The Utility of Arterial Spin Labeling MRI in Medial Temporal Lobe as a Vascular Biomarker in Alzheimer’s Disease Spectrum: A Systematic Review and Meta-Analysis

We sought to systematically review and meta-analy the role of cerebral blood flow (CBF) in the medial temporal lobe (MTL) using arterial spin labeling magnetic resonance imaging (ASL-MRI) and compare this in patients with Alzheimer’s disease (AD), individuals with mild cognitive impairment (MCI), and cognitively normal adults (CN). The prevalence of AD is increasing and leading to high healthcare costs. A potential biomarker that can identify people at risk of developing AD, whilst cognition is normal or only mildly affected, will enable risk-stratification and potential therapeutic interventions in the future. All studies investigated the role of CBF in the MTL and compared this among AD, MCI, and CN participants. A total of 26 studies were included in the systematic review and 11 in the meta-analysis. Three separate meta-analyses were conducted. Four studies compared CBF in the hippocampus of AD compared with the CN group and showed that AD participants had 2.8 mL/min/100 g lower perfusion compared with the CN group. Eight studies compared perfusion in the hippocampus of MCI vs. CN group, which showed no difference. Three studies compared perfusion in the MTL of MCI vs. CN participants and showed no statistically significant differences. CBF measured via ASL-MRI showed impairment in AD compared with the CN group in subregions of the MTL. CBF difference was significant in hippocampus between the AD and CN groups. However, MCI and CN group showed no significant difference in subregions of MTL.


