Cerebral blood flow is an earlier indicator of perfusion abnormalities than cerebral blood volume in Alzheimer's disease

Abstract

The purpose of this study was to elucidate whether cerebral blood flow (CBF) can better characterize perfusion abnormalities in predementia stages of Alzheimer's disease (AD) than cerebral blood volume (CBV) and whether cortical atrophy is more associated with decreased CBV or with decreased CBF. We compared measurements of CBV, CBF, and mean cortical thickness obtained from magnetic resonance images in a group of healthy controls, patients with mild cognitive impairment (MCI) who converted to AD after 2 years of clinical follow-up (MCI-c), and patients with mild AD. A significant decrease in perfusion was detected in the parietal lobes of the MCI-c patients with CBF parametric maps but not with CBV maps. In the MCI-c group, a negative correlation between CBF values and cortical thickness in the right parahippocampal gyrus suggests an increase in CBF that depends on cortical atrophy in predementia stages of AD. Our study also suggests that CBF deficits appear before CBV deficits in the progression of AD, as CBV abnormalities were only detected at the AD stage, whereas CBF changes were already detected in the MCI stage. These results confirm the hypothesis that CBF is a more sensitive parameter than CBV for perfusion abnormalities in MCI-c patients.

Introduction

Microvasculature abnormalities have a key role in the pathogenesis of Alzheimer's disease (AD).¹ Several neuroimaging studies have pointed to cerebral blood hypoperfusion in the parietotemporal association areas of patients with mild cognitive impairment (MCI) as a predictor of rapid conversion to AD.² However, while there is consensus on the hypoperfusion pattern of parietotemporal areas for AD, results for the medial temporal lobes are inconsistent, probably owing to differences in the spatial resolution of the imaging techniques used.² Blood perfusion abnormalities in patients with AD seem to be associated with local neuronal tissue degeneration caused by the AD neuropathology, but also with reduced neuronal activity in areas functionally connected to those atrophied regions. According to the so-called disconnection hypothesis, functional changes observed in the parietotemporal and frontal regions might be related to reduced neuronal inputs from the medial temporal lobe,³ the primary structure affected by the AD neuropathology. Although perfusion biomarkers are not validated for early diagnosis of AD,⁴ magnetic resonance imaging (MRI) of perfusion could prove to be a powerful tool in the characterization and tracking of AD and a good alternative to nuclear medicine.^{5, 6}

Magnetic resonance imaging enables blood perfusion to be assessed through cerebral blood flow (CBF) and cerebral blood volume (CBV). Cerebral blood volume is an indicator of blood vessel lumen size and density, whereas CBF is the amount of blood reaching the tissue per unit of time and thus contains blood velocity information. Therefore, CBF could be more sensitive to physiologic perfusion changes than CBV.⁷ Most perfusion MRI studies in AD only reported either CBF^{8, 9} or CBV;^{10, 11} it remains unclear which parameter of the two constitutes an earlier and more sensitive biomarker of AD. To our knowledge, the few studies that measured both variables reported a statistically significant reduction in CBF but a nonsignificant reduction in CBV in the parietotemporal and frontal cortex of AD patients.^{12, 13}

We hypothesized that CBF could be a more sensitive indicator of hypoperfusion than CBV in predementia stages of AD, when neuronal activity is already reduced but AD neuropathology has not largely spread. The purpose of this study was to elucidate whether CBF can better characterize perfusion abnormalities in predementia stages of AD than CBV and whether cortical atrophy is more associated with decreased CBV or with decreased CBF. We therefore compared CBV, CBF, and mean cortical thickness measured using MRI in a group of healthy controls, patients with MCI who converted to dementia due to AD (MCI-c), and patients with AD and mild dementia.

Materials and Methods

All studies were performed under an agreed protocol approved by the Hospital General Universitario Gregorio Marañón Ethics Committee, in accordance with the guidelines detailed in Ley 14/2007, de 3 de julio, de Investigación Biomédica (Biomedical Research Act). Written informed consent was obtained from all participants. In the case of the patients, the study neurologist informed about the study procedures to the patient and the caregiver (next of kin). The study neurologist established the capacity/ability of the patient to consent.

