Changes of Cerebral Perfusion and Functional Brain Network Organization in Patients with Mild Cognitive Impairment

Abstract

Disruptions of the functional brain network and cerebral blood flow (CBF) have been revealed in patients with mild cognitive impairment (MCI). However, the neurophysiological mechanism of hypoperfusion as well as the reorganization of the intrinsic whole brain network due to the neuropathology of MCI are still unclear. In this study, we aimed to investigate the changes of CBF and the whole brain network organization in MCI by using a multimodal MRI approach. Resting state ASL MRI and BOLD MRI were used to evaluate disruptions of CBF and underlying functional connectivity in 27 patients with MCI and 35 cognitive normal controls (NC). The eigenvector centrality mapping (ECM) was used to assess the whole brain network reorganization in MCI, and a seed-based ECM approach was proposed to reveal the contributions of the whole brain network on the ECM alterations. Significantly decreased perfusion in the posterior parietal cortex as well as its connectivity within the default mode network and occipital cortex were found in the MCI group compared to the NC group. The ECM analysis revealed decreased EC in the middle cingulate cortex, parahippocampal gyrus, medial frontal gyrus, and increased EC in the right calcarine sulcus, superior temporal gyrus, and supplementary motor area in the MCI group. The results of this study indicate that there are deficits in cerebral blood flow and functional connectivity in the default mode network, and that sensory-processing networks might play a compensatory role to make up for the decreased connections in MCI.

Keywords

Cerebral blood flow eigenvector centrality functional connectivity functional MRI mild cognitive impairment

INTRODUCTION

Mild cognitive impairment (MCI) is considered as the transition stage between cognitively normal aging and dementia [1]. The functional disruptions of the brain, as reflected by glucose metabolism measured by fluorodeoxyglucose positron emission tomography (FDG-PET), functional connectivity detected by blood oxygenation level dependent functional magnetic resonance imaging (BOLD-fMRI), and cerebral blood flow (CBF) measured by arterial spin labeling (ASL) MRI, have been widely reported in patients with MCI as well as Alzheimer’s disease (AD) [2 –6].

The BOLD signal is considered as a measure of changes in blood oxygenation and blood flow related to underlying neuronal activity, while the ASL technique provides a quantitative measure of cerebral blood flow without the need of contrast agent injection. The CBF measured by ASL-MRI was considered to be closely coupled with glucose metabolism measured by PET, and has been reported to show similarly disrupted patterns in MCI and AD [6, 7]. Patients with MCI have been consistently reported to show decreased perfusion in the posterior cingulate cortex/precuneus region [8, 9], which is the core component of the default mode network (DMN) identified by BOLD-fMRI. Disruptions of the DMN have also been widely revealed in MCI and AD, and were suggested to be associated with the amyloid deposition in the brain [10 –13]. However, the association between hypoperfusion/hypometabolism and brain functional network alterations in MCI is still unclear.

Only a small number of studies have investigated the influence of CBF on the BOLD fMRI activity in patients with AD or MCI. By evaluating the BOLD fMRI response and ASL perfusion response during resting state and an encoding task, Fleisher et al. [14] firstly reported the increased resting state CBF in the medial temporal lobes would influence the differences in BOLD activations in subjects with a high AD risk. Li et al. [15] evaluated the effect of donepezil treatment by using ASL and BOLD fMRI in patients with mild AD, and found that 12-week donepezil therapy could enhance the CBF around the posterior cingulate cortex (PCC) as well as its underlying functional connectivity to the parahippocampal, temporal, prefrontal, and parietal cortices. These results imply that the altered CBF would influence the BOLD activity, especially in the DMN, in patients with AD/MCI. Recently, Liang et al. [16] investigated the relationship between CBF and functional connectivity strength of the whole brain in healthy adults, and also reported a tight relationship between CBF and the functional topology of the brain in the DMN and executive control network. Investigating the changes in CBF and brain network organization in MCI by using multi-modal techniques may extend our understanding of the mechanism of the disease.

In addition, previous studies mainly focused on the DMN abnormalities of AD/MCI; the DMN disruptions should be a primary locus of AD/MCI pathology but not the only one. Beside the DMN abnormalities, widespread within-network and between-network connectivity disruptions have been demonstrated to be associated with the pathology [17 –19]. Investigating the abnormalities of the functional network across the whole brain would provide more information into the cognitive impairment in MCI. Recently, graph theory based approaches have been widely used to investigate the whole brain network. The eigenvector centrality mapping (ECM) derived from graph theory approaches provides an efficient method to characterize the intrinsic connectivity of the whole brain on a voxel-wise level [20]. As a model-free, data-driven approach, this method has been used to investigate the brain network alterations of AD [21] and early small vessel disease [22], and evaluate the effect of transcranial direct current stimulation treatment in MCI [23].

In this study, we aimed to investigate the changes in CBF as well as functional connectivity of the whole brain in MCI by using a multimodal neuroimaging approach. Firstly, the CBF abnormalities of MCI were evaluated using pseudo-continuous ASL (pCASL) MRI. Then, the underlying functional connectivity disruptions in the region with hypoperfusion were assessed using resting-state BOLD-fMRI. Furthermore, the disrupted organization of the whole brain network of MCI was investigated by using ECM on a voxel-wise level, and a seed-based ECM approach was proposed to further evaluate the influence of the whole brain network on the ECM changes in MCI.

MATERIALS AND METHODS

Participants

Participants were recruited from a subgroup of the RISK (Risk Index for Subclinical brain lesions in Hong Kong) study, which is a community study to explore the risk index for screening subclinical brain lesions [24]. All the participants were community-dwelling and had sufficient communication competency for cognitive testing, including the Cantonese version of Mini-Mental State Examination (CMMSE) [25], the Hong Kong version Montreal Cognitive Assessment (HK-MoCA) [26], Barthel Index [27], and Lawton’s Instrumental Activity of Daily Living (IADL) [28].