Introduction
Dementia is a neurodegenerative syndrome that predominantly affects older adults. It affects cognition irreversibly while consciousness remains intact. Cognitive domains, namely episodic memory, language, executive functions, and visual-spatial skills deteriorate progressively [1]. Patients become gradually more dependent on relatives and carers for essential daily-living tasks. As life expectancy is extended, the prevalence rate of dementia increases. According to the World Health Organisation, 50 million people worldwide are living with dementia, and this number is anticipated to reach 152 million by 2050 [2]. In 2019, approximately 885,000 dementia patients were living in the UK and their overall care cost for that year was estimated at GBP 34.7 billion [3]. By 2040, UK patients will have risen to 1.6 million and the cost of their care will have reached GBP 94.1 billion [3]. As no preventive treatment is available, a better understanding of the disease process is needed if we are to decelerate the degenerative process by applying pharmaceutical interventions. To this end, research on potential biomarkers and preclinical diagnosis will underpin the research in treatment.
The neurodegenerative process of AD begins years or even decades before symptomatic disease. According to NIA-AA, it can manifest with various symptoms. These include reduced ability to acquire and maintain new information, impaired management of complex tasks, reduced visuospatial skills, impaired language function, as well as personality changes [4]. AD is considered a continuum in which earlier asymptomatic stages can be detected on a biological level [5]. Towards this direction, the National Institute on Aging and Alzheimer's Association (NIA-AA) proposed a research framework of diagnostic biomarkers: the AT(N) classification. AT(N) incorporates several imaging and biological tests that act as biomarkers. Individual biomarkers are detected in various stages of the AD continuum and are complementary to the clinical criteria in the diagnosis of AD dementia and its earlier stages [6,7].
Recent research has revealed emerging biomarkers related to the vascular component of AD. According to the vascular hypothesis, neurodegeneration in AD is the outcome of vascular disturbances in the brain [8]. Cerebrovascular impairment plays a key role in AD pathogenesis. Chronic brain hypoperfusion causes oxidative stress and inflammation [9]. Oxidative stress can cause blood-brain barrier impairment which reduces the clearance of beta amyloid. In addition, there is a higher expression of proinflammatory cytokines-IL1β and TNFa-in brain cells of MCI and AD patients, compared with controls. In turn, accumulation of beta amyloid triggers-through the production of IL-6-the differentiation of microglial cells, resulting in further neuroinflammation and blood-brain barrier dysfunction [10]. Badji et al. (2020) suggested the incorporation of biomarkers which reflect vascular deficiencies, such as lacunes, WMH, microbleeds, and cerebral blood flow changes [11]. Identifying groups of people with the aforementioned changes offers a wider window for interventions to delay cognitive impairment.
CBF is considered a potential vascular biomarker (V). CBF can be measured through different imaging modalities [12]. Invasive available techniques are single-photon emission tomography (SPECT), H2 (15)O positron emission tomography (PET), Xenon-enhanced CT, and dynamic susceptibility contrast magnetic resonance imaging (DSC-MRI). These modalities use exogenous contrast agents [12]. SPECT and FDG PET have been widely used in the diagnosis of MCI and AD for over 3 decades [13]. SPECT showed a specificity of 80% and a sensitivity of 90% in distinguishing AD patients from cognitively normal controls. For the same purpose, FDG-PET's sensitivity was 85% and specificity 89% [14]. O' Brien et al. directly compared the efficacy of the aforementioned modalities, showing statistically significant superiority of FDG PET compared with SPECT [15]. Cost-effectiveness of both SPECT and PET is still an area of debate. There are groups who maintain that early diagnosis with these two modalities is not cost-effective due to the lack of definitive treatment and the low estimation of quality-adjusted years. On the other hand, studies utilizing PET in the US and SPECT in the EU showed that early diagnosis of AD can reduce the annual patient healthcare cost by 15%. This reduction is believed to be greater than the cost of annual testing with these modalities. Reference [13] presents DSC-MRI, a relatively novel technique in current clinical practice. This technique utilizes gadolinium-based agents to measure CBF, cerebrovascular volume (CBV), and mean transit time [16]. DSC-MRI is a functional method, which detected reduced regional CBV in AD patients compared with healthy controls, with a reported sensitivity and specificity of approximately 90% [17].
There is also a non-invasive MRI technique, arterial spin labeling MRI (ASL-MRI). ASL-MRI captures changes in tissue perfusion of the whole brain or specific regions of interest using arterial water molecules as an endogenous tracer [18]. ASL can be easily added to conventional brain MRI, requiring 5 min of additional scan time [19]. Preliminary data indicate that ASL-MRI can be used as a predictor of cognitive impairment, as it can detect vascular changes that structural MRI overlooks. During the last decade, significant efforts to standardise ASL-MRI have been made, although the recommended implementation is not universally applied [20,21]. Studies have shown an overlap of hypometabolism patterns acquired via FDG-PET and hypoperfusion through ASL-MRI between AD patients and differentiation of AD from cognitively normal individuals. Although ASL-MRI has been shown to have a lower sensitivity compared with SPECT and FDG-PET, ASL-MRI is still considered a promising alternative due to its advantages [22][23][24]. Another aspect that should be taken into consideration is that MRI is a more accessible modality in comparison with both PET and SPECT. According to the official stats of Eurostat, availability of MRI ranged among EU countries from 0.5 to 3.4 scanners per 100,000 of population. The same measure for G-cameras-including SPECT-was 0.3-1.8. PET scanners ranged from 0 in Lichtenstein to a maximum of 0.8 per 100,000 of population in Denmark [25].
Perfusion patterns in the precuneus and posterior cingulate gyrus have shown a straightforward correlation to cognitive decline [26,27]. Medial temporal lobe (MTL) is another region of research interest that has already been studied via structural MRI measuring its volume. Atrophy of the MTL is an established biomarker that should not be used in isolation for early diagnosis of AD [28]. CBF in the MTL was also studied and has shown possible correlation between it and the evolution of AD [29,30]. Due to the limited number of publications and their complexity, the potential of perfusion in MTL for assessing AD progression could not be adequately assessed by reviews published in the previous decade [27]. However, recently there has been an increase in the number of studies aiming to evaluate MTL perfusion as a potential biomarker in AD. Many of these studies used ASL-MRI due to its aforementioned advantages. We believe that a systematic analysis of these newly emerging data would provide a comprehensive evaluation of the relationship between MTL perfusion and AD.
This systematic review focuses on the use of ASL-MRI as a potential vascular biomarker (V). Our aim is to evaluate the available data of CBF changes in the medial temporal lobe measured via resting state ASL-MRI in adults with cognitive decline. To accomplish this, we compared perfusion patterns in participants with AD with cognitively normal elderly adults (CN) and MCI to CN.

