Mild Cognitive Impairment and Alzheimer Disease: Patterns of Altered Cerebral Blood Flow at MR Imaging

Abstract

Purpose: To examine regional cerebral blood flow (rCBF) in incident mild cognitive impairment (MCI) and Alzheimer disease (AD) by using continuous arterial spin-labeling (CASL) magnetic resonance (MR) imaging.

Materials and Methods: This study was approved by the local institutional review board and was compliant with HIPAA regulations. Informed consent was obtained. rCBF was measured in 38 control subjects, 29 MCI patients, and 37 AD patients who were participating in a longitudinal epidemiologic study. Multisection CASL MR imaging with alternating single and double adiabatic inversion pulses and ramp-sampled echo-planar imaging were performed to acquire 19 contiguous axial sections. Voxel-level rCBF was compared among groups by using an analysis of variance design; clusters of voxels with significant group differences were identified. Multiple regression models controlled for age, sex, and presence of hypertension and related the mean rCBF in those clusters to the presence of MCI and AD.

Results: MCI and AD patients had decreased rCBF in the posterior cingulate gyrus (P = .01) with extension to the medial precuneus compared with that in control subjects. MCI patients had increased rCBF in the left hippocampus (P < .001), right amygdala (P = .007), and rostral head of the right caudate nucleus and ventral putamen and globus pallidus (P = .003) compared with that in control subjects. AD patients had decreased rCBF relative to that in control subjects and MCI patients in the left inferior parietal (P = .005), left lateral frontal (P < .001), left superior temporal (P = .001), and left orbitofrontal (P = .003) cortices. AD patients had increased rCBF in the right anterior cingulate gyrus (P = .02) compared with that in control subjects.

Conclusion: The transition from normal cognition to AD is associated with dynamic pathologic processes in the brain, and this is reflected by both decreases and increases in rCBF. Increases in rCBF suggest a cellular and vascular compensatory process associated with incipient AD.

© RSNA, 2009

Alzheimer disease (AD) is a neurodegenerative disorder that leads to cerebral structural changes and alterations in neurotransmitter systems, especially in the cortical association areas and the limbic system (1–5). These structural changes become evident as patients with AD go through the transition from normal cognition to dementia, a clinical phase referred to as mild cognitive impairment (MCI) (6–11). Similar observations are made at a functional level; there is decreased regional cerebral blood flow (rCBF) or glucose metabolism in specific brain regions of patients with AD (12–15) and patients with MCI (16–19).

All of these findings, however, are complicated by the fact that in both normal aging and in the transition to dementia, there is a range of factors that modify an individual's risk to show any degree of cognitive dysfunction. Cognitive reserve or compensatory processes have been invoked as possible explanations for some of these risk modifiers (20–22), but there is no true consensus as to the critical factors that influence the progression to dementia. What is clear, however, is that this is a dynamic process. The brain is not simply “dying” but is undergoing a substantial change as the functions of specific regions are modified or lost and the brain must adapt to these progressive injuries (23–25).

The bulk of the research in these fields has come from well-executed studies derived from samples in referral clinics. These studies have a variety of advantages in that the samples are well characterized and usually represent individuals with few substantial comorbid medical conditions. However, a disadvantage of the studies is the fact that individuals drawn from referral clinics must have had clinical symptoms severe enough to warrant an evaluation. The patient, family, or treating physicians must have been aware of some degree of change. Patients classified as having MCI who typically are recruited from referral clinics thus frequently have the earliest manifestation of a full-blown dementia syndrome. Consequently, it is not surprising to find that these individuals have alterations in both the structure and function of brain regions typically seen in more severe clinical stages.

The purpose of this study was to evaluate rCBF in participants drawn from a community-based study of dementia—the Cardiovascular Health Study Cognition Study (26). We specifically report rCBF changes in patients who developed AD during the course of observation in the Cardiovascular Health Study Cognition Study, as well as in patients who developed MCI during that period. We acquired whole-brain maps of rCBF by using continuous arterial spin-labeling (CASL) perfusion magnetic resonance (MR) imaging and detected voxel-level and region-level differences in rCBF abnormalities by using region-based statistical analyses. CASL MR imaging is a noninvasive technique that requires neither injections nor ionizing radiation (27–30), and it has been successfully used to measure rCBF in patients with AD (31,32).