The inclusion criteria were: (1) aged 65–80 years; (2) no clinical dementia (CMMSE >22 for elderly with more than 2 years of education, CMMSE >20 for elderly with 1 to 2 years of education, and CMMSE >19 for elderly with no education [29]), and (3) intact daily life activities defined by a score of 20 on the 20-point Barthel Index and <2 on the IADL scale. The exclusion criteria were (1) history of clinical stroke or transient ischemic attack; (2) any history of profound sensory deficits, psychiatric, and/or other neurological disorders, and (3) presenting any contraindications to MRI scanning. Patients with MCI were diagnosed with education-adjustment HK-MoCA (the HK-MoCA score was added one point for subjects with education years ≤6) less than 22, which has been reported as an optimal cut-off score to detect cognitive impairment for elderly in Hong Kong [26]. Twenty-seven elderly were found to have MCI while thirty-five were cognitively healthy and served as the normal control group (NC) in this study. There was no significant group difference in age, gender, and educational levels between the two groups. The demographics and clinical characteristics of the two groups are presented in Table 1. The study was approved by the Joint Chinese University of Hong Kong-New Territories East Cluster Clinical Research Ethics Committee (CUHK-NTEC CREC) following the ethical standards and procedural requirements described in the Hospital Authority Guide on Research Ethics and the Standard Operating Procedure of the CUHK-NTEC CREC. Written informed consent was obtained from all of the participants.

Image acquisition

MRI images were acquired using a 3 Tesla Philips MRI scanner (Achieva TX, Philips Medical System, Best, the Netherlands) with an 8-channel SENSE head coil. A three-dimensional T1-weighted anatomical image was firstly obtained using the following sequence: repetition time (TR) = 7.5 ms, echo time (TE) = 3.5 ms, flip angle = 8°, field of view (FOV) = 250×250 mm², matrix = 240×240, 305 sagittal slices with slice thickness 0.6 mm. Functional ASL images were acquired using a pseudo-continuous ASL (pCASL) sequence [30]: TR = 4000 ms, TE = 14 ms, flip angle = 90°, FOV = 240×240 mm², matrix = 80×80, 17 slices, slice thickness = 7 mm, label duration = 1650 ms, post labeling delay = 1525 ms, background suppression inversion pules at 1680 and 2760 ms after the saturation pulse. Twenty pairs of perfusion labeled and control scans were obtained. An additional equilibrium magnetization image (M0) was acquired prior to the pCASL images collection with a long TR = 8000 ms. The resting state BOLD fMRI images were acquired with a T2-weighted gradient echo-planar imaging (EPI) sequence: TR = 2050 ms, TE = 25 ms, flip angle = 90°, FOV = 205×205 mm²,matrix = 64×64, 47 slices, and slice thickness = 3.2 mm. A total of 210 volumes of BOLD fMRI were obtained for each participant. Participants were instructed to keep their eyes open and focus on the cross of the screen during the pCASL and BOLD fMRI acquisition.

Data preprocessing

Both BOLD fMRI and ASL images were preprocessed using Statistical Parametric Mapping software (SPM8, http://www.fil.ion.ucl.ac.uk/spm/).

BOLD-fMRI preprocessing

The first ten volumes were discarded to allow for T1 equilibration and the remaining 200 functional images were corrected for timing offsets between different slices and further realigned to the first volume for rigid-body head motion correction. No subject was excluded with head movement larger than 1.5 mm of translation or 1.5° of rotation in any direction. The individual T1-weighted anatomical image was then co-registered to the mean realigned functional image and subsequently segmented into gray matter, white matter, and cerebrospinal fluid by using a unified segmentation method [31]. A study-specific anatomic template was created using the Diffeomorphic Anatomical Registration Exponentiated Lie algebra (DARTEL) toolbox [32]. The corrected images were spatially normalized to the standard MNI space by using the nonlinear normalization parameters estimated by the DARTEL toolbox. The normalized functional images were resampled to 3×3×3 mm³ and spatially smoothed with a 6 mm full-width half maximum (FWHM) Gaussian kernel.

To remove the possible effect of low-frequency drifts and high-frequency physiological noises during functional connectivity analysis, all the voxel-wise BOLD fMRI time-series were detrended and temporal band-pass filtered (0.01∼0.1 Hz) using the Resting-State Data Analysis Toolkit (REST) [33]. The nuisance signals, including six head-motion profiles, ventricle, and white matter signals, along with their first derivatives, were also regressed out from each voxel’s time series. In addition, a careful movement correction was implemented to minimize the effect of participant motion on functional connectivity estimation [34, 35]. The framewise displacement (FD) of each scan was computed based on the six realignment parameters. Scans in which the FD was larger than 0.5 mm were flagged. All the flagged scans, as well as one preceding and two succeeding time points of the flagged scan, were excluded during the functional connectivity estimation. Based on this procedure, an average of 4.99% and 6.44% of all scans was excluded for the NC and MCI group separately. There was no significant difference between the two groups in the mean FD before (p = 0.184) and after (p = 0.275) this motion correctionprocedure.

ASL preprocessing

For each participant, all the control and labeled pCASL images were realigned with the first control image for rigid-body head motion correction. No participant was excluded under a head-motion criterion of 1.5 mm and 1.5°. The averaged surround subtraction of control and labeled (ΔM) images were estimated [36]. Subsequently, both the ΔM andM0 images were co-registered to individual T1-weighted anatomical image for further CBF quantification.

CBF quantification

The quantitative CBF image was estimated based on a single compartment kinetic model [37]: $CBF = \frac{6000 λ Δ M e^{\frac{PLD}{T_{1, blood}}}}{2 α T_{1, blood} M_{0} (1 - e^{\frac{τ}{T_{1, blood}}})}$ (1) where, λ = 0.9 is the blood-brain partition coefficient, α= 0.85 is the labeling efficiency, ΔM and M₀are the image intensity of the mean control-label surround subtraction image and equilibrium magnetization image, T_1,blood= 1.65 is the longitudinal relaxation time of blood, PLD is the post label delay, and τ is the label duration.

To compensate for the partial volume effects (PVE) of the CBF maps, partial volume correction was implemented to ΔM by using a linear regression method [38] based on the gray matter and white matter maps derived from the T1-weighted anatomical image segmentation. The gray matter-corrected CBF maps were nonlinearly spatially normalized to the standard MNI space, re-sampled to 3×3×3 mm³ and spatially smoothed with a 6 mm FWHM Gaussian kernel by using DARTEL toolbox. The spatially normalized CBF images were then transformed to z-score by using the mean and standard deviation across all voxels in the brain of each individual for further statistical analysis.