Subjects

Patients for this study were recruited prospectively by a senior neurologist with expertise in behavioral neurology during routine clinical practice in the behavioral neurology clinic of a teaching hospital. The study sample comprised patients with amnesic MCI (n=16) that converted to dementia due to AD and patients with mild Alzheimer's dementia (n=12). The control subjects (n=20) were chosen from among caregivers attending the behavioral neurology clinic and from among researchers' acquaintances. Demographic and cognitive data are shown in Table 1.

Table 1. Demographic and clinical data of the participants.

	Controls, n=20	MCI-c, n=16	Alzheimer, n=12
Age in years (s.d.)	71.65 (7.04)	72.94 (6.53)	77.42 (6.97)
Sex (F:M)	11:9	11:5	4:8
Years of education (s.d.)	9.1 (4.36)	6.56 (2.90)	7.08 (5.38)
MMSE (s.d.)a	27.55 (2.09)	21.75 (4.40)b	19.08 (3.15)b
CDR	0	0.5	1

CDR, Clinical Dementia Rating; MCI-c, mild cognitive impairment converters; MMSE, Mini Mental State Examination.

^a(P<0.0001) Analysis of variance of group differences.

^bSignificant differences (P<0.0001) with the control group in the post hoc Dunnett's test.

The potential existence of cognitive impairment was investigated in all the participants by means of a semi-structured interview and a physical and neurologic examination. All the participants underwent a formal battery of neuropsychological tests including the California Verbal Learning Test, the Frontal Assessment Battery, a verbal fluency test, Addenbrooke's Cognitive Examination, and the Rey-Osterrieth Complex Figure. Subjects were excluded if they presented any medical, psychiatric, or neurologic condition (except the possibility of AD) that could affect cognition. General inclusion criteria were age over 60 years and ability to read and write. The specific criteria for inclusion in each of the study groups are detailed below.

Control group

Control subjects were cognitively normal, and performed within the normal limits in the cognitive tests, according to age and education. They had a score of 0 in the Clinical Dementia Rating and a score above 24 in the Mini Mental State Examination.

Mild Cognitive Impairment converters group

This group was sub-sampled from a total of 37 patients with MCI. Mild cognitive impairment was diagnosed using the criteria of Winblad et al¹⁴ The selection criteria were as follows: (1) cognitive impairment had to be supported by an abnormal performance (1 to 1.5 s.d. below the expected performance for age and education) in one or more tests from the neuropsychological battery; (2) patients were at stage 0.5 in the Clinical Dementia Rating; and (3) after 2 years of clinical follow-up, the patients converted to dementia due to AD and fulfilled the same criteria as the mild AD group.

Mild Alzheimer's Dementia group

Probable AD was diagnosed using the criteria established by the National Institute of Neurological and Communicative Disorders and Stroke—Alzheimer Disease and Related Disorders Association.¹⁵ Hence, deterioration in memory and other cognitive functions had to be documented, and instrumental activities of daily living had to be affected. The diagnosis of mild AD was always supported by a score of 1 (i.e., mild dementia) in the Clinical Dementia Rating.

Image Acquisition

Magnetic resonance imaging data were acquired using a 1.5T system (Philips Medical Systems, Best, The Netherlands). The imaging protocol included a volumetric T1-weighted three-dimensional gradient echo, which was used for tissue segmentation (FA=30° TR=16 milliseconds, TE=4.6 milliseconds; matrix size=256 × 256; FOV=256 × 256 mm and 100 slices with slice thickness=1.5 mm).

Perfusion images were obtained using an echo-planar imaging sequence (EPI factor=61, FA=40° TR=1,439 milliseconds, TE=30 milliseconds; matrix size=128 × 128; FOV=230 × 230 mm; section thickness=5 mm) after injection of a bolus of gadolinium chelate (10 mL of gadobutrol; Gadovist Bayer-Schering AG, Berlin, Germany). Forty volumes (30 slices each) per subject were obtained during administration of the gadolinium contrast (automatic injector, 4 mL/second), which was followed by administration of 30 mL of saline solution (4 mL/second).

