predictor from amyloid-positive MCI to dementia.

Advanced MRI techniques

Diffusion-weighted MRI images (DWI) are applied to identify regions with diffusion limitations. Tissues that are highly cellular or affected by cellular swelling have lower diffusion coefficients than normal. DWI has been particularly significant for the diagnosis of CJD and VaD, and it also allows the detection of differential diagnoses, such as ischemia, encephalitis, and neoplasia [9]. Shiga et al. [73] concluded that DWI has higher sensitivity in detecting CJD than FLAIR, T2W, EEG, CSF proteins, or neuron-specific enolase. DWI provides quantitative lesion assessment by evaluating the apparent diffusion coefficient (ADC). In CJD, DWI shows limited diffusion in at least two cortical areas (“ribboning”) and/or restricted diffusion in the caudate nucleus, followed by the putamen and thalamus. Using DWI would enable distinguishing different types of CJD, such as “pulvinar sign” in variant CJD—a signal on FLAIR and DWI of higher intensity in the posterior than in the anterior putamen [74]. Taoka et al. [75] noted that ADC of the uncinate fascicles also linked well with cognitive performance and appeared to be useful as biomarkers for detecting the progress of AD.

Diffusion tensor imaging (DTI) studies on the microstructural features of WM and its derived metrics have been widely used in the investigation of AD and MCI. Derived metric, such as fractional anisotropy (FA), is indicative of water diffusion along WM tracts that is radially constrained by the myelin sheath enclosing axons (Figure 8) [76].

Different parts of corpus callosum are affected in various types of dementia since different parts of the cortex are affected, and degeneration of WM fibers represents a consequence of neuronal loss in the cortex [77]. Shim et al. [78] demonstrated the genu FA values as an important predictor of global cognition in all dementia and MCI groups. Atrophy and FA alteration in the genu and splenium are commonly seen in AD, and these abnormalities are considered to occur in the initial stage of AD. The genu receives axons from the prefrontal cortex and myelinates later than the splenium, which receives axons from the temporoparietal lobes, which typically exhibit atrophy and hypometabolism in AD.

Derived DTI metrics are also the sensitive cSVD indicators for detecting early cognitive deterioration. Subtle changes in some WM tracts can be identified from the preclinical stage of vascular MCI, with a localized to generalized spreading pattern during development [79].

Magnetic resonance spectroscopy (MRS) provides a noninvasive biochemical examination of the brain in vivo (Figure 9). Metabolites N-acetyl aspartate (NAA), myo-inositol (MI), total choline (Cho), and total creatine (Cr) are present in the brain in amounts suitable for detection and have been commonly studied. It has been shown that in AD, the NAA peak decreases while MI and Cho peak increases, particularly in the posterior cingulate and HC. This pattern is inconsistent across the studies, and only the NAA drop seems to be significant [80]. The MRS results have shown some variation, which is higher in the MTL due to the bone at the base of the skull as well as the air in the sinuses, which is why its clinical efficacy in dementia is lacking [1]. In the literature, limitations to MRS usage are technical variances in data collection, varied voxel composition, the existence of underlying pathology in each voxel, constraints of lower field MRS, and voxel location changes. Even if there are some alternatives (ultra-high field MRI), standardization is needed [81, 82].

Some commercial MRS processing software packages use linear-combination modeling (LCM) of in vitro spectrum to maximize the use of prior information in the analysis. LCModel is among the most quoted. However, the application of software also involves some drawbacks, such as a moderate agreement between algorithms for the same metabolites and poor understanding between standardized model parameters and MRS results [83]. Although MRS does not add much to the initial diagnosis, a meta-analysis concluded that it may play a role in forecasting progress when comparing before and after MRS results [84].

Functional MRI (fMRI) generates brain activity via a bold oxygen level-dependent (BOLD) signal that measures blood flow and volume shifts. In comparison to controls, AD patients show no or less stimulation of the HC and other MTL structures while doing memory tasks. Activation of the parietal and posterior cingulate sulci suggests some compensation. While offering insight into pathophysiology, fMRI clinical usage is not supported due to limitations such as a low signal-to-noise ratio and the debatable relevance of BOLD as a neuronal activity measure. Variability may be caused by uncontrolled hemodynamic factors [85].