Eigenvector centrality mapping analysis

To identify the aberrant of the whole-brain functional network in MCI, a voxel-wise eigenvector centrality analysis was used. ECM is an assumption-free, data-driven graph theory approach to evaluate the node centrality of the whole brain on voxel-wise level [20]. Eigenvector centrality (EC) is derived from the eigenvector decomposition, and reflects the relative influence of a node on all nodes connected to it. A node with high eigenvector centrality indicates it is connected with nodes that are highly connected. Thus, ECM could take both the direct and indirect connections into account, and characterize the entire pattern of the network [39]. In addition, unlike other centrality measures (such as degree centrality, functional connectivity strength and betweenness centrality), ECM does not depend on the threshold of the correlation matrix.

For each participant, a whole-brain network was generated on voxel level. Each voxel was defined as a node in the graph, and the Pearson correlations between the time series of any pairs of voxels were estimated. Only the positive correlations were retained. Correlations with Euclidean distance less than 10 mm between two voxels were also set to zero due to the short-distance correlations may associate with slicing, motion, or smoothing artifacts [40]. This analysis was confined within a group-specific mask (N_voxels = 41 486) which was created by multiplying a gray matter mask (with threshold of 0.2 for the mean gray matter probability map of all 62 subjects) with the automated anatomical labeling (AAL) atlas [41]. The eigenvector centrality was estimated by using the Brain Connectivity Toolbox (BCT, https://sites.google.com/site/bctnet/) [42]. Finally, the z-score transformation was performed to each ECM for further between-group comparison.

Seed-based functional connectivity analysis

In order to evaluate the functional connectivity disruptions in the region with hypoperfusion, the region with significantly decreased CBF values in the MCI group was selected as the seed for the seed-based functional connectivity analysis. A region of interest (ROI) with a 6-mm radium sphere centered on the maximal peak voxel of the significant cluster was defined, and the Pearson correlation of the average time course of the ROI and the other voxels in the whole brain were estimated.

The ECM clusters identified by the between-group comparison reflect the differences in centrality of a brain region, but could not reveal which regions contribute to these centrality changes. In order to evaluate the contributions of the whole brain network on the identified ECM clusters, a seed-based ECM approach was proposed. Eigenvector centrality of node i is defined as the i-th entry of the eigenvector V corresponding to the largest eigenvalueλ for the adjacency matrix A, then AV = λV, which is equivalent to $V = \frac{1}{λ} AV$ , and $v_{i} = \frac{1}{λ} \sum_{j = 1}^{N} a_{ij} v_{j}$ . The eigenvector centrality of a node is equal to the sum of all the connections between this node and nodes on the whole brain weighted by their eigenvector centrality and the inverse of the largest eigenvalue. Thus, the seed-based ECM (sECM) is defined as the Pearson correlation of the average time course of the seed and the other voxels in the whole brain, weighted by the inverse of the largest eigenvalue multiplied by the eigenvector centrality of the voxels in the wholebrain, i. e. ${sECM}_{j} = \frac{1}{λ} a_{j} v_{j}$ (2) where, a_j is the Pearson correlation between seed and voxel j, v_j indicates the eigenvector centrality of voxel j, and λ is the largest eigenvalue.

The seed-based ECM reflects the contributions of the voxels in the whole brain to the eigenvector centrality of the seed. It would take into account not only the direct connection between the seed and voxel j but also the indirect connections between the seed and other voxels via voxel j. Six 6-mm radium sphere ROIs centered on the coordinates of the maximal peak voxel of the significant clusters with ECM differences were selected for seed-based ECM estimations.

Statistical analysis

The between-group comparisons were assessed by using an independent two sample t-test with gender, age, and education years as covariates. Regarding the seed-based functional connectivity analysis, the between group comparisons were restricted in the areas with positive correlations. One sample t-test on the functional connectivity or seed-based eigenvector centrality maps across the groups was used to identify regions with significant positive correlations. Multiple comparisons correction was implemented by using the Monte Carlo simulations to achieve a cluster-level false-positive rate of 0.05 (p < 0.01 uncorrected, 10000 iterations). With Regard to the post-hoc seed-based ECM analysis, the voxel-wise threshold was set to p < 0.01/6 = 0.0017 to correct for the number of seeds based on the Bonferroni correction.

RESULTS

CBF analysis

The group averaged CBF maps of NC and MCI are shown in Fig. 1. Between-group comparison showed significantly decreased CBF in the surrounding regions of the PCC, i.e., precuneus/middle cingulate cortex (MCC), in the MCI group compared to the NC group (p < 0.01, cluster size >810 mm³) (Fig. 2A). No significantly increased CBF was found in MCI. In order to evaluate the functional connectivity changes in the region with hypoperfusion, the functional connectivity maps with a 6-mm radium sphere centered on the maximal peak voxel of the identified CBF cluster (x = –3, y = –45, z = 33) were estimated by using BOLD fMRI (Supplementary Figure 1). The MCI group showed significantly decreased connectivity in the left superior medial frontal gyrus, right inferior frontal gyrus, anterior cingulate cortex, superior precuneus, bilateral parahippocampus, bilateral middle/superior temporal lobe, and occipital cortex (p < 0.01, cluster size >702 mm³) from the region with hypoperfusion, i.e., the precuneus/MCC (Fig. 2B and Supplementary Table 1). No significantly increased connectivity was found.

ECM differences

Figure 3 shows the group averaged ECM of the whole brain for the NC and MCI groups separately. Consistent with previous studies, high EC values were mainly distributed in the occipital cortex and precuneus. The distributions of ECM showed similar patterns for the NC and MCI groups. Between-group analysis revealed significantly decreased EC values in the MCI group compared to the NC group in the MCC, left parahippocampal gyrus, and superior medial frontal gyrus. Increased EC values in the right calcarine sulcus, right superior temporal gyrus, and supplementary motor area (SMA) were also found in the MCI group (p < 0.01, cluster size >810 mm³) (Fig. 4 and Supplementary Table 2).