MATERIALS AND METHODS

Cardiovascular Health Study Cognition Study

CASL MR imaging was performed in Cardiovascular Health Study Cognition Study participants who were assessed between 2002 and 2003, and these participants formed the basis of the present study (33) (Fig 1). MR imaging of the brain was performed 1–2 weeks after clinical evaluation. The neurologic and neuropsychologic evaluations were similar in 1998–1999 and 2002–2003 (26). The characteristics of the Cardiovascular Health Study cohort and the Pittsburgh Cardiovascular Health Study Cognition Study cohort have been described previously (26,33). A description of the clinical examination is shown in Appendix E1 (http://radiology.rsnajnls.org/cgi/content/full/2503080751/DC1). The clinical and radiologic components of the study were approved by the local institutional review board of the University of Pittsburgh, and it was compliant with Health Insurance Portability and Accountability Act regulations.

Figure 1:

Chart of MR imaging of the brain in participants of the Cardiovascular Health Study Cognition Study. All participants underwent MR imaging of the brain between 1992 and 1994 and between 1998 and 1999. A third MR imaging examination was performed in 150 participants between 2002 and 2003. * = See Participant section for exclusion details.

Participants

We examined the CASL data in 104 (69%) of 150 participants who underwent MR imaging of the brain between 2002 and 2003 (Fig 1). The remaining 46 participants were excluded from analysis because of the following reasons: evidence of structural central nervous system lesions (eg, brain infarcts, central nervous system neoplasms, prior brain surgery) (two in the MCI group, one in the dementia group), history of strokes or head trauma encephalopathy (two in the MCI group, one in the dementia group), patient had other type of dementia besides AD (three in the dementia group), consumption of caffeine within 8 hours prior to examination (one in the normal group), inability to segment structural MR images by using semiautomated tools (four in the normal group, two in the MCI group, two in the dementia group), improper placement of the CASL MR imaging labeling plane in a position that was not orthogonal to both carotid arteries (four in the normal group, two in the MCI group, three in the dementia group), a greater than 20% difference between left and right carotid arterial mean velocities (four in the normal group, three in the MCI group, one in the dementia group), excessive structural MR imaging artifacts caused by patient motion (four in the MCI group, four in the dementia group), or excessive image artifact (caused by, eg, hair oil or dental implant) (one in the normal group, one in the MCI group).

MCI and Dementia Criteria

MCI was diagnosed according to the Cardiovascular Health Study Cognition Study diagnostic criteria for probable MCI (26). The diagnosis of dementia was made by the Adjudication Committee (O.L.L., L.H.K., J.T.B.) by using all available data from each participant (33). All participants with dementia included in this study met criteria for probable AD (34).

CASL MR Imaging

All MR imaging data were acquired with a 1.5-T system (Signa, LX version; GE Healthcare; Milwaukee, Wis) at the University of Pittsburgh Medical Center MR Research Center after each participant provided informed consent (either directly or by their caregiver) and passed MR imaging safety screening. The CASL acquisition, cerebral blood flow (CBF) calculation, and image analysis used in the Cardiovascular Health Study Cognition Study MR imaging study (W.D., H.M.G., and O.T.C.) were summarized previously for a study of cognitively healthy control subjects (35). Additional information regarding the rCBF calculations is presented in Appendix E1 (http://radiology.rsnajnls.org/cgi/content/full/2503080751/DC1).

Imaging protocol.—Coronal T1-weighted three-dimensional spoiled gradient-recalled echo (SPGR) images covering the whole brain were acquired for gray and white matter segmentation and image registration and normalization (field of view, 24 × 18 cm; matrix, 256 × 192; echo time, minimum full k-space [5 msec]; repetition time, 25 msec; flip angle, 40°; section thickness, 1.5 mm; spacing, 0 mm; sections, 124; receiver bandwidth, 16 kHz; number of signals acquired, one; acquisition time, 7 minutes 44 seconds).

A coronal two-dimensional slab MR angiogram was acquired to visualize the carotid and vertebral arteries and prescribe the location of the transverse spin-labeling plane (see Appendix E1 for details, http://radiology.rsnajnls.org/cgi/content/full/2503080751/DC1). The spin-labeling plane was prescribed by using the MR angiogram to ensure that the plane was transverse to both internal carotid arteries for efficient labeling. The labeling plane was typically placed near the cervicomedullary junction, about 10 cm below the center of the imaging volume and 5.5 cm below the inferior edge of the imaging volume. The velocity of the arterial flow through the labeling plane was measured by using fast two-dimensional phase contrast cine MR imaging.

Multisection CASL MR imaging with alternating single and double adiabatic inversion pulses (3.7-second pulse train at 92% duty cycle) and ramp-sampled echo-planar imaging were performed to acquire 19 contiguous axial sections (36). The postlabeling delay was 700 msec prior to multisection image acquisition to permit us to image the entire cerebrum during the multisection acquisition without the presence of substantial intraluminal artifact, as verified with statistical analysis of each image set. Images were acquired sequentially from superior to inferior to avoid radiofrequency perturbation of the endogenous tracer as it moves superiorly into the brain and to minimize signal intensity discontinuities associated with interleaved acquisitions. The adiabatic inversion pulse sequence was repeated 50 times for signal averaging of the pairs of label and control acquisitions.

Corrections of relaxation-related effects (eg, off-resonance saturation) were required to allow direct comparisons of different image sections because each section was acquired at a different time after the radiofrequency label or control irradiation (Figure E1, http://radiology.rsnajnls.org/cgi/content/full/2503080751/DC1). Tissue T1 was measured by using saturation recovery echo-planar imaging with off-resonance saturation and without off-resonance saturation.

A partial volume correction was applied for each voxel by dividing its CBF value by the fraction of its gray matter content on the probabilistic segmentation maps. The correction assumes that the white matter does not substantially contribute to the perfusion signal of the voxel at 1.5 T within our experimental conditions (Figure E2, http://radiology.rsnajnls.org/cgi/content/full/2503080751/DC1).

Statistics.—The SPGR acquisition in each participant was aligned to the standard colin27 reference brain by using a fully deformable registration method that is superior for the intersubject registration to the method provided by using statistical parametric mapping software (Statistical Parametric Mapping 2 [SPM2]; Wellcome Department of Imaging Neuroscience, London, England) (37,38). An echo-planar mean image for each participant was registered to the corresponding SPGR image by using SPM2; the registration parameters were used to transform the subject rCBF maps into the space of the SPGR image. The parameters of the fully deformable registration method that transformed the SPGR image to the colin27 brain then were used to transform the rCBF maps from the space of the SPGR image to the space of the colin27 brain. In this way, each voxel of rCBF data from each participant was transformed to coincide with gray matter voxels in the standard colin27 space.

Because the brain volume coverage may have varied between individual participant rCBF maps, voxels in the standard colin27 space had a variable number of transformed participant rCBF maps that coincided with them. Therefore, the statistical analysis included only those gray matter voxels that had perfusion data from a majority of participants in each group (ie, at least 19 of 38 control subjects, 15 of 29 patients with MCI, and 19 of 37 patients with AD). We refer to the set of analyzed voxels as the brain volume mask. The rCBF maps were smoothed by using a 6-mm Gaussian kernel prior to subsequent analyses.

The smoothed rCBF maps were masked so that only voxels in the brain volume mask were tested for group differences. At each brain volume mask voxel, the effect of clinical group on rCBF at that voxel was tested by using one-way analysis of variance written in MATLAB (MathWorks, Natick, Mass). A two-tailed critical P value of .02 was chosen for this voxel-level analysis to identify significant differences between groups. A cluster-level P value was obtained for each cluster by calling the appropriate SPM2 subroutine. This cluster-level correction was performed to guard against false-positive findings caused by the large number of voxel-level comparisons performed. The correction took into account the voxel-level threshold and the size and shape of the cluster (39). SPM2 subroutines were used to display the clusters with a corrected cluster level of P < .05 onto sections of the colin27 brain.

Cluster-level post hoc analyses were performed to determine whether any differences detected at the cluster-level analysis were still significant after controlling for potential confounders. For each participant, the mean rCBF in each region of interest was calculated; these per-participant mean region-of-interest rCBF values were the dependent variables in subsequent statistical models. Because of the variability in CASL section coverage described above, some participant regions of interest contained fewer than 100 voxels of coinciding rCBF data. Mean rCBF values for these participant regions of interest were discarded from the analysis to guard against unreliable measurement of region-level CBF.