Materials and Methods
We followed the PRISMA 2020 statement to conduct this systematic review [31]. The research sources were PubMed, Embase/Ovid, and Cochrane library. The databases were searched from inception to 14 February 2021. The search strategy consisted of the following terms: (Cognit* OR Alzheimer* OR Memor*) AND (cerebral blood flow OR CBF OR arterial spin labeling OR ASL OR perfusion) AND (magnetic resonance imaging OR MRI) No limitations were added. The full search strategy is presented in Appendix A.
We included all studies that fulfilled the following criteria: 1.
Observational studies (cohort, case-control, or cross-sectional) or baseline results of interventional studies 2.
Human studies written in English 3.
Studies that recruited cognitively normal adults as control group 4.
Patients diagnosed at any stage of sporadic Alzheimer's disease and/or participants with mild cognitive impairment, validated with at least one cognitive test 5.
Participants who underwent arterial spin labeling magnetic resonance imaging in the resting state (rsASL-MRI).
We excluded conference abstracts, case reports, reviews, and letters to the editor. Studies were excluded if participants with cognitive decline were recruited due to organic pathology (artery occlusion, carotid artery stenosis/stiffness, stroke, intracranial artery stenosis, or CADASIL patients). Postoperative and Parkinson's disease dementia or cognitive decline in diabetes patients were also beyond our research scope. Early-onset and autosomal dominant AD studies and patients diagnosed with other types of dementia were also excluded.
When more than one study included participants from the same cohort/database, we used the article with the largest sample size or the article that provided the most detailed information on CBF values and cognitive assessment, after discussion between two reviewers (EMK, DCI).
We imported the full reference and the abstracts of the search results in Endnote (Version X9). After de-duplication, two reviewers (EMK, DCI) independently screened the titles and abstracts on the basis of the selection criteria. A third reviewer (VSV) provided adjudication, where necessary. The eligible and inconclusive articles were then assessed on a full-text level using a predesigned algorithm (Appendix B). The selection process is depicted in the Prisma flow diagram recording the reasons of exclusion (Figure 1 in Results section). stenosis, or CADASIL patients). Postoperative and Parkinson's disease dementia or cognitive decline in diabetes patients were also beyond our research scope. Early-onset and autosomal dominant AD studies and patients diagnosed with other types of dementia were also excluded.
When more than one study included participants from the same cohort/database, we used the article with the largest sample size or the article that provided the most detailed information on CBF values and cognitive assessment, after discussion between two reviewers (EMK, DCI).
We imported the full reference and the abstracts of the search results in Endnote (Version X9). After de-duplication, two reviewers (EMK, DCI) independently screened the titles and abstracts on the basis of the selection criteria. A third reviewer (VSV) provided adjudication, where necessary. The eligible and inconclusive articles were then assessed on a full-text level using a predesigned algorithm (Appendix B). The selection process is depicted in the Prisma flow diagram recording the reasons of exclusion ( Figure 1 in Results section). The following data were extracted from each eligible study: study details (name of the first author, year of publication, study design), participant characteristics (mean age, sample size, number of female participants, cerebrovascular risk factors, statistical adjustments on age, sex, cerebrovascular risk factors or other parameters, cognitive assessment tests, and the definition of AD or MCI used), MRI data (ASL acquisition type and protocol, MRI specifications, CBF estimation and analysis method, statistical approach, and confounding variables), outcome measurements (mean CBF values (relative or absolute) expressed in millilitres per 100 g per min in a specific region of interest (MTL) and qualitative results of CBF (hypoperfusion/hyperperfusion) in MTL). The data were recorded on predesigned Microsoft Excel (Microsoft 365 version 2202) spreadsheets by two researchers independently. The following data were extracted from each eligible study: study details (name of the first author, year of publication, study design), participant characteristics (mean age, sample size, number of female participants, cerebrovascular risk factors, statistical adjustments on age, sex, cerebrovascular risk factors or other parameters, cognitive assessment tests, and the definition of AD or MCI used), MRI data (ASL acquisition type and protocol, MRI specifications, CBF estimation and analysis method, statistical approach, and confounding variables), outcome measurements (mean CBF values (relative or absolute) expressed in millilitres per 100 g per min in a specific region of interest (MTL) and qualitative results of CBF (hypoperfusion/hyperperfusion) in MTL). The data were recorded on predesigned Microsoft Excel (Microsoft 365 version 2202) spreadsheets by two researchers independently.
The Joanna Briggs Institute critical appraisal checklist (JBI) was used to assess the risk of bias of the eligible studies [32]. This tool was the most appropriate for this systematic review, as we included not only cohort and case-control studies but also cross-sectional studies. This checklist detects bias in the design, conduct, and analysis of the studies. Most of the included studies were cross-sectional and the eight domains assessed were as follows: Questions for the cross-sectional studies are the following: Q1: Were the criteria for inclusion in the sample clearly defined? Q2: Were the study subjects and the setting described in detail?
Q3: Was the exposure measured in a valid and reliable way? Q4: Were objective, standard criteria used for measurement of the condition? Q5: Were confounding factors identified? Q6: Were strategies to deal with confounding factors stated? Q7: Were the outcomes measured in a valid and reliable way? Q8: Was appropriate statistical analysis used? The possible answers were: "Yes", "No", "Unclear", or "Not applicable". Study quality was assessed by two reviewers independently and any disagreement was resolved through discussion.
A meta-analysis comparing AD vs. CN and MCI vs. CN in relation to CBF flow was also undertaken. Heterogeneity across studies was assessed using the I2 statistic from the standard X2 test. A random-effects model was used in cases of high heterogeneity (I2 > 50%). A sensitivity analysis was also undertaken by excluding each study sequentially. The statistical threshold was set at p < 0.05.