Seed-based ECM analysis

In order to identify brain regions contributing to the altered EC values in MCI, we assessed the seed-based ECM for the identified significantly altered clusters. Significant seed-based EC alterations were found in MCI (Fig. 5 and Supplementary Table 3). The left parahippocampus showed significantly decreased connectivity with the MCC, medial frontal gyrus, left middle temporal gyrus, and occipital lobe in the MCI group compared to the NC group (p < 0.0017, cluster size >729 mm³, see Fig. 5A). The MCC showed significantly decreased connectivity with the medial frontal gyrus and precuneus in the MCI group (p < 0.0017, cluster size >729 mm³, see Fig. 5B). The medial frontal gyrus showed decreased connectivity with the MCC, precuneus, left parahippocampa gyrus, and left middle frontal gyrus in the MCI group compared to the NC group (p < 0.0017, cluster size >675 mm³, see Fig. 5C). In addition, the MCI group also showed significantly increased connectivity between the right calcarine sulcus and the SMA, right middle frontal gyrus, and right precentral gyrus (p <0.0017, cluster size >756 mm³, see Fig. 5D). The right superior temporal gyrus showed significantly increased connectivity with SMA, MCC, precuneus, bilateral postcentral gyrus, and right middle temporal gyrus, as well as middle/superior frontal gyrus in the MCI group compared to the NC group (p < 0.0017, cluster size > 756 mm³, see Fig. 5E). The SMA showed significantly increased connectivity with the right superior/middle temporal gyrus, right pre/postcentral gyrus, and calcarine in the MCI group compared to the NC group (p <0.0017, cluster size > 729 mm³, seeFig. 5F).

DISCUSSION

In this study, we investigated the CBF abnormalities of the brain in patients with MCI by using pCASL imaging, and evaluated the changes of underlying functional connectivity for the region with hypoperfusion as well as the whole brain network organization by using BOLD fMRI. The MCI patients showed significantly reduced perfusion around the PCC, as well as decreased connectivity between the hypoperfused region and regions in the default mode network, i.e., medial frontal gyrus, ACC, superior precuneus, parahippocampus, and superior/middle temporal lobe, and the occipital cortex. These alterations in MCI may also contribute to the altered organization of the whole brain network assessed by using eigenvector centrality mapping.

Hypoperfusion in MCI

Significantly reduced perfusion in the precuneus/MCC has been consistently reported in patients with MCI and AD, as well as preclinical AD in previous studies [3 , 9]. Recently, Xekardaki et al. [2] found the reduced perfusion in the PCC in both MCI patients and healthy elderly with deteriorated cognitive function at 18-month follow up, indicating that the PCC hypoperfusion may occur in the very early stage of dementia. In addition, a recent study investigated the association of amyloid disruptions with cerebral perfusion also reported that the amyloid in the brain may cause more loss of CBF in the early stages of AD [43]. Thus, it is quite likely that the precuneus/MCC hypoperfusion in MCI might be due to the functional and synaptic loss caused by the amyloid disruptions.

Underlying functional connectivity disruptions in the region with hypoperfusion

The identified hypoperfusion region is considered as a core component of the DMN, which has been widely reported to show functional connectivity alterations in MCI, AD, and preclinical AD [5 , 45]. In order to reveal the underlying functional connectivity disruptions in the region with hypoperfusion, the functional connectivity between this region and the whole brain were evaluated. Seed-based functional connectivity analysis revealed that, compared to the healthy controls, region with hypoperfusion showed significantly decreased connectivity with the left medial frontal gyrus, ACC, superior precuneus, bilateral parahippocampus, bilateral middle/superior temporal lobe, and occipital cortex in the MCI group. A recent study has reported that for the patients with mild AD, CBF in the MCC and PCC was found to be increased after 12-week donepezil treatment, and CBF increased regions also showed improved connections with the left parahippocampal gyrus, ventral medial prefrontal cortex, ventral anterior cingulate cortex, and inferior parietal cortex [15]. As the regions around the PCC are vulnerable to amyloid deposition in AD/MCI, the connectivity of these regions were reported to be decreased with other regions spread across the medial frontal region, hippocampus/parahipocampus, superior precuneus, and middle temporal gyrus, as well as the visual cortex in AD/MCI patients [11 , 46–48]. In line with these studies, the results in the present study indicated that the region with hypoperfusion in MCI showed underlying functional connectivity deficits. The BOLD signal is considered to measure the loss of oxygen from hemoglobin in the brain. The hypoperfusion would alter the local deoxyhemoglobin concentration in the brain, which would further lead to the changes of BOLD signal intensity as well as the functional connectivity of the whole brain [49]. However, no significant increased connectivity was found in current study, which may be caused by the variation of ROI selection. In this study, region with reduced perfusion in the MCC/precuneus was selected as the seed, while other studies selected the ROIs based on anatomical atlas or previous studies. The ROI selected based on hypoperfusion in the current study should reflect the underlying functional connectivity disruptions in the hypoperfused region directly. The decreased connectivity found in this study illustrates the functional connectivity deficits of brain in the region with hypoperfusion.

ECM differences

The MCI group showed significantly decreased EC values in the MCC, medial frontal gyrus, and left parahippocampus. Intriguingly, these regions with decreased EC values were located around the regions that showed decreased connectivity with the hypoperfusion region or around the hypoperfusion region. It is possible that the hypoperfusion and the decreased connectivity in DMN may contribute to the ECM alterations in MCI. A further seed-based ECM analysis revealed the contributions of the whole brain network on the identified eigenvector centrality disruptions. The MCC region showed decreased connections with the medial frontal gyrus and superior precuneus in the MCI group compared to the NC group. Moreover, both the left parahippocampal gyrus and medial frontal gyrus also showed the decreased connections with the MCC in the MCI group. These results may imply that the regions around MCC play an important role in both cerebral perfusion and functional brain network reorganization in MCI.