By using a multivariate linear regression model, we examined the independent contributions of age, sex, education, and race and the presence of apolipoprotein ɛ4 allele, heart disease, diabetes, and hypertension to mean per-participant rCBF in each region of interest. We found in bivariate models that age, presence of hypertension, and sex had the most significant effects on mean region-of-interest rCBF, and thus when we compared the normal, MCI, and AD groups, we controlled for these factors by using three-way analysis of covariance. Analysis of variance and χ² tests were used to compare demographic and clinical characteristics among the normal, MCI, and dementia groups (J.T.B. and O.L.L.).

RESULTS

The demographic characteristics of the participants are shown in Table 1. Participants with probable AD had lower Modified Mini-Mental State Examination (40) scores than patients with MCI and healthy control subjects; otherwise, the participant groups were similar to one another.

Table 1.

Demographic and Clinical Characteristics of Participants

Note.—Unless otherwise indicated, data are numbers of participants, with percentages in parentheses.

^* Analysis of variance and χ² tests.

^† Data are means ± standard deviations, with ranges in parentheses.

^‡ Data are means ± standard deviations.

^§ Control and MCI groups different from AD group (P < .001).

^‖ Told by physician.

^# According to American Diabetic Association.

^** History of myocardial infarction, angina, or congestive heart failure.

Per-participant global mean CBF, computed for each participant by averaging CBF over all brain volume mask voxels for that participant, differed significantly among the three groups (F_2,101 = 5.09, P = .008). Post hoc analyses showed differences between normal and AD (t = 2.23, P = .03) and between MCI and AD (t = −3.16, P = .002) groups. However, there were no significant differences between normal and MCI groups (t = 1.06, P = .29). Statistically significant group differences according to voxel-level analysis are shown on orthogonal sections of the colin27 brain (Figs 2–4). With regard to patients with MCI versus healthy control subjects (Fig 2), patients with MCI had decreased rCBF in the posterior cingulate gyrus with extension to the medial precuneus and increased rCBF in the left hippocampus, right caudate nucleus and putamen and globus pallidus (striatum), and right amygdala compared with that in healthy control subjects. With regard to the AD versus MCI groups (Fig 3), the AD group had further decreases in rCBF in the left superior parietal, lateral frontal, superior temporal (anterior), and orbitofrontal cortices, as well as in the left hippocampus, left thalamus, and right amygdala compared with that in the MCI group. With regard to the AD group versus control group (Fig 4), patients with AD had decreased rCBF in the posterior cingulate gyrus with extension to the medial precuneus, along with the inferior parietal cortex, left lateral frontal cortex, left superior temporal cortex, and left orbitofrontal cortex. Patients with AD had increased rCBF in the right anterior cingulate gyrus. Cluster-level statistics for the significant clusters are shown in Table 2.

Figure 2:

A–D, Images show statistically significant clusters (with a corrected cluster level of P < .05) according to analysis of variance overlaid in color on the surface section of the colin27 brain. Yellow indicates the most significant difference in rCBF. Images show sagittal, coronal, and axial planes from left to right.

Figure 3:

A–F, Images show statistically significant clusters (with a corrected cluster level of P < .05) according to analysis of variance overlaid in color on the surface section of the colin27 brain. Yellow indicates the most significant difference in rCBF. Images show sagittal, coronal, and axial planes from left to right.

Figure 4:

A–E, Images show statistically significant clusters (with a corrected cluster level of P < .05) according to analysis of variance overlaid in color on the surface section of the colin27 brain. Yellow indicates the most significant difference in rCBF. Images show sagittal, coronal, and axial planes from left to right.

Table 2.