Search Results
The search selection process is depicted in the PRISMA flow diagram in Figure 1 [31]. We initially identified 9024 references. After de-duplication, 7564 articles were screened at the title/abstract level. We excluded 5950 references because of non-human studies, irrelevant target disease, no MRI exposure, or unclear study design. We screened 1614 references for full-text eligibility. We excluded 1588 studies, as they did not meet the eligibility criteria: no ASL-MRI (n = 603), no AD spectrum (n = 928), no control group (n = 13), presenile AD patients (n = 4), or no MTL regions (n = 31). We also excluded papers due to sample overlap (n = 9). Lastly, we excluded the studies that did not provide CBF results for MTL or its subregions (n = 31).
We assessed the eligibility of several ADNI publications. However, due to high risk of overlap, we decided to include only one study that was more suitable based on the eligibility criteria and studied a large sample of participants.

Included Studies
At the end of the selection process, 26 studies were eligible. We present demographic and participant characteristics in Table 1  . Thirteen studies recruited only AD patients, ten studies recruited only MCI participants, and three studies recruited participants from both diagnostic groups. All studies had an observational design and were published after 2000. Sixteen studies were conducted from North America (the US and Canada), seven from Asia, and the remaining three from Europe. APOE4 carriers and cardiovascular risk factors were under-reported, with only three studies including both variables. The sample size of patients diagnosed with AD varied from 12 to 71 participants. The range of MCI participants was also wide, from 9 to 105 volunteers. The mean age of cognitively declined participants ranged from 65 to 84 years.
Imaging characteristics and CBF analysis are depicted in Table 2. A total of 21 out of 26 studies used 3 Tesla MRI scanners, four studies used 1.5 Tesla scanners, while one study was conducted with 4 Tesla scanners. The ASL sequences applied were pulsed ASL (PASL), continuous ASL (CASL), and pseudo-continuous ASL (PCASL) in 9, 5, and 11 studies, respectively. Quantification of CBF was conducted with a wide range of methods, with some studies using more than one approach. A voxel-wise method was applied in 10 out of 26 studies. Regions of interest (ROIs) or other regional approaches were conducted in 20 studies. More than one method was used in five studies. The most frequently used variables as confounders were age and gender (11 out of 26 studies). Partial volume correction was applied in the majority of the studies (20 out of 26).   The main findings of studies that recruited AD patients are summarised in Table 2. Four studies showed decreased perfusion in the MTL for the AD group [34,35,39,40], with two of them reaching statistical significance [34,35]. In contrast, Huang et al. 2018 found higher CBF in the MTL of AD patients without reaching statistical significance [46]. Hypoperfusion was observed in the hippocampus (7 out of 16 studies) [38,44,47,[49][50][51]. Four studies showed lower perfusion in the parahippocampal region [35,47,50,51]. Alsop et al. 2008 noted hyperperfusion in hippocampal and parahippocampal regions, which was statistically significant [35]. Lastly, perfusion was decreased in the amygdala in two studies, without reaching statistical significance when comparing AD with CN [48,52].
Moreover, we present the main findings of MCI participants compared with cognitively normal volunteers in Table 2. Two studies showed decreased perfusion in MTL [37,53], but only in one was the difference statistically significant [37]. In addition, higher CBF in the MTL subregions was observed in the Westerberg et al. and Alexopoulos et al. studies, with a significant and a non-significant difference, respectively [33,55]. Furthermore, decreased CBF in the hippocampus was noted in five studies [38,40,44,46,54]. After CBF correction, the difference was non-significant in two of the studies [43,57]. Increased CBF in the hippocampus was described in three studies [33,41,56]. Finally, one study showed a reduction [44], and two an increase of CBF in the parahippocampal region [43,55].

Critical Appraisal
As shown in Figure 2, answers noted in green are "Yes". Red circles indicate "No", and yellow was used when the provided information was "unclear". Eleven studies had a total score of 100% positive answers. Eight studies scored 87.5% and the remaining six studies scored 62.5%. Overall, the risk of bias for the majority of included studies was low. Six studies had a moderate risk of bias as they did not fulfil or provided limited information regarding confounding factors and the study's setting.
JBI risk of bias quality assessment, green circle indicating that the study addressed well the specific question, red circle that it did not address well the question, and yellow that it provided no information with regard to that question.

Summary of Main Findings and Meta-Analysis
The volume and the quality of the results on ASL-MRI for detection of MCI participants as well as for AD patients were satisfactory. The results are presented in the summary of findings table (Appendix C). Thirteen studies recruited MCI participants, and 16 studies recruited AD patients. Meta-analysis was difficult to conduct with the inclusion of all studies, as they focused on different subregions of the MTL. Thus, we performed subgroup meta-analyses in the hippocampus for AD-CN groups (Figure 3) as well as in the hippocampus and MTL for MCI-CN groups (Figures 4 and 5). We included these regions in our meta-analysis, as three or more studies provided quantified CBF results, expressed in ml/min/100 g tissue. The CBF values were extracted from text, tables, graphs, or box plots.
ipants (41). In Question 2 (Q2), we considered a high risk of bias in 11 studies (42%) because the information regarding the setting was limited [34][35][36]39,[41][42][43]45,47,49,50,54]. Confounding factors were adjusted for most of the studies (60%). Four studies (15%) did not adjust for confounding factors and were assessed as having a high risk of bias [33][34][35]39]. Another four studies (15%) provided little information regarding the same domain and had an unclear risk of bias [45,46,50,54].  JBI risk of bias quality assessment, green circle indicating that the study address well the specific question, red circle that it did not address well the question, and yell that it provided no information with regard to that question.

Summary of Main Findings and Meta-Analysis
The volume and the quality of the results on ASL-MRI for detection of MCI part pants as well as for AD patients were satisfactory. The results are presented in the su mary of findings table (Appendix C). Thirteen studies recruited MCI participants, and studies recruited AD patients. Meta-analysis was difficult to conduct with the inclus of all studies, as they focused on different subregions of the MTL. Thus, we perform subgroup meta-analyses in the hippocampus for AD-CN groups ( Figure 3) as well as the hippocampus and MTL for MCI-CN groups (Figures 4 and 5). We included these gions in our meta-analysis, as three or more studies provided quantified CBF results, pressed in ml/min/100 g tissue. The CBF values were extracted from text, tables, grap or box plots. This meta-analysis shows that there is significant difference of 2.82 mL/min/10 between AD and CN in the hippocampus, with AD showing lowed cerebral perfusi This remained statistically different even following the sensitivity analysis.  This meta-analysis shows that there is significant difference of 2.82 mL/min/100 between AD and CN in the hippocampus, with AD showing lowed cerebral perfusio This remained statistically different even following the sensitivity analysis. There is no statistical difference in cerebral perfusion between MCI and CN. This meta-analysis shows that there is significant difference of 2.82 mL/min/10 between AD and CN in the hippocampus, with AD showing lowed cerebral perfusi This remained statistically different even following the sensitivity analysis. There is no statistical difference in cerebral perfusion between MCI and CN. This meta-analysis shows that there is significant difference of 2.82 mL/min/100 g between AD and CN in the hippocampus, with AD showing lowed cerebral perfusion. This remained statistically different even following the sensitivity analysis.
There is no statistical difference in cerebral perfusion between MCI and CN. There is no statistical difference between MCI and CN in MTL perfusion.

Alzheimer's Disease and ASL
A total of 14 out of 16 studies noted a significantly lower perfusion in the medial temporal lobe of AD patients. These results are in agreement with the hypoperfusion patterns in medial temporal regions depicted in other perfusion imaging modalities, such as PET, SPECT, and DSC [24] and would suggest that hypoperfusion could partly explain the development of AD. Supporting this is a study by Musiek et al. who correlated perfusion patterns using ASL to glucose uptake using fluorodeoxyglucose positron emission tomography (FDG-PET) in AD patients [69]. In their small study, qualitative and quantitative similarities were found, indicating a close relationship between perfusion and metabolic state in different brain regions of AD patients. Adding further evidence for this, a study by Binnewijzend et al. (2016) correlated the lower CBF to later AD stages and a weaker cognitive performance [70]. The number of studies showing statistically significant lower CBF in the MTL of AD patients compared with controls, would demonstrate the important causative relation. CBF can, thus, in the future, become a helpful test in clinical practice in identifying patients at risk. This will be particularly vital if an early disease-modifying therapy becomes available for AD. Of course, CBF assessment is not limited to MRI, and other modalities, such as SPECT and PET, could aid in the earlier diagnosis of AD and, thus, potentially earlier treatment [39].
Of the 16 studies, two showed lack of hypoperfusion in the MTL of AD participants. They specifically showed increase of perfusion in the MTL [46] and the hippocampus [35] but without reaching statistical significance. This transient CBF raise could be explained as the participants recruited had a new onset of AD. Vascular compensatory mechanisms, induced by abnormal neuronal activity, may have still been in place [71]. Inflammatory molecules expression, the disruptions of the blood-brain barrier, the vascular dilation and the high metabolic demands of the atrophic regions could also explain the above [72]. These processes will be eventually exhausted, and low perfusion would be evident in every region of the demented brain.
In conclusion, ASL MRI can detect reduced MTL perfusion in patients at a later stage of AD. However, the results should be interpreted with caution in patients with a new onset of AD, where compensatory mechanisms are still actively influencing CBF measurement. Thus, ASL MRI should complement clinical judgement when managing patients with cognitive decline.

Mild Cognitive Impairment and ASL
The perfusion status of MTL in MCI participants is not as straightforward as in the sample of people with dementia. In three studies, MCI participants had a higher CBF value than the control group. These results were statistically significant in two out of the three studies (66%). In contrast, nine of the included studies presented a statistically significant higher CBF in the control group. It is interesting that in one study, hypoperfusion in the hippocampus was preserved at a significant level after corrections for vascular risk factors and adjusting for partial volume effect between cognitively declined and cognitively normal adults [44]. The wide variability of results on CBF difference between MCI and cognitively normal subjects cannot be overlooked. Although there is no clear explanation for the above, there are certain factors that may influence ASL MRI readings. Cardiovascular risk factors and APOE4 mutation seem to impact CBF readings. Specifically, by recruiting only cognitively normal participants and dividing them into high risk and low risk groups for AD, Fleisher et al. found that the cerebral blood flow in MTL was increased in the high-risk group [73]. In addition, people without dementia who are carriers of the APOE4 allele demonstrated higher CBF in subregions of MTL compared with non-carriers, increasing the possibility of AD conversion in the future [74].
The above hypothesis seems to attract significant interest from the scientific community. Aging, cardiovascular risk factors, APOE mutation, and capillary dysfunction seem to be in the spotlight of recent research. The results of this research show that microcirculation changes due to ageing, hypertension, and genetic risk factors for AD (e.g., the APOE4 allele) are responsible for remodelling of small vessels and, thus, the lowering of oxygen efficacy in the brain [75]. Initially, the brain compensates for the metabolic needs of the neural networks by increasing the CBF of the affected regions. The rise of CBF is sufficient for mild capillary dysfunction. When capillary changes become more evident, the oxygenation demand cannot be compensated for with higher perfusion rates [76]. At this stage, cognitive decline symptoms are pronounced, and CBF is decreased. Recent work confirms this hypothesis by comparing the CBF in MTL of cognitive normal adults, people with objective subtle cognitive decline, and MCI participants [77]. Objective cognitive is defined by Thomas et al. as "performance lesser than 1 standard deviation (SD) below the age, education, and sex-adjusted mean on: (1) two neuropsychological total scores in two different cognitive domains, (2) one neuropsychological total score and one process score, or (3) two process scores". The group with the objective cognitive decline showed higher CBF in hippocampus than MCI as well as higher CBF than the cognitively normal group, which is suggestive of mild capillary dysfunction.
The above theory could provide a valid explanation of the CBF variation between MCI and cognitively normal participants in different studies. In addition, high heterogeneity of the included studies could be another reason for the results, and we should always consider the characteristics of the MCI participants (demographics and stage of the disease at the recruitment) and the variety of disease presentation within and between subjects [71]. However, no matter how reasonable the explanations might be, variability in results is a factor that remains present and should be taken into consideration when interpreting ASL MRI readings in MCI patient.

Limitations
We should acknowledge the limitations of our systematic review. We focused our review on English-written studies, implementing a restriction to available included studies. Another weakness is that all but one eligible study were cross-sectional. In this type of study, the cause-outcome/causal relationship cannot be identified, as the disease and the exposure variables are studied simultaneously. The variability of the acquisition protocols and the imaging analysis performed were considerable. These differences in imaging protocols should be considered when we interpret the results of our review. Included studies assessed various comorbidities, namely HTN, stroke, heart disease, and diabetes; however, prediabetes and variable glycemic control were not assessed. A meta-analysis was also conducted. This had significant heterogeneity, but a random-effects model was utilised to account for this. Lastly, the sample size of two studies was less than 30 participants. A very small sample size could bias the outcome, as there is a possibility of our assumptions being confirmed by chance [78].

Recommendations for Future Studies
Despite the limitations, this work confidently addresses whether "CBF in the medial temporal lobe is different between cognitively normal and cognitively declined participants". This is based on the results of the included studies and without applicability concerns regarding the studied population, imaging test, or cognitive status. This is because cross-sectional studies can demonstrate differences between the studied groups, but the causal relationship remains unclear. For this reason, longitudinal studies are essential to evaluate the effectiveness of the imaging test and to assess the potential role of the CBF changes as an additional biomarker of differentiation between AD and cognitively normal adults.
The potential benefit of incorporating ASL-MRI in general practice should also be assessed. When an MRI is indicated in clinical practice, adding an ASL sequence can be beneficial and of a low additional cost. On the other hand, the plethora of choices in ASL parameters can be a drawback. Multisite comparisons are difficult and expensive to perform. It would be beneficial if the medical community could come together and, through guidelines, mutually agree on a specific protocol so that comparisons of acquired results can be performed directly. Towards this vision, the European working group "ASL in Dementia" and the equivalent USA Perfusion Study group-International Society for Magnetic Resonance in Medicine-have provided indicative acquisition parameters for ASL to standardise the method [21]. The implementation of this consensus could simplify the use of ASL in clinical practice and encourage the adoption of the method in more settings.
ASL-MRI is a perfusion technique that enables the non-invasive quantification of cerebral blood flow. Quantified results could be more objective than descriptive results when researchers define thresholds or a range of values to discriminate cognitive decline from cognitively normal adults [21]. To accomplish this, the brain atrophy and partial volume effect (PVE) should be taken into consideration. PVE correction is essential when CBF values are extracted. From the included studies, only four studies adjusted their results according to PVE [43,53,54,57]. Researchers should account for PVE as a confounder in the design of future studies [20]. PVE-corrected and PVE-uncorrected results can be extremely different, and comparisons among participants of different studies could be a challenge.

Role in Clinical Practice
Arterial spin labeling is a perfusion MRI modality that is relatively inexpensive and easy to perform. Cohort studies and diagnostic test accuracy studies are mandatory to set a common language among clinicians when they interpret ASL results. If these steps are accomplished, ASL-MRI could be considered as an additional test-part of the existing MRI acquisition protocol-to detect people at risk for AD with greater precision and hopefully at an earlier stage. This can be particularly helpful if a therapeutic intervention would become available, whereby early initiation is likely to offer improved outcomes. Authors should discuss the results and how they can be interpreted from the perspective of previous studies and of the working hypotheses. The findings and their implications should be discussed in the broadest context possible. Future research directions may also be highlighted.

Conclusions
In this comprehensive systematic review and meta-analysis, we show that there is significant difference in the cerebral blood flow of the hippocampus between AD and CN groups. There was no difference in either the cerebral perfusion in the hippocampus or MTL between MCI and CN participants. This work suggests that incorporating cerebral perfusion in routine MRI brain imaging could be beneficial for identifying patients at earlier stages of AD.

Quality concerns
Patients' characteristics were not always adequately described and confounding factors were moderately reported. Concerns regarding reproducibility of the investigated method were generally low.

Limitations
Lack of standard methodology in ASL-MRI process. Variety of regions of interest (ROIs) among studies

Conclusions
There is need for conducting longitudinal studies with a standardised methodological protocol of ASL-MRI with larger population samples.