As a core component that is vulnerable to the reduced metabolism and perfusion, the regions around PCC have been widely reported to show reduced centrality in subjects with amyloid deposition, MCI, and AD [4, 50]. Schaefer et al. [22] investigated the ECM of patients with cerebral small vessel disease, which is suggested to contribute to cognitive impairment and dementia, and also found a similarly decreased eigenvector centrality in the MCC and medial frontal cortex [51]. Using resting-state magnetoencephalography, significantly decreased EC in the parietal hub areas was also found in higher frequency bands in AD patients [52]. Recently, Binnewijzen et al. [21] used the ECM to evaluate the brain network disruptions of AD, and reported that compared to controls, no significantly decreased parietal EC was found in AD by using a voxel-wise comparison, but the mean EC in the parietal was significantly decreased in AD [53]. They explained it might be due to the influence of signal preprocessing. In their study, no head motion parameters or other nuisance signals was removed before ECM estimation. In our study, in order to eliminate the influence of physiological noise, signals of six head-motion profiles, ventricle, and white matter signals, along with their first derivatives, were regressed out from each voxel’s time series before ECM estimation. A careful movement correction approach (FD) was also implemented to minimize the effect of participant motion, since the participant motion may increase the short-distance connectivity while decrease the long-distance connectivity [34]. Moreover, the 1.5 Tesla magnetic fields they used to acquire the fMRI data may also affect the sensitivity for detecting the functional connectivity aberrations [54]. A previous study has reported that fMRI images acquired under 1.5 Tesla may conduct false negative results [55]. Our images acquired under 3.0 T magnetic fields could provide a higher spatial resolution and sensitivity to detect centralityabnormalities.

Decreased EC values over the left parahippocampus and medial frontal cortex in the MCI group may be attributed to the disconnections between these regions and MCC/PCC. The parahippocampal gyrus is considered as the primary hub of the DMN in the medial temporal lobe and plays a mediation role to the functional connectivity between hippocampus and PCC [56]. The disrupted extensive anatomical connections between PCC and hippocampus/parahippocampus have also been demonstrated in AD and MCI in previous studies using diffusion tensor imaging [57 –59]. The decreased medial frontal EC was mainly attributed to the reduced connections with posterior regions, i.e. MCC, superior precuneus, and parahippocampus, which gives further evidence for the impairment of long-distance connections in MCI/AD [50 , 60–62].

In contrast, the EC values in the right calcarine sulcus, right superior temporal gyrus, and SMA were significantly higher in the MCI group than the NC group. Seed-based ECM analysis demonstrated that both the right calcarine sulcus and right superior temporal gyrus showed increased connections with regions around the SMA extending to the middle frontal gryus in the MCI group compared to the NC group. It is intriguing that these regions are involved in the sensory (i.e., visual, auditory, and movement) processing. Sheline et al. [11] speculated that the impairments of DMN might result in the dysregulation of the sensory processing control, since one role of the DMN is considered to be vigilance and modulation of attention for sensory-processing [63]. The increased centrality in the sensory-processing regions may reflect a compensatory effect. These identified increased EC clusters were also reported to play a compensatory role in previous task-related fMRI studies. The calcarine was identified as a key component of the posterior compensatory network in cognitively proficient elderly with hippocampal atrophy during a spatial working memory task [64]. The right superior medial temporal was reported to play a compensatory role during a recognition memory task [65]. The supplementary motor area showed a crucial role in the long-term musical memory preservation in AD [66]. In fact, previous studies have demonstrated these regions, especially the SMA and calcarine sulcus, showed little hypometabolism and were the last to degenerate during AD pathology [67 –69]. It is possible that additional connections (i.e., occipital-frontal and temporal-frontal connections) in these intact or relatively intact regions might be established to cope with the decreased connections in MCI. It reflects a compensatory effect of the brain in the incipient stage of MCI pathology.

Limitations

Several limitations of this study should be noted. First, even though the results in current study demonstrated the changes of cerebral perfusion and functional brain network organization in MCI, we could not clarify the causality between CBF loss and functional brain network reorganization. Long-term longitudinal studies are needed to reveal whether the functional brain network reorganization is a consequence of the CBF loss. It is worth noted that one study reported the CBF but not BOLD alterations in the medial temporal lobes in MCI patients and healthy elderly with AD risk [70]. It may imply that the CBF disruptions occur prior to the BOLD disruptions. Moreover, in this study, the ASL and BOLD fMRI images were acquired separately. Simultaneous ASL-BOLD fMRI image acquisition would be informative to reveal the direct association between perfusion and intrinsic functional brain network reorganization in MCI. In addition, it has been demonstrated that the amyloid deposition and ApoE genotype are related with the functional brain network reorganization as well as CBF abnormalities. No amyloid or ApoE genotype is considered in this study. It would be interesting to reveal the relationships between intrinsic functional network reorganization and amyloid, ApoE, as well as CBF in MCI. Last but not least, as MCI is a heterogeneous clinical condition with multiple etiological causes, further studies are required to explore the generalizability of our findings in various MCI subtypes.

CONCLUSION

In conclusion, reduced perfusion around the PCC and decreased functional connectivity between the hypoperfusion region and the whole brain were identified in the patients with MCI. Further whole brain network analysis using eigenvector centrality mapping revealed the decreased eigenvector centrality in the default mode network and increased centrality in the sensory-processing related networks in the MCI group compared to the NC group. The results demonstrated the deficits of cerebral blood flow as well as functional brain network disruptions in the DMN in MCI. In addition, the sensory-processing networks might play a compensatory role in MCI, and additional occipital/temporal -frontal connections might be established to make up for the decreased connections in the DMN.

Footnotes

ACKNOWLEDGMENTS

This project was partially supported by grants from the Research Grants Council of the Hong Kong Special Administrative Region, China (Project No.: CUHK 14113214, CUHK 416712, and CUHK471911), a grant from the National Natural Science Foundation of China (Project No. 81271653), and a grant from The Science, Technology and Innovation Commission of Shenzhen Municipality (Project No.: CXZZ20140606164105361).

Authors’ disclosures available online ().

Appendix

References

Petersen

, Roberts

, Knopman

, Boeve

, Geda

, Ivnik

, Smith

, Jack

(2009) Mild cognitive impairment ten years later. Arch Neurol 66, 1447–1455.

