Journal of Cerebral Blood Flow & Metabolism (2007) 27, 15631572 & 2007 ISCBFM All rights reserved

0271-678X/07 $30.00
www.jcbfm.com

Aging effects on cerebral blood and cerebrospinal fluid flows

Souraya Stoquart-ElSankari1, Olivier Baledent2, Catherine Gondry-Jouet3, Malek Makki4, Olivier Godefroy1 and Marc-Etienne Meyer2
1

Department of Neurology, Amiens University hospital, Amiens Cedex, France; 2Department of Imaging and Biophysics, Amiens University hospital, Amiens Cedex, France; 3Department of Radiology, Amiens University hospital, Amiens Cedex, France; 4Department of Radiology, Wayne State University, Childrens Hospital of Michigan, Detroit, Michigan, USA

Phase-contrast magnetic resonance imaging (PC-MRI) is a noninvasive reliable technique, which enables quantification of cerebrospinal fluid (CSF) and total cerebral blood flows (tCBF). Although it is used to study hydrodynamic cerebral disorders in the elderly group (hydrocephalus), there is no published evaluation of aging effects on both tCBF and CSF flows, and on their mechanical coupling. Nineteen young (mean age 2764 years) and 12 elderly (7169 years) healthy volunteers underwent cerebral MRI using 1.5 T scanner. Phase-contrast magnetic resonance imaging pulse sequence was performed at the aqueductal and cervical levels. Cerebrospinal fluid and blood flow curves were then calculated over the cardiac cycle, to extract the characteristic parameters: mean and peak flows, their latencies, and stroke volumes for CSF (cervical and aqueductal) and vascular flows. Total cerebral blood flow was (P < 0.01) decreased significantly in the elderly group when compared with the young subjects with a linear correlation with age observed only in the elderly group (R2 = 0.7; P = 0.05). Arteriovenous delay was preserved with aging. The CSF stroke volumes were significantly reduced in the elderly, at both aqueductal (P < 0.01) and cervical (P < 0.05) levels, whereas aqueduct/cervical proportion (P = 0.9) was preserved. This is the first work to study aging effects on both CSF and vascular cerebral flows. Data showed (1) tCBF decrease, (2) proportional aqueductal and cervical CSF pulsations reduction as a result of arterial loss of pulsatility, and (3) preserved intracerebral compliance with aging. These results should be used as reference values, to help understand the pathophysiology of degenerative dementia and cerebral hydrodynamic disorders as hydrocephalus.
Journal of Cerebral Blood Flow & Metabolism (2007) 27, 15631572; doi:10.1038/sj.jcbfm.9600462; published online 21 February 2007 Keywords: aging; CBF; CSF; PC-MRI

Introduction
The increase in average length of life is one of the most obvious benefits of scientific and medical progress of the past century: the number of people aged above 65 years in the developing countries is expected to increase from 249 millions in 2000 to 690 millions in 2030 (CDC, Public Health and Aging, 2003). Simultaneously, the prevalence of dementia
Correspondence: Dr S Stoquart-ElSankari, Department of Neurology, Centre Hospitalier Universitaire Nord, Place Victor Pauchet, Amiens Cedex 80054, France. E-mail: stoquart-elsankari.soraya@chu-amiens.fr and Dr O Baledent, Department of Imaging and Biophysics, Centre Hospitalier Universitaire Nord, Place Victor Pauchet, Amiens Cedex 80054, France. E-mail: baledent.olivier@chu-amiens.fr Received 15 September 2006; revised 21 December 2006; accepted 22 December 2006; published online 21 February 2007

increases dramatically with age: from 1.2% in people aged from 65 to 69 years, to 28.5% over 90 years (Lobo et al, 2000). Although the exact frequencies of different types of dementia remain difficult to assess because of differences in the used criteria, Alzheimer disease (AD) and vascular dementia (VD) are the two most common subtypes of dementia, whereas others, such as frontotemporal dementia or normal-pressure hydrocephalus (NPH), remain rare. Although AD accounts for 60% to 70% of all dementia (Bornebroek, 2004), its pathogenesis remains poorly understood, and the limits with normal aging are still unclear. Morphological alterations consist of rapid global cortical atrophy (Resnick et al, 2003), which prevails in entorhinal and hippocampal regions. Furthermore, morphological studies of the choroid plexuses, the cerebrospinal fluid (CSF)-producing secretory system, showed

Aging effects on CBF and CSF flows S Stoquart-ElSankari et al 1564

structural alterations in AD, for example fibrosis (Jellinger et al, 1976) and thickening of the basement membrane (Serot et al, 2000). These modifications are likely to be related to decrease CSF secretion levels of transthyrein, vitamin B12, and folate (Ikeda et al, 1990; Reynolds, 1979; Serot et al, 1997), inducing CSF hydraulic disorders observed with cisternography in AD (Bartolini et al, 1982). Normal-pressure hydrocephalus is a rarer cause of cognitive deficits, responsible for 0% to 5% of all dementia (Vanneste, 2000), probably underestimated, partly because of inconsistent definitions. Recently, evidence-based guidelines for the clinical diagnosis of idiopathic NPH have been developed (Relkin et al, 2005) to promote earlier and accurate diagnosis, and improve treatment options. The diagnosis criteria are clinical (consisting of the clinical triad of gait disturbance, mental deterioration, and urinary incontinence) and neuroimaging criteria, especially on magnetic resonance imaging (MRI) (severe ventricular enlargement out of proportion to cerebral atrophy, and CSF flow voiding sign, mainly in the aqueduct). Although NPH is the sole curable cause of dementia, with spectacular improvement after shunting in 50% of idiopathic and 70% of secondary NPH cases (Vanneste, 1994), the diagnosis remains uncertain in a large group of patients, even after the analysis of clinical and MRI data, mainly because of the frequent association to periventricular and ischemic white matter lesions. In these patients, MRI quantification of CSF flows is helpful to idiopathic NPH diagnosis. Some studies have even suggested a correlation between the increase in CSF volume mobilized through the aqueduct during the cardiac cycle (CC) and the postsurgical improvement (Bradley et al, 1996). Nevertheless, further studies are required to confirm