Image Analysis

Perfusion Image processing

Cerebral blood volume and CBF parametric maps were obtained using automatic detection of arterial input function¹⁶ plus fitting of the concentration curve to a gamma function.¹⁷

T1-weighted image processing

For each subject, the anatomic T1 image was used to estimate regional mean cortical thickness and gray matter (GM) volume with the FreeSurfer package (version 4.5.1, http://www.surfer.nmr.mgh.harvard.edu). Brain masks were estimated from skull-stripped baseline images that were generated using the VBM8 toolbox (available at: http://www.dbm.neuro.uni-jena.de/vbm) for the SPM8 package (Wellcome Trust Centre for Neuroimaging, London, UK; available at: http://www.fil.ion.ucl.ac.uk/spm). This algorithm produces a skull-stripped ‘p0' image that consists of brain tissue classified into GM, white matter, and cerebrospinal fluid. The ‘p0' images were binarized and visually inspected to manually correct brain segmentation errors. The brain mask images were introduced into the default FreeSurfer processing pipeline because they provide more accurate skull stripping.

FreeSurfer processing

The white and gray cortical surfaces were reconstructed from the raw unaligned images in native space using the methods described by Fischl and Dale.¹⁸ The reconstruction procedure was supervised and corrected when necessary by an operator who was blind to the subject's diagnosis. Mean cortical thickness and GM volume for different regions of interest were obtained by applying the Desikan–Killiany atlas.¹⁹

T1-weighted MRI was co-registered with the CBF and CBV parametric maps of perfusion images using mutual information methods (Figure 1).²⁰ Gyrus-based ROI masks were then applied to perfusion maps and an average CBF (or CBV) value per ROI was computed.

Figure 1.

Cerebral blood flow map co-registered to a T1-weighted image.

Measurements

In this study, we evaluated cortical GM volume and mean cortical thickness of the bihemispheric frontal, parietal, temporal, and occipital lobes. The individual gyral parcellations were re-labeled into the lobes by assigning a set of gyri to each lobe. See Desikan et al¹⁹ for the assignment schedule. The cingulate cortex was evaluated as anterior and posterior division and temporal lobes were also dissociated in the medial and lateral aspects according to the Desikan–Killiany atlas. Whole-brain cortical GM volume was calculated by summing brain lobe volumes and the cingulated cortex volume. This measurement was used as a covariate of mean cortical thickness values in the ROI analysis. For the mean cortical thickness measurement, we performed an ROI analysis of each gyrus independently.

We also segmented volumes for the amygdalae and hippocampi according to Fischl et al.²¹

The average for perfusion data (CBV and CBF) was calculated for each of the structural regions described above. Cerebral blood volume and CBF data were also calculated for global cortical volume to normalize the intensity of regional perfusion data.

Perfusion Image Data Normalization and Thresholding

Because of the large physiologic variability in global perfusion measurements across the subjects owing to biologic and experimental factors, data must be normalized before comparative analysis of CBV or CBF parametric maps. The most widely used normalization method computes the ratio of an ROI value to the average CBV or CBF of all voxels within a reference region. In this study, we normalized by whole-brain cortical GM. In CBV maps, pixels with signal intensity higher than the 98th percentile of the brain values were excluded from all regional analyses, as they probably included blood vessels.^{5, 10}

Statistical Analysis

Differences in demographic and clinical data were tested using an analysis of variance (ANOVA) model.

Group differences in whole-brain cortical GM volume were tested using an analysis of covariance (ANCOVA) model with age as a covariate. Lobar differences were tested using the same ANCOVA model, to determine which lobes contributed most to the reduction in cortical volume. Another ANCOVA model including age and whole-brain cortical GM volume as covariates was used to test for differences in mean cortical thickness and hippocampi and amygdalae volumes between the three groups (controls, MCI-c, and AD). The adjusted values for mean cortical thickness provide us with information about which regions are more deeply affected by the disease in the frame of global atrophy.

In order to test for regional differences in CBV and CBF mean values, we used an ANOVA model, as no effects of age or whole-brain cortical GM volume were observed in perfusion values.

Whenever the ANCOVA or ANOVA model revealed significant differences between groups, we used a post hoc procedure (Dunnett's test) to compare means between each patient group and controls. In variables showing significant differences between group means, the effect size (Cohen's coefficient, d) was calculated to assess the magnitude of those differences. The correlation between cortical thickness and perfusion variables (CBF, CBV) was studied using Pearson's correlation. Data were analyzed using Statistical Analysis System version 9.0 (SAS Institute, Cary, NC, USA).