Xekardaki

, Rodriguez

, Montandon

, Toma

, Tombeur

, Herrmann

, Zekry

, Lovblad

, Barkhof

, Giannakopoulos

, Haller

(2015) Arterial spin labeling may contribute to the prediction of cognitive deterioration in healthy elderly individuals. Radiology 274, 490–499.

, Antuono

, Jones

, Xu

, Wu

, Ward

, Li

(2007) Perfusion fMRI detects deficits in regional CBF during memory-encoding tasks in MCI subjects. Neurology 69, 1650–1656.

Drzezga

, Becker

, Van Dijk

KRA

, Sreenivasan

, Talukdar

, Sullivan

, Schultz

, Sepulcre

, Putcha

, Greve

, Johnson

, Sperling

(2011) Neuronal dysfunction and disconnection of cortical hubs in non-demented subjects with elevated amyloid burden. Brain 134, 1635–1646.

Sheline

, Raichle

(2013) Resting state functional connectivity in preclinical Alzheimer’s disease. Biol Psychiatry 74, 340–347.

Musiek

, Chen

, Korczykowski

, Saboury

, Martinez

, Reddin

, Alavi

, Kimberg

, Wolk

, Julin

, Newberg

, Arnold

, Detre

(2012) Direct comparison of fluorodeoxyglucose positron emission tomography and arterial spin labeling magnetic resonance imaging in Alzheimer’s disease. Alzheimers Dement 8, 51–59.

, Rowley

, Wu

, Alsop

, Shankaranarayanan

, Dowling

, Christian

, Oakes

, Johnson

(2010) Reliability and precision of pseudo-continuous arterial spin labeling perfusion MRI on 3.0 T and comparison with O-15-water PET in elderly subjects at risk for Alzheimer’s disease. NMR Biomed 23, 286–293.

Dai

, Lopez

, Carmichael

, Becker

, Kuller

, Gach

(2009) Mild cognitive impairment and Alzheimer disease: Patterns of altered cerebral blood flow at MR imaging. Radiology 250, 856–866.

Binnewijzend

, Kuijer

, Benedictus

, van der Flier

, Wink

, Wattjes

, van Berckel

, Scheltens

, Barkhof

(2013) Cerebral blood flow measured with 3D pseudocontinuous arterial spin-labeling MR imaging in Alzheimer disease and mild cognitive impairment: A marker for disease severity. Radiology 267, 221–230.

10.

Sperling

, LaViolette

, O’Keefe

, O’Brien

, Rentz

, Pihlajamaki

, Marshall

, Hyman

, Selkoe

, Hedden

, Buckner

, Becker

, Johnson

(2009) Amyloid deposition is associated with impaired default network function in older persons without dementia. Neuron 63, 178–188.

11.

Sheline

, Raichle

, Snyder

, Morris

, Head

, Wang

, Mintun

(2010) Amyloid plaques disrupt resting state default mode network connectivity in cognitively normal elderly. Biol Psychiatry 67, 584–587.

12.

Hedden

, Van Dijk

, Becker

, Mehta

, Sperling

, Johnson

, Buckner

(2009) Disruption of functional connectivity in clinically normal older adults harboring amyloid burden. J Neurosci 29, 12686–12694.

13.

Mormino

, Smiljic

, Hayenga

, Onami

, Greicius

, Rabinovici

, Janabi

, Baker

, Yen

, Madison

, Miller

, Jagust

(2011) Relationships between beta-amyloid and functional connectivity in different components of the default mode network in aging. Cereb Cortex 21, 2399–2407.

14.

Fleisher

, Podraza

, Bangen

, Taylor

, Sherzai

, Sidhar

, Liu

, Dale

, Buxton

(2009) Cerebral perfusion and oxygenation differences in Alzheimer’s disease risk. Neurobiol Aging 30, 1737–1748.

15.

, Antuono

, Xie

, Chen

, Jones

, Ward

, Franczak

, Goveas

, Li

(2012) Changes in regional cerebral blood flow and functional connectivity in the cholinergic pathway associated with cognitive performance in subjects with mild Alzheimer’s disease after 12-week donepezil treatment. Neuroimage 60, 1083–1091.

16.

Liang

, Zou

, He

, Yang

(2013) Coupling of functional connectivity and regional cerebral blood flow reveals a physiological basis for network hubs of the human brain. Proc Natl Acad Sci U S A 110, 1929–1934.

17.

Elman

, Madison

, Baker

, Vogel

, Marks

, Crowley

, O’Neil

, Jagust

(2014) Effects of beta-amyloid on resting state functional connectivity within and between networks reflect known patterns of regional vulnerability. Cereb Cortex 26, 695–707.

18.

Brier

, Thomas

, Snyder

, Benzinger

, Zhang

, Raichle

, Holtzman

, Morris

, Ances

(2012) Loss of intranetwork and internetwork resting state functional connections with Alzheimer’s disease progression. J Neurosci 32, 8890–8899.

19.

Lou

, Shi

, Wang

, Tam

, Chu

, Mok

, Cheng

, Lam

(2015) Decreased activity with increased background network efficiency in amnestic MCI during a visuospatial working memory task. Hum Brain Mapp 36, 3387–3403.

20.

Lohmann

, Margulies

, Horstmann

, Pleger

, Lepsien

, Goldhahn

, Schloegl

, Stumvoll

, Villringer

, Turner

(2010) Eigenvector centrality mapping for analyzing connectivity patterns in fMRI data of the human brain. PLoS One 5, e10232.

21.

Binnewijzend

, Adriaanse

, Van der Flier

, Teunissen

, de Munck

, Stam

, Scheltens

, van Berckel

, Barkhof

, Wink

(2014) Brain network alterations in Alzheimer’s disease measured by eigenvector centrality in fMRI are related to cognition and CSF biomarkers. Hum Brain Mapp 35, 2383–2393.

22.

Schaefer

, Quinque

, Kipping

, Arelin

, Roggenhofer

, Frisch

, Villringer

, Muellerl

, Schroeter

(2014) Early small vessel disease affects frontoparietal and cerebellar hubs in close correlation with clinical symptoms–a resting-state fMRI study. J Cereb Blood Flow Metab 34, 1091–1095.