Summary of Cluster-Level Statistics for Decreased and Increased rCBF in Participants

The rCBF for the cluster-level post hoc analysis was expressed in milliliters per 100 g per minute of tissue, and the analysis was controlled for age, sex, and presence of hypertension (Table 3). There were statistically significant differences in the rCBF among groups in the lateral frontal, orbitofrontal, superior temporal, and inferior and superior parietal cortices and medial precuneus, as well as in the anterior and posterior cingulate gyrus, putamen and globus pallidus, amygdala, and hippocampus. The z scores of patients with MCI and patients with AD relative to those of control subjects are shown in Figure 5. There were significant effects of age on rCBF in the left orbitofrontal cortex (P = .01), of sex on rCBF in the left hippocampus (P = .02) and right orbitofrontal cortex (P = .01), and of the presence of hypertension on rCBF in the left hippocampus (P = .02) and left inferior frontal cortex (P = .01).

Table 3.

Region-based Analysis of Mean rCBF in Brain Regions for Normal, MCI, and AD Groups Adjusted for Age, Sex, and Presence of Hypertension

Note.—Data are mean rCBF values ± standard deviations expressed as milliliters per 100 g per minute.

^* MCI and AD groups different from normal group.

^† Normal group different from patients with early AD.

^‡ MCI group different from normal group.

^§ MCI group different from normal and AD groups.

^‖ AD group different from MCI and normal groups.

Figure 5:

rCBF z scores in patients with MCI and early AD relative to control subjects. L = left, R = right, ACG = anterior cingulate gyrus, Amyg = amygdala, BG = basal ganglia (ventral caudate nucleus, globus pallidus, and putamen), Hipp = hippocampus, IP = inferior parietal cortex, LF = lateral frontal cortex, OF = orbitofrontal cortex, PCG = posterior cingulate gyrus with extension to the precuneus; SP = superior parietal cortex, ST = superior temporal cortex, Thal = thalamus.

DISCUSSION

The transition from normal cognition to clinical dementia reflects dynamic pathologic processes in the brain, and this is expressed as both decreases and increases in rCBF. Decreases in blood flow (and by implication, decreases in brain function) were observed both subcortically and cortically in patients with early AD. We found increases in rCBF in portions of the limbic system, which suggests some form of compensatory mechanism during the early expression of the clinical syndrome, especially during the MCI stage.

One strength of this study was that all the participants were part of a longitudinal community-based cohort study rather than a specialty clinic, and we had detailed clinical and radiologic data for as many as 13 years prior to CASL MR imaging, including neuropsychologic testing. None of the patients with MCI or patients with AD had comorbid conditions that could, in and of themselves, have caused the alterations in cognition. However, most important, from our perspective, none of the patients had a physician diagnose the dementia, and none had sought an evaluation for changes in cognitive function. This means that these patients—both MCI and AD—were in the earliest phase of the clinical syndrome, and thus our data provide unique information about the first changes in rCBF that occur in the transition from normal cognition to dementia.

rCBF is generally considered to be a reasonable proxy measure for brain functional activity (41). As the activity of neurons changes, so does the demand for oxygen, and it is the changing demand for energy that moderates the rCBF (42). Thus, brain atrophy is generally accompanied by decreases in rCBF, and in the case of AD and related dementias, this is seen as decreases in flow (at single photon emission computed tomography) or metabolism (at positron emission tomography [PET]) in the temporoparietal regions in patients with AD (12,13,15) and patients with MCI (16–18).

However, reduced rCBF can coexist with increased rCBF in the early stages of the neurodegenerative process (43). Specifically, rCBF is elevated in patients with AD in the medial temporal lobe (including the hippocampus), which suggests a compensatory increase in neural activity (43). A recent study (44) conducted in a referral clinic cohort with perfusion-weighted MR imaging showed increased rCBF in the medial temporal lobe, amygdala, and anterior cingulate gyrus in patients with MCI. Mildly impaired patients with AD tend to have normal functional connectivity of memory circuits (mesial temporal lobe), although they have suprathreshold activity in the dorsolateral prefrontal and parietal cortices as the complexity of memory tasks increases (20,21). Thus, the data from the present study, in the context of prior work, are fully consistent with the model that the transition from normal cognition to dementia is marked by a stage during which the brain appears to be attempting to compensate for the structural and functional lesions that occur as a result of the neurodegenerative process.

Brain structural changes caused by the neurodegenerative process are not necessarily synchronous with functional change, and a certain degree of synaptic activity may remain near normal levels, even in regions experiencing substantial loss of neurons and amyloid deposition. While some investigators found an association between amyloid-binding ligands and atrophy, they did not necessarily find correlation between amyloid deposits and regional glucose metabolism (45). Similarly, acetylcholinesterase activity was reduced in all areas of the neocortex in moderately impaired patients with AD, while atrophy and abnormal fluorine 18 fluorodeoxyglucose metabolism were restricted to the cortical association areas (46). Finally, we also found an asymmetry in the rCBF abnormalities, which had been found previously in fluorine 18 fluorodeoxyglucose PET studies in MCI and early AD (19) and in amyloid deposition (47), which may be a manifestation of the early neurodegenerative process.

Although the focus of many studies has been on the structure and function of the temporal regions, there is increasing evidence that the basal ganglia are abnormal very early in the degenerative process of AD. Studies (48,49) have shown that amyloid deposition occurs not only in the frontal and parietal cortical areas, but also in the basal ganglia, especially in the striatum of patients with mild AD. We found that our patients with MCI had increased rCBF in the ventral regions of the putamen, globus pallidus, and the medial part of the head of the caudate nucleus. These areas are part of what is known as the rostral ventral striatum (50,51), and together with the ventral pallidum (striatopallidal system), form a complex cortical-subcortical system that modulates behavior and cognitive functioning. Therefore, the increased rCBF is occurring in a specific brain circuit that plays an essential role in the modulation of behavior and cognition (52). This does not mean that the neurons of ventral striatum are intact; indeed, our findings suggest the opposite. The increased rCBF is a response to pathologic damage (53). However, it is possible that the absence of the ability to mount this response may result in a faster progression from normal cognition to AD or in a patient manifesting more severe symptoms of dementia.

The main limitations of the study involved the low intrinsic perfusion signal-to-noise ratio of CASL compared with that of exogenous contrast material (gadolinium-based) methods, the assumptions made in the perfusion quantification (eg, tracer arrival time), the short tracer life that precludes accurate measurements of white matter perfusion at 1.5 T, the uncertainty and reproducibility of the perfusion technique (approximately 10%–15%), and the variability of perfusion within and between individuals that limits the utility of the technique in being able to help diagnose disease and changes in individuals. Some of the variability may be related to the conditions of imaging, alertness and physiologic state of the research participant, and medications and diet (including caffeine and nicotine) that may affect CBF. The perfusion signal-to-noise ratio and trac life can be increased by using magnetic field strengths greater than 1.5 T at the risk of increased radiofrequency specific absorption rate and by using a faster image acquisition sequence (eg, spiral acquisitions and/or parallel imaging) to image the brain during the peak tracer signal. Finally, a correlation between rCBF and neuropsychologic tests, which could further our understanding of the rCBF in the transition from normalcy to dementia, will be conducted in a separate study.

The benefits of the CASL method are the superior accuracy and reproducibility of CASL compared with those of exogenous perfusion methods, the simplicity and safety of CASL imaging (no gadolinium-based contrast material is used), and the ease of quantification compared with that of exogenous contrast material–based kinetics. The CASL technique can provide greater perfusion signal-to-noise ratio, larger brain coverage per acquisition, and better quantification than the pulsed arterial spin-labeling technique (54).

The increased and decreased rCBF implies that central nervous system metabolism is highly variable during the transition from normal cognition to AD through the MCI state. Our findings, as well as those from others, support the hypothesis that there is an early compensatory cellular mechanism (coupled with a vascular mechanism) associated with the AD pathologic process that appears to be more pronounced in the MCI state (ie, prior to some threshold of neuropathology beyond which such compensation is not possible).

ADVANCE IN KNOWLEDGE

• Results of continuous arterial spin-labeling (CASL) MR imaging performed in a carefully characterized group of patients with mild cognitive impairment and early Alzheimer disease suggest changes in brain blood flow within specific brain regions, which may help characterize the disease process.

IMPLICATION FOR PATIENT CARE

• CASL MR imaging is a noninvasive technique that requires neither injections nor ionizing radiation; thus, it may provide a more generally available method for depicting alterations in cerebral blood flow.

Abbreviations

• AD = Alzheimer disease

• CASL = continuous arterial spin labeling

• CBF = cerebral blood flow

• MCI = mild cognitive impairment

• rCBF = regional CBF

• SPGR = spoiled gradient-recalled echo