Results

Demographic Data

The demographic and clinical data of the participants are shown in Table 1. No significant differences between groups were found for any of the variables studied, except in the Mini Mental State Examination score, which was lower in the AD and MCI-c groups than in the control group (P<0.0001).

Structural Measurements

Considering the age effect in the ANCOVA, AD and MCI-c patients had smaller global cortical volumes than controls. Dunnett's test showed that those differences were more significant in MCI (P<0.01) than in AD (P<0.05). In a lobe-based analysis, ANCOVA revealed significant differences in the frontal lobes (P<0.05), right cingulated cortex (P<0.05), parietal lobes (P<0.01), and temporal lobes in their lateral (P<0.01) and medial aspect (P<0.01). Only the occipital lobes did not have a role in global cortical atrophy.

Analysis of covariance for the adjusted values of the mean cortical thickness was only significant in the medial temporal lobes (P<0.01) and posterior cingulated cortex (P<0.05). However, when each gyrus was analyzed independently, cortical thinning was detected in some of the gyri in the parietal and frontal lobes in the patient groups. Table 2 shows the effect size and statistical significance (Dunnett's test) for gyrus-based ROI differences. Both patient groups also had lower GM volumes in the hippocampi and amygdalae after adjustment for age and global cortical volume (Table 3). Although MCI-c patients had more widespread atrophy than AD patients, the atrophy in the latter group was clearly more pronounced in the left pars triangularis, bilateral entorhinal cortex, left amygdala, and left hippocampus (Tables 2 and 3).

Table 2. Cohen's coefficient/Dunnett's test P value for between-group differences.

		Cortical thickness		CBF		CBV
Cortical regions	Hemisphere	Controls-MCI-c	Controls-AD	Controls-MCI-c	Controls-AD	Controls-MCI-c	Controls-AD
Frontal
Pars triangularis	Left	−0.95/0.021	−1.73/0.002	—	—	—	—
Pars opercularis	Left	—	—		−1.29/0.005	—	—
Pars orbitalis	Right	—	—		—	—	0.80/0.035
Paracentral	Left	−1.39/0.004	—		—	—	—
	Right	—	—	−0.81/0.022	—	—	—
Precentral	Right	−1.42/0.008	—	—	—	—	—
Rostral medial	Left	—	—	—	−1.48/0.002	—	—
	Right	—	—	—	−1.11/0.006	—	—
Parietal
Superior gyrus	Left	—	—	−1.21/0.003	—	—	—
	Right	—	—	−0.88/0.021	—	—	—
Supramarginal	Left	—	—	—	−1.25/0.003	—	—
Precuneus	Left	−1.32/0.011	—	—	—	—	—
	Right	−1.26/0.014	—	−0.84/0.031	—	—	—
Temporal (lateral)		—	—	—	—	—	—
Superior gyrus	Left	—	—	—	−1.44/<0.001	—	−1.23/0.003
	Right	—	—	—	−1.10/0.016	—	—
Banks	Left	—	—	—	−1.19/0.004	—	−0.78/0.020
Medial gyrus	Left	—	—	—	−1.17/0.004	—	—
Temporal (medial)
Entorhinal cortex	Left	−1.35/0.001	−2.49/<0.0001	—	—	—	—
Entorhinal cortex	Right	−1.31/0.005	−2.02/<0.001	—	—	—	—
Parahippocampal gyrus	Right	−1.03/0.008	−1.35/0.016	0.98/0.021	0.84/0.040	—	0.85/0.023
Fusiform gyrus	Right	—	—	—	1.55/<0.001	—	—
Posterior Cingulate		—	—	—	—	—	—
Isthmus division	Right	—	−1.26/0.014	—	—	—	—
Occipital
Lingual gyrus	Left	—	—	—	0.89/0.032	—	−0.94/<0.01

AD, Alzheimer's disease; CBF, cerebral blood flow; CBV, cerebral blood volume; MCI-c, mild cognitive impairment converters.

Negative coefficient=control values greater than patient values; positive coefficient=control values less than patient values.

Table 3. Cohen's coefficient/Dunnett's test P value for between-group differences.