23.

Meinzer

, Lindenberg

, Phan

, Ulm

, Volk

, Flöel

(2015) Transcranial direct current stimulation in mild cognitive impairment: Behavioral effects and neural mechanisms. Alzheimers Dement 11, 1032–1040.

24.

Wong

, Law

, Liu

, Wang

, Lo

, Lau

, Wong

, Mok

(2015) Montreal cognitive assessment: One cutoff never fits all. Stroke 46, 3547–3550.

25.

Chiu

, Lee

, Chung

, Kwong

(1994) Reliability and validity of the Cantonese version of the Mini-Mental State Examination: A preliminary study. Hong Kong J Psychiatr 4, 25–28.

26.

Wong

, Xiong

, Kwan

, Chan

, Lam

, Wang

, Chu

, Nyenhuis

, Nasreddine

, Wong

, Mok

(2009) The validity, reliability and clinical utility of the Hong Kong Montreal Cognitive Assessment (HK-MoCA) in patients with cerebral small vessel disease. Dement Geriatr Cogn Disord 28, 81–87.

27.

Mahoney

, Barthel

(1965) Functional evaluation: The Barthel Index. Md State Med J 14, 61–65.

28.

Lawton

, Brody

(1969) Assessment of older people: Self-maintaining and instrumental activities of daily living. Gerontologist 9, 179–186.

29.

Lam

, Tam

, Leung

, Lui

, Fung

, Chiu

, Chan

, Ng

, Chan

(2010) Combined clinical and cognitive criteria to identify mild cognitive impairment in a southern Chinese community. Alzheimer Dis Assoc Disord 24, 343–347.

30.

Dai

, Garcia

, de Bazelaire

, Alsop

(2008) Continuous flow-driven inversion for arterial spin labeling using pulsed radio frequency and gradient fields. Magn Reson Med 60, 1488–1497.

31.

Ashburner

, Friston

(2005) Unified segmentation. Neuroimage 26, 839–851.

32.

Ashburner

(2007) A fast diffeomorphic image registration algorithm. Neuroimage 38, 95–113.

33.

Song

, Dong

, Long

, Li

, Zuo

, Zhu

, He

, Yan

, Zang

(2011) REST: A toolkit for resting-state functional magnetic resonance imaging data processing. PLoS One 6, e25031.

34.

Power

, Barnes

, Snyder

, Schlaggar

, Petersen

(2012) Spurious but systematic correlations in functional connectivity MRI networks arise from subject motion. Neuroimage 59, 2142–2154.

35.

Cheng

, Rolls

, Gu

, Zhang

, Feng

(2015) Autism: Reduced connectivity between cortical areas involved in face expression, theory of mind, and the sense of self. Brain 138, 1382–1393.

36.

Liu

, Wong

(2005) A signal processing model for arterial spin labeling functional MRI. Neuroimage 24, 207–215.

37.

Alsop

, Detre

, Golay

, Gunther

, Hendrikse

, Hernandez-Garcia

, Lu

, MacIntosh

, Parkes

, Smits

, van Osch

, Wang

, Wong

, Zaharchuk

(2015) Recommended implementation of arterial spin-labeled perfusion MRI for clinical applications: A consensus of the ISMRM perfusion study group and the European consortium for ASL in dementia. Magn Reson Med 73, 102–116.

38.

Asllani

, Borogovac

, Brown

(2008) Regression algorithm correcting for partial volume effects in arterial spin labeling MRI. Magn Reson Med 60, 1362–1371.

39.

Geerligs

, Saliasi

, Renken

, Maurits

, Lorist

(2014) Flexible connectivity in the aging brain revealed by task modulations. Hum Brain Mapp 35, 3788–3804.

40.

Power

, Schlaggar

, Lessov-Schlaggar

, Petersen

(2013) Evidence for hubs in human functional brain networks. Neuron 79, 798–813.

41.

Tzourio-Mazoyer

, Landeau

, Papathanassiou

, Crivello

, Etard

, Delcroix

, Mazoyer

, Joliot

(2002) Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. Neuroimage 15, 273–289.

42.

Rubinov

, Sporns

(2010) Complex network measures of brain connectivity: Uses and interpretations. Neuroimage 52, 1059–1069.

43.

Mattsson

, Tosun

, Insel

, Simonson

, Jack

Jr , Beckett

, Donohue

, Jagust

, Schuff

, Weiner

, Alzheimer’s Disease Neuroimaging Initiative (2014) Association of brain amyloid-beta with cerebral perfusion and structure in Alzheimer’s disease and mild cognitive impairment. Brain 137, 1550–1561.

44.

Greicius

, Srivastava

, Reiss

, Menon

(2004) Default-mode network activity distinguishes Alzheimer’s disease from healthy aging: Evidence from functional MRI. Proc Natl Acad Sci U S A 101, 4637–4642.

45.

, Wu

, Wang

, Zhang

, Dong

, Yao

, Li

(2010) Impairment and compensation coexist in amnestic MCI default mode network. Neuroimage 50, 48–55.

46.

Bai

, Watson

, Yu

, Shi

, Yuan

, Zhang

(2009) Abnormal resting-state functional connectivity of posterior cingulate cortex in amnestic type mild cognitive impairment. Brain Res 1302, 167–174.

47.

Zhang

, Wang

, Xing

, Liu

, Ma

, Yang

, Zhang

, Teng

(2009) Detection of PCC functional connectivity characteristics in resting-state fMRI in mild Alzheimer’s disease. Behav Brain Res 197, 103–108.

48.

Dennis

, Thompson

(2014) Functional brain connectivity using fMRI in aging and Alzheimer’s disease. Neuropsychol Rev 24, 49–62.

49.

Montagne

, Nation

, Pa

, Sweeney

, Toga

, Zlokovic

(2016) Brain imaging of neurovascular dysfunction in Alzheimer’s disease. Acta Neuropathol 131, 687–707.

50.

Dai

, Yan

, Li

, Wang

, Cao

, Lin

, Shu

, Xia

, Bi

, He

(2015) Identifying and mapping connectivity patterns of brain network hubs in Alzheimer’s disease. Cereb Cortex 25, 3723–3742.

51.

Snyder

, Corriveau

, Craft

, Faber

, Greenberg

, Knopman

, Lamb

, Montine

, Nedergaard

, Schaffer

, Schneider

, Wellington

, Wilcock

, Zipfel

, Zlokovic

, Bain

, Bosetti

, Galis

, Koroshetz

, Carrillo

(2015) Vascular contributions to cognitive impairment and dementia including Alzheimer’s disease. Alzheimers Dement 11, 710–717.

52.

de Haan

, van der Flier

, Wang

, Van Mieghem

, Scheltens

, Stam

(2012) Disruption of functional brain networks in Alzheimer’s disease: What can we learn from graph spectral analysis of resting-state magnetoencephalography? Brain Connect 2, 45–55.

53.

Adriaanse

, Wink

, Tijms

, Ossenkoppele

, Verfaillie

, Lammertsma

, Boellaard

, Scheltens

, van Berckel

, Barkhof

(2015) The association of glucose metabolism and eigenvector centrality in Alzheimer’s disease. Brain Connect 6, 1–8.

54.

Kruger

, Kastrup

, Glover

(2001) Neuroimaging at 1.5 T and 3.0 T: Comparison of oxygenation-sensitive magnetic resonance imaging. Magn Reson Med 45, 595–604.

55.

Skouras

, Gray

, Critchley

, Koelsch

(2014) Superficial amygdala and hippocampal activity during affective music listening observed at 3 T but not 1.5 T fMRI. Neuroimage 101, 364–369.

56.

Ward

, Schultz

, Huijbers

, Van Dijk

KRA

, Hedden

, Sperling

(2014) The parahippocampal gyrus links the default-mode cortical network with the medial temporal lobe memory system. Hum Brain Mapp 35, 1061–1073.

57.

Zhang

, Schuff

, Jahng

, Bayne

, Mori

, Schad

, Mueller

, Du

, Kramer

, Yaffe

, Chui

, Jagust

, Miller

, Weiner

(2007) Diffusion tensor imaging of cingulum fibers in mild cognitive impairment and Alzheimer disease. Neurology 68, 13–19.

58.

Zhou

, Dougherty

Jr , Hubner

, Bai

, Cannon

, Hutson

(2008) Abnormal connectivity in the posterior cingulate and hippocampus in early Alzheimer’s disease and mild cognitive impairment. Alzheimers Dement 4, 265–270.

59.

Ito

, Sasaki

, Takahashi

, Uwano

, Yamashita

, Higuchi

, Goodwin

, Harada

, Kudo

, Terayama

(2015) Detection of early changes in the parahippocampal and posterior cingulum bundles during mild cognitive impairment by using high-resolution multi-parametric diffusion tensor imaging. Psychiatry Res 231, 346–352.

60.

Liu

, Yu

, Zhang

, Liu

, Duan

, Alexander-Bloch

, Liu

, Jiang

, Bullmore

(2014) Impaired long distance functional connectivity and weighted network architecture in Alzheimer’s disease. Cereb Cortex 24, 1422–1435.

61.

Wang

, Li

, Liang

, Zhang

, Li

, Shu

, Wang

, Zhang

(2013) Amnestic mild cognitive impairment: Topological reorganization of the default-mode network. Radiology 268, 501–514.

62.

Wang

, Zuo

, Dai

, Xia

, Zhao

, Jia

, Han

, He

(2013) Disrupted functional brain connectome in individuals at risk for Alzheimer’s disease. Biol Psychiatry 73, 472–481.

63.

Buckner

, Andrews-Hanna

, Schacter

(2008) The brain’s default network: Anatomy, function, and relevance to disease. Ann N Y Acad Sci 1124, 1–38.

64.

Valenzuela

, Turner

, Kochan

, Wen

, Suo

, Hallock

, McIntosh

, Sachdev

, Breakspear

(2015) Posterior compensatory network in cognitively intact elders with hippocampal atrophy. Hippocampus 25, 581–593.

65.

Lenzi

, Serra

, Perri

, Pantano

, Lenzi

, Paulesu

, Caltagirone

, Bozzali

, Macaluso

(2011) Single domain amnestic MCI: A multiple cognitive domains fMRI investigation. Neurobiol Aging 32, 1542–1557.

66.

Jacobsen

, Stelzer

, Fritz

, Chetelat

, La Joie

, Turner

(2015) Why musical memory can be preserved in advanced Alzheimer’s disease. Brain 138, 2438–2450.

67.

Forster

, Grimmer

, Miederer

, Henriksen

, Yousefi

, Graner

, Wester

, Forstl

, Kurz

, Dickerson

, Bartenstein

, Drzezga

(2012) Regional expansion of hypometabolism in Alzheimer’s disease follows amyloid deposition with temporal delay. Biol Psychiatry 71, 792–797.

68.

Jack

Jr , Holtzman

(2013) Biomarker modeling of Alzheimer’s disease. Neuron 80, 1347–1358.

69.

Benzinger

, Blazey

, Jack

Jr , Koeppe

, Su

, Xiong

, Raichle

, Snyder

, Ances

, Bateman

, Cairns

, Fagan

, Goate

, Marcus

, Aisen

, Christensen

, Ercole

, Hornbeck

, Farrar

, Aldea

, Jasielec

, Owen

, Xie

, Mayeux

, Brickman

, McDade

, Klunk

, Mathis

, Ringman

, Thompson

, Ghetti

, Saykin

, Sperling

, Johnson

, Salloway

, Correia

, Schofield

, Masters

, Rowe

, Villemagne

, Martins

, Ourselin

, Rossor

, Fox

, Cash

, Weiner

, Holtzman

, Buckles

, Moulder

, Morris

(2013) Regional variability of imaging biomarkers in autosomal dominant Alzheimer’s disease. Proc Natl Acad Sci U S A 110, E4502–E4509.

70.

Bangen

, Restom

, Liu

, Wierenga

, Jak

, Salmon

, Bondi

(2012) Assessment of Alzheimer’s disease risk with functional magnetic resonance imaging: An arterial spin labeling study. J Alzheimers Dis 31, S59–S74.