Three-dimensional pseudo-continuous ASL provides us with absolute and relative cerebral blood flow (CBF) measures and might be able to identify functional deficits in the brain, similar to FDG-PET and SPECT [86]. Its values have been demonstrated to be comparable with fMRI-BOLD imaging; however, unlike BOLD, ASL offers a quantitative evaluation reflecting brain physiology and neuronal activity [87]. Maps of CBF are often qualitatively assessed with image statistical approaches. Hypoperfusion in the posterior cingulate and precuneus, along with the temporoparietal association cortex, seems to be common AD finding (Figure 10) [88]. Hypoperfusion is commonly identified in bilateral frontal lobes of FTD patients. In comparison to FTD, AD patients had apparent hypoperfusion in the posterior cingulate cortex, and both the sensitivity and specificity in discriminating presenile AD from FTD exceeded 70% [87]. The lower precuneus and cuneus perfusion in LBD has been related to poor overall clinical scores, and the hypoperfusion pattern seems to be comparable to the hypometabolism pattern on FDG-PET [89].

Furthermore, ASL has substantial advantages over PET and SPECT: it is noninvasive and does not expose to ionizing radiation, intravenous contrast agents, or radioactive isotopes; it can be conducted fast with most MRI scanners and repeated quickly. Furthermore, perfusion and structural images could be obtained in the same scan session [86].

The ASL values might change with aging. According to the existing research, younger patients had higher CBF values than seniors. CBF decreases with age and may be caused by reduced cerebral metabolism induced by brain tissue loss. The patient’s age, as well as vascular risk factors, such as carotid disease, should be considered while examining the absolute CBF perfusion values [90].

Use of MRI—is there a purpose?

Indicated on time, MRI could lead to earlier diagnosis, thus enabling effective treatment and modification of risk factors. Research indicates that a high-quality diet, regular exercise, and a healthier lifestyle can delay the progression of dementia, reduce the chance of developing symptoms, and positively impact MCI [91]. Dietary changes, such as the Asian and Mediterranean diets that are high in unsaturated fats and antioxidants, as well as diets high in vitamins (E, C, B6, D, B12, and folate), fruits and vegetables, accompanied by caloric restriction, have also provided some beneficial results [92].

Clinical trials and therapy evaluation might certainly benefit from neuroimaging. Current studies, mainly considering AD, rely on clinical assessments of cognition and behavior, which vary from day to day and from examiner to examiner. According to Chao et al. [86], ASL-detected hypoperfusion can forecast eventual clinical and cognitive deterioration and may be beneficial in selecting candidates for future AD therapy studies. So far, cholinesterase inhibitors have been mainly used. However, recently the most intensively studied medications include amyloid beta-directed antibodies, such as aducanumab, a medication for early AD or MCI, authorized by FDA. It is the first medication aimed at the underlying AD pathophysiology. According to the results of clinical trials, a reduction in the plaques associated with AD is likely to reduce cognitive and clinical deterioration [93]. An MRI is required to identify ARIA caused by amyloid-modifying drugs. ARIA that could be caused by amyloid beta-directed antibodies include ARIA-E: parenchymal oedema and sulcal effusions, and ARIA-H: microhaemorrhages and localized superficial siderosis. To enable early detection of these side effects, the FDA suggested MRI prior to the application of the initial, 7th, and 12th treatments [7].

Another essential topic is the application of artificial intelligence (AI) algorithms in collaboration with MRI methods. Studies have already shown that employing AI alongside MRI evaluation in achieving precise diagnosis of dementia, including early identification and distinction between different types of dementia, increases diagnostic accuracy. So far, AI has been rarely applied; however, due to its high diagnostic value, it will undoubtedly be used in everyday dementia diagnostic assessment [94].

Conclusion

MRI has become one of the key diagnostic elements, not only in diagnosis but also in monitoring the therapeutic response of patients with dementia. Semiquantitative scales have been quite useful in the assessment of structural MRI; nevertheless, for even greater precision, they need to be complemented by other advanced techniques. Volumetric analysis and ASL have demonstrated a capacity to improve not only the diagnosis but also the precision in the prognosis of dementia. ASL may be used in cognitive decline prediction and may be effective in selecting candidates for future dementia therapies. However, further research is necessary for even greater usability of the abovementioned MRI techniques.