the prognostic role of MRI for shunt-responsive idiopathic NPH, as well as the predictive values of other tests, as the external lumbar drainage or the measure of CSF outflow resistance (Marmarou et al, 2005). Phase-contrast magnetic resonance imaging (PCMRI) enables reliable (Barkhof et al, 1994), noninvading, and rapid measurements of CSF and blood flows, and benefits in knowledge of the mechanical coupling between the cerebral blood and the CSF flows throughout the CC, and their temporal coordinated succession in healthy young subjects (Baledent et al, 2001; Greitz, 1993; Henry-Feugeas et al, 2000) (Figure 1). An unbalance in this mechanical coupling is presumed to be responsible not only for pathological states such as NPH (Greitz, 2004) or AD (Uftring et al, 2000), but also for morphological changes reported in normal aging (Greitz, 1969). Age-related structural and dynamic changes in cerebral and vascular compartments have been widely studied. Greitz (1969) reported cerebral atrophy in elderly patients, even those free from cerebrovascular disease, to be responsible for brain distension, probably because of myelin loss. Simultaneously, alterations in composition of connective tissues and muscles of vessel walls are observed with aging in both animal and human models (Kalaria, 1996), resulting in cerebral hypoperfusion (Buijs et al, 1998; Kashimada et al, 1994; Scheel et al, 2000). Finally, structural modifications in choroid plexuses have been reported, mainly thickening of the basement membrane, fibrosis, and amyloidal deposition in the choroid blood vessels, resulting in a decrease of CSF secretion and turnover (Serot et al, 2000; Silverberg et al, 2003). To the best of our knowledge, few authors (Barkhof et al, 1994; Luetmer et al, 2002; Uftring et al, 2000)

Phase 1
ICSPF

Phase 2
5% after ICSPF

Phase 3
11-20 % after ICSPF

Phase 4
20-21-25 % after ICSPF

Phase 5
25 - 35 % after ICSPF

Brain Equilibrium Pressure

Arterial peak flow C2-C3 CS peak flow V4 outflow and jugula peak flow Jugular Aqueduc Monro V3 peak flow

Figure 1 Time evolution of CSF and blood dynamic flows throughout the CC. The systolic phase of CC is divided into five levels with occurrence latencies expressed in % of CC. Phase 1 starts with the systolic arterial Fi peak in the internal carotid arteries (ICSPF), which instantaneously increases cerebral blood volume. In phase 2, starting at 5% of the CC, the cervical CSF is the first to respond to brain expansion by flushing through the subarachnoid spaces. Consequently, subarachnoid cerebral spaces pressure drops, resulting in venous jugular flush peak flow which occurs on phase 3, between 11% and 20% of the CC. On the aqueduct, third ventricle (V3) and Monroe levels CSF flush peak flows occur later (20% to 25% of CC), forming phase 4, and their effects are negligible with regard to cervical CSF flows. Finally, between 25% and 35% of CC, phase 5 occurs, characterized with a state of equilibrium cerebral pressure, with entering arterial flows equal to venting venous flows, and cervical CSF flows almost negligible. Then, the successive reverse phases begin with a reverse fill CSF flow on the aqueductal level (Baledent et al, 2001).
Journal of Cerebral Blood Flow & Metabolism (2007) 27, 15631572

Aging effects on CBF and CSF flows S Stoquart-ElSankari et al 1565

have studied the effects of aging on CSF flows and their coupling to vascular and CSF cerebral flow dynamics; moreover, none of them has found CSF flow pattern modifications. In addition, only Luetmer et al (2002) studied the CSF stroke volume, representing the CSF volume mobilized through the aqueduct throughout the CC, that is aqueductal compliance, and they found no significant modification with aging (Luetmer et al, 2002). Finally, previous studies did not include cervical CSF flow evaluation, because of the complex anatomy of the subarachnoid spaces, although many authors pointed to the major role of spinal compartment in the early systolic regulation of the brain expansion (Baledent et al, 2001; Greitz et al, 1992). In this study, we used an in-house automated segmentation algorithm, which enables reliable and reproducible flow measurements of both cervical and aqueductal CSF flows (Baledent et al, 2001, 2004). Our purpose was to evaluate the effects of aging on both vascular and CSF cerebral flows, and on the dynamic coupling between these compartments. The potential clinical output is reducing the agerelated effects in the evaluation of pathological states such as NPH, AD, or cerebrovascular diseases.

USA). Patients were supine. Conventional morphologic sequences were acquired in each patient depending on clinical request. Then, flow images were acquired with a two-dimensional cine PC-MRI pulse sequence with retrospective peripheral gating, so that the 32 frames analyzed covered the entire CC. The MRI parameters were as follows: echo time TE = 6 msecs to 9 msecs; repetition time TR = 20 msecs; flip angle = 251 for vascular flows, 201 for CSF flows; field of view FOV = 16 12 mm2; matrix = 256 128; and slice thickness = 6 mm. