Correlation of regional cerebral blood ow with brain structural changes in mild cognitive impairment and Alzheimer's disease

Brain atrophy and structural changes due to aging contribute strongly in the pathophysiology of mild cognitive impairment (MCI) and Alzheimer’s Disease (AD). Regional cerebral blood ow (CBF) impairment is believed to be one of the initial changes in the AD continuum. In this study, we investigated the association between CBF and brain structural changes associated with aging and neurodegeneration. Methods Data from three groups of participants including 39 control normal (CN), 82 MCI, and 28 AD subjects were downloaded from the Alzheimer’s disease Neuroimaging Initiative (ADNI). Magnetic resonance images (MRI) of participants were automatically segmented by FreeSurfer V 7.0 software and arterial spin labeling (ASL) MRI was applied to measure CBF and investigate effect of aging and structural changes on CBF in experimental groups. One way ANOVA and Pearson correlation coecient were used to compare data and nd correlation between CBF and structural changes in the brain. Results AD patients signicantly lower Mini-Mental State Examination (MMSE) score (p=0.001) and more APOE-ε4 carriers (p=0.001) as compared to the MCI and CN groups. Our ndings revealed a wide spread signicant correlation between the CBF and structural changes, including cortical volume, subcortical volume, surface area, and in all participants, particularly AD patients after adjusting for age, sex, and genotyping status.


Introduction
Alzheimer's disease (AD) is a type of neurodegenerative disease characterized by learning and memory impairment, personality changes, and executive dysfunction 1 . Although mental status examination and neuropsychological tests have been applied to assess the symptoms of mild cognitive impairment (MCI) and AD patients, neuroimaging methods such as magnetic resonance imaging (MRI) are used to identify structural changes in the AD brain 2 . However, these modalities are not adequate to diagnose AD because of the overlaps with brain changes in the normal aging. Neuroimaging ndings have revealed that brain atrophy and structural changes due to aging contribute strongly in the pathophysiology of AD and MCI 3,4 and can support clinicians to diagnose AD 5 . Neuroimaging studies have shown that the rate of wholebrain atrophy is a valuable marker to predict the progression of AD in patients with MCI 6 . In addition, AD patients typically show atrophy in the medial temporal lobe 7 , entorhinal cortex 8 , and posterior cingulate cortex 9 .
Benjanin et al., reported that structural changes are associated with hypometabolism in the large posterior neocortical regions of the brain in AD patients 10 . Moreover, brain atrophy has been reported in the vast majority of neurological diseases. Previous studies indicated a signi cant correlation between the pattern of brain atrophy with Parkinson's disease 11,12 and amyotrophic lateral sclerosis (ALS).
Therefore, brain atrophy in MRI can be a reliable index to discriminate AD from other neurodegenerative diseases, particularly in the early onset of the disease. It has been well-documented that amyloid-beta (Aβ) and tau protein are involved in the pathophysiology of AD and responsible for neural degeneration and changes in the brain structure 13 .
Cerebral small vessel disease, which is common in AD, has been reported as a pivotal cause of cognitive decline. White matter hyperintensities of presumed vascular origin (WMHs) can be observed on magnetic resonance imaging (MRI) using T2-weighted Fluid Attenuated Inversion Recovery (FLAIR) sequences.
Although WMHs are related to the brain atrophy and cognitive decline in AD, their mechanistic role is not understood 14 . A decline in the regional cerebral blood ow (CBF) is proposed to be one of the initial changes in the AD process 15 . However, the cause of this decline is not yet fully understood. Mattsson et al. reported an association between the Aβ and CBF pattern and proposed that declined CBF is an early consequence of neural death prior to considerable grey matter loss. It has been shown that brain atrophy in the AD patients is due to the accumulation of Aβ plaques, resulting in microvasculature disease and CBF impairment 16 .
With regard to the role of amyloid plaques and neuro brillary tangles in the AD brain pathology, investigating correlation between CBF and structural changes of the brain might reveal a different pathological pathway involved in the cortical and subcortical changes of the brain. Therefore, present study was undertaken to investigate the association between CBF with the structural changes in the cortical and subcortical regions of the brain in three groups of individuals, including normal control (NC), MCI, and AD subjects.

Materials And Methods
Participants and data acquisition The participants' information was collected from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database (http://www.loni.usc.edu/), launched by the Michael W. Weiner in 2003, as a public-private partnership. Primary endpoint of ADNI has been to examine whether serial MRI, positron emission tomography (PET), biological marker examinations, and clinical/neuropsychological assessments can be combined to determine the progression of MCI and early AD. All participants with available data and imaging results were enrolled in this study. Demographic and clinical data of 149 subjects including 28 AD patients, 82 MCI patients, and 39 healthy subjects were obtained from the ADNI and analyzed in this study. The diagnostic status, Mini-Mental State Examination (MMSE) score, and the result of apolipoprotein E gene (APOE) genotyping for each subject were collected, as well.
MRI acquisition (T1, FLAIR) All ADNI individuals had T1 structural MRI and T2-FLAIR acquired on a 1.5 Tesla (T) or 3T scanner from Siemens, Philips, or General Electric. The ADNIMRI protocol is described in detail by Jack et al. 17 . Accordingly, axial 3T FLAIR was acquired with voxel sizes of 0.85994 × 0.8594 × 5 mm. Each image underwent quality control after acquisition at the Mayo Clinic (Rochester, MN), including image quality assessment, protocol compliance checks, and inspection for clinically signi cant medical abnormalities.
Automatic segmentation with Freesurfer V 7.0 Volumetric segmentation was performed using FreeSurfer image analysis suite, which is freely available for download online (http://surfer.nmr.mgh.harvard.edu/). The processing consisted of motion correction of multiple T1-weighted images, removal of non-brain tissues, automated Talairach transformation, segmentation of subcortical white matter and gray matter structures, automated topology correction, and intensity normalization.
Arterial spin labeling (ASL)-MRI processing ASL-MRI, as a completely non-invasive procedure, was used to measure the CBF by exploiting the endogenous spins of arterial water as a proxy for blood ow. Also, labeling magnetization of arterial spins was inverted selectively. The Center for Imaging of Neurodegenerative Diseases (CIND) prepared perfusion-weighted images (PWI), computed a map of CBF, and conducted a regional analysis. ASL-MRI pre-processing consisted of three steps. The rst step involved motion correction that ASL images were converted from DICOM to NiFTI format. In the second step, PWI computations were done and the ASL images were separated into the two groups of tagged and untagged, and the mean of each group was computed and saved. Next, the difference in the mean of the two groups was determined to obtain the PWI. The rst untagged image was used as the reference for water density and termed "M0". M0 was used to calibrate the ASL signal for CBF computations and estimate the transformation from ASL-MRI and structural MRI, as an intermediate frame. The third step was intensity scaling of PWI, as well as the M0 image.

Statistical analysis
Before the statistical analysis, variables without a normal distribution were log-transformed to meet the normality assumption. Demographic variables were compared between the groups using the ANOVA test.
Local associations between the CBF and structural variables were investigated using correlation models.
Partial correlation models were implemented for each association separately by adding CBF and structural variables (i.e., subcortical and cortical volume, thickness, and surface area) for each region and entering age, sex, and APOE-ε4 genotyping status as controlled variables. The bootstrapping method was used for addressing type I errors due to multiple comparisons. The signi cance level was set at 0.05, and statistical analysis was performed in SPSS version 22.

Results
Participants' characteristics Table 1 presents demographic and clinical data of the participants. There was no signi cant difference in terms of age, education, and sex among the groups. However, AD patients had signi cantly lower MMSE scores and more APOE-ε4 carriers as compared to the MCI and CN groups.

Local correlation between CBF and subcortical volume
After implementing controlled correlation models, we found a signi cant local correlation between the CBF and subcortical volume in the groups ( Table 2). In the AD group, only a local correlation was found in the fourth ventricle, whereas local correlations in the MCI group were observed in the third and fourth left and right lateral ventricles, as well as the right nucleus accumbens. Indeed, signi cant correlations were observed in the right pallidum, right vessel, and white matter of the left hemisphere cerebellum in the CN group.

Local correlation between CBF and thickness
In all AD patients, negative correlations were found in two regions, including posterior segment of the left middle frontal gyrus and the caudal part of the right anterior cingulate cortex ( Table 3). The results of Pearson's correlation analysis showed a positive correlation between the CBF and thickness in several regions in the MCI group, including left entorhinal area, left and right lateral occipital cortices, left and right superior parietal lobules, right inferior parietal lobule, posterior part of the right middle frontal gyrus, right superior frontal gyrus, right inferior temporal gyrus, right pericalcarine, right postcentral gyrus, right precentral gyrus. In contrast, we found a negative correlation between right anterior cingulate cortex thickness and CBF. In healthy subjects, signi cant correlations were found between the left entorhinal area and the rostral part of the left anterior cingulate cortex with CBF.
Local correlation between CBF and cortical volume Pearson correlation coe cient revealed a signi cant correlation between the CBF and cortical volume in the MCI and CN groups bot not in the AD group (

Discussion
In present study, our ndings revealed a signi cant correlation between the CBF and structural changes, including cortical volume, subcortical volume, surface area, and thickness in all groups after adjusting for age, sex, and APOE genotyping status. To the best of our knowledge, this is the rst study investigating the correlation between the CBF and structural changes, including cortical volume, subcortical volume, and surface area in patients with cognitive impairments.

Recently, Kim et al. revealed no signi cant association between the CBF and thickness in MCI patients
and healthy people 18 . Generally, the CBF decline is one of the earliest events observed in patients with AD 19 . Evidence on CBF changes in patients with MCIs revealed that an increase or decrease in perfusion can be an early marker of neurodegeneration and may re ect metabolic demand changes in regions that are involved in cognitive function, including the temporal lobe, parietal lobe, frontal lobe, posterior cingulate gyrus, and precuneus 20,21 . Neuroimaging studies have also found a decreased cortical cerebral blood ow (by 25 %) in patients with AD. Several mechanisms for this hypoperfusion have been proposed including constriction of brain arterioles, loss of vascular density and changes in neural activity patterns and/or in neurovascular coupling 22 . Also, alterations in cortical and subcortical regions, especially in the medial temporal lobe, have been observed, which are thought to be an indirect marker of neuronal damage in the preclinical phase of AD and can be detected by MRI 23,24 . Moreover, cortical and subcortical volume, surface area, and thickness may help diagnose AD and accurately predict the cognitive decline as a result of neurodegeneration 25,26 .
Our ndings revealed that changes in the CBF can predict structural changes in various brain regions, including the cortical and subcortical areas in AD, MCI, and healthy individuals. The correlations were mostly found in the medial temporal, temporal, parietal, occipital, and frontal regions, which are thought to be sensitive areas for the early pathology of AD 27 . More involved regions were detected in our study, including the precentral gyrus, pericalcarine cortex, entorhinal cortex, supramarginal gyrus, fusiform, pallidum, and ventricles. Notably, we found a correlation between the CBF and cortical volume and surface area in the posterior cingulate cortex and temporal pole of MCI patients, which are the main regions in the default mode network (DMN) 28 . It should be mentioned that this association was mostly observed in the preclinical stage of AD. Generally, the decline in CBF leads to the brain dysfunction, structural changes in the AD pathogenesis, and even death 29 .
According to our results, the CBF may be responsible for further structural changes and atrophy in the pathogenic course of AD; however, the mechanisms of this hypoperfusion in the early stages of neurodegeneration are unclear. Evidence shows that the CBF decline can lead to Aβ and hyperphosphorylated tau accumulation 30 . On the other hand, Aß monomers have found to drive vasoconstriction in brain arterioles and potentially contributes to a reduction in resting cerebral blood ow and, therefore, CBF decline could be result of Aβ pathology 31 . Recent studies suggest that Aβ can impair the fundamental mechanisms of blood supply regulation 32,33 . However, several factors might account for this dysregulation in AD, such as impairment of endothelium-dependent responses, hypercontractile phenotype of cerebral smooth muscle cells, and vascular oxidative stress 34 . In this regard, Michels et al. observed a signi cant relationship between the CBF and APOE-ε4, independent of Aβ accumulation in MCI and normal elderly individuals 35 . Moreover, other studies found CBF alterations in the right parahippocampal gyrus, bilateral cingulate gyri, and frontal regions in APOE-ε4 carriers 36 .
We found strong correlations between the CBF and structural changes in the main brain regions involved in AD development, including the cingulate gyrus, temporal gyrus, and parietal lobule, which differ between healthy individuals and patients with AD and MCI 24,35 . Although the role of tau pathology in neurodegeneration seems to be stronger than Aβ, both are associated with early pathological changes in AD 37 . According to our hypothesis, CBF may independently predict structural changes due to AD progression. However, it is still debatable whether the altered CBF is the cause or consequence of atrophy and structural changes 17,38 .
In contrast to our ndings, Luckhaus et al. found no signi cant association between atrophy and CBF in the early pathogenesis of AD 37 . Conversely, another study reported a signi cant correlation between CBF and cortical thickness in the predementia stages 39 . Another study investigated the patterns of atrophy and hypoperfusion in MCI patients, compared to the controls, and found that both cerebral perfusion and gray matter structure reduced in the entorhinal cortex and the isthmus cingulate cortex. However, in several regions, the CBF decline and atrophy were observed independently in MCI patients 39 .

Conclusion
According to the present results, CBF decline, as measured by ASL-MRI, might predict future structural changes and causes neurodegeneration associated with AD development, regardless of Aβ or tau accumulation. However, there are some unresolved issues regarding the role of Aβ and tau pathologies in blood ow dysregulations and structural changes underlying the pathology of CBF impairment in the AD mechanisms; therefore, further research is needed in this area.