		Gray matter volume		CBF		CBV
Subcortical regions	Hemisphere	Controls-MCI-c	Controls-AD	Controls-MCI-c	Controls-AD	Controls-MCI-c	Controls-AD
Hippocampus	Left	−1.48/0.026	−1.98/<0.001	—	—	—	—
	Right	−1.80/0.006	−1.67/0.010	—	—	—	—
Amygdale	Left	−1.56/0.003	−1.81/0.001	—	—	—	—
	Right	−1.46/0.005	−1.31/0.035	—	—	—	—

AD, Alzheimer's disease; CBF, cerebral blood flow; CBV, cerebral blood volume; MCI-c, mild cognitive impairment converters.

Negative coefficient=control values greater than patient values.

Cerebral Blood Volume and Cerebral Blood Flow Data

We did not observe statistically significant differences between patients and controls for mean CBV or CBF values in the whole-brain cortex. For regional CBV data, ANOVA revealed statistical differences in the left occipital lobe (P<0.05) and left lateral temporal lobe (P<0.01), with lower mean values in the AD group than in the controls. No significant differences were found for CBV data in the MCI-c group. For regional CBF data, ANOVA revealed statistical differences between groups in the parietal lobes (P<0.05), left lateral temporal lobe (P<0.05), and right medial temporal lobe (P<0.05), where these differences revealed a hyperperfusion pattern: both patient groups had higher mean values than controls. Dunnett's post hoc analysis revealed perfusion differences (only in CBF) in the MCI-c group in the parietal and temporal lobes; perfusion differences (CBV or CBF) in the AD group were also detected in the frontal and occipital lobes (Figures 2 and 3). Table 2 shows the effect size and statistical significance based on Dunnett's test for gyrus-based ROI differences.

Figure 2.

Representation of regions of interest (ROIs) showing significant differences in the post hoc test (Dunnett) between patients with mild cognitive impairment and controls. P value of Dunnett's test and effect size of the differences in represented ROIs are shown in Table 2. CBF, cerebral blood flow; GM, gray matter.

Figure 3.

Representation of regions of interest (ROIs) showing significant differences in the post hoc test (Dunnett) between patients with Alzheimer's disease and controls. P value of Dunnett's test and effect size of the differences in represented ROIs are shown in Table 2. CBF, cerebral blood flow; CBV, cerebral blood volume; GM, gray matter.

Correspondence between Perfusion Values and Atrophy

In the MCI-c group, CBF abnormalities coincided with a reduction in cortical thickness only in the right precuneus and right parahippocampal gyrus (Figure 2). The correlation between CBF data and the mean cortical thickness in the right precuneus was not significant, regardless of whether we used raw values for mean cortical thickness or adjusted values after correcting for age and whole-brain cortical GM volume. However, a significant negative correlation was detected between CBF and adjusted values of mean cortical thickness in the right parahippocampal gyrus (r=–0.61; P=0.012).

In AD, an increase in CBV and CBF coincided with cortical thinning in the right parahippocampal gyrus (Figure 3), although no correlation was found between variables with raw or adjusted values for cortical thickness.

Discussion

Cortical thinning in the initial stages of AD does not necessarily match hypoperfused brain regions, thus reinforcing the contribution of perfusion MRI as independent variable in the characterization of AD (Figure 2 and Figure 3). Our study suggests that CBF deficits appear before CBV deficits in the progression of AD, since CBV abnormalities were only detected at the stage of AD, whereas CBF changes were already detected in the MCI stage. Therefore, our results confirm the hypothesis that CBF is a more sensitive parameter than CBV for perfusion abnormalities in patients with MCI. CBF could reflect neuronal dysfunction, whereas CBV could reflect changes in microvasculature architecture as a consequence of prolonged neuronal dysfunction.²²

Structural Measurements

The analysis by lobar regions of interest showed a characteristic pattern of atrophy in early AD, which was previously described in the literature using the same brain parcellation method.²³ The medial temporal lobes and posterior cingulated cortex seem to present higher atrophy in early stages of disease, as significant differences between groups persisted after correcting for global cortical atrophy. Changes in the posterior cingulate cortex could be closely related to those in the entorhinal cortex in patients with AD.²⁴ Atrophy in the AD group was more pronounced than in the MCI-c group in the entorhinal cortex, left hippocampus, left amygdala, and left pars triangularis. This observation could be explained by longer disease duration or stronger tissue involvement in AD than in MCI-c.

Perfusion Measurements

A significant decrease in CBF in parietal lobes was detected in the MCI-c group, although no significant changes in CBV were detected in this group. Were it confirmed that hypoperfusion in the parietal lobes is a predictor for rapid conversion from MCI to AD,²⁵ our results would suggest that CBF could be a more useful biomarker than CBV for progression of MCI to AD. A reduction in capillary diameter in response to neuronal deactivation could have a greater effect on CBF values than on CBV, probably because of the Hagen–Poiseuille equation, according to which flow in a tube (vessel) of a radius R changes with a factor R⁴, whereas the corresponding volume of the tube changes with a factor R². Therefore, cerebral changes would imply a relatively greater change in flow than in volume.²⁶ Prolonged neuronal deactivation or the spread of AD neuropathology across the brain could have a more intense effect on microvasculature architecture (capillary density), thus increasing CBV differences in more advanced stages of AD.²⁷ Cerebral blood volume could be more related to tissue atrophy.¹¹

A hyperperfusion pattern was detected in the medial temporal lobes in both patient groups, suggesting the existence of compensatory mechanisms.^{28, 29} In the MCI-c group, this hyperperfusion was detected as an increase in CBF, whereas in AD it was also detected as an increase in CBV. The increase in CBV in the AD group could represent changes in the microvasculature architecture. On neural activation, astrocytes not only regulate capillary diameters³⁰ but they also stimulate the mitogenic activity of capillary endothelial cells, suggesting that prolonged functional hyperemia may initiate angiogenesis that could result in increased capillary density.³¹ This observation could explain the findings of Bell and Ball,²² who detected areas of the medial temporal lobes with higher levels of capillary density in the brains of AD patients than in the brains of controls. In those areas, capillary density correlated positively with neuronal degeneration. Hyperperfusion has also been attributed to inflammatory processes.⁸ Inflammation studies in AD have reported activated microglia surrounding β-amyloid deposits within the brain which may be responsible for a locally induced chronic inflammatory response.^{32, 33} Inflammation could explain the increase in capillary density in those areas with upregulated inflammatory responses.^{34, 35}

Correspondence between Perfusion Values and Atrophy in Patients

Although the MCI-c group presented more widespread atrophy than AD patients, the AD group showed more hypoperfused regions than the MCI-c group. These findings could support the idea that atrophy does not explain cognitive deficits per se and that vascular factors have a key role in the decline of cognition in AD;³⁶ therefore, treatments that stabilize or increase brain perfusion could delay cognitive deterioration or even improve cognitive functions.^{37, 38, 39} In MCI-c, a negative correlation between CBF values and cortical thickness in the right parahippocampal gyrus suggests an increase in CBF that is dependent on cortical atrophy. In AD, this correlation disappears, suggesting a breakdown of compensatory mechanisms⁴⁰ or a reduction in inflammatory processes in more advanced stages of the disease. Future studies in this field could elucidate the etiology of hyperperfusion in predementia and early stages of AD.

Our study is subject to a series of limitations. Statistical power is limited by the reduced sample size, and differences in spatial resolution between T1 and perfusion image studies may cause tissue registration mismatch; therefore, errors in the cortical parcellation of CBF and CBV maps cannot be controlled. The most important limitation of this study is that we performed a cross-sectional comparison of two groups of patients in different clinical situations instead of a longitudinal study. Longitudinal mapping of perfusion parameters could prove more interesting for the characterization of perfusion abnormalities during progression of AD. The main asset of our study is the multimodal nature of the data (T1 and perfusion images), which allowed us to apply an easily reproducible method for cortical parcellation, by incorporating structural information in the perfusion data. In addition, our access to information on the clinical progression of MCI-c enabled us to obtain the greater value of CBF compared with CBV as a perfusion deficit biomarker in a group of MCI patients who converted to dementia due to AD.

In summary, in the MCI converter group, we observed a significant decrease in CBF in the parietal lobes and a significant increase in CBF in the right medial temporal lobe. The negative correlation between CBF values and cortical thickness in the right parahippocampal gyrus suggested an increase in CBF that was dependent on cortical atrophy in predementia stages of AD. As perfusion deficits were detected in CBF and not in CBV, our results suggest that CBF maps can better characterize perfusion abnormalities in predementia stages of AD than CBV.

The authors declare no conflict of interest. Sponsorship comes exclusively from public research funding, and there is no commercial or financial involvement that might present a conflict of interest in connection with this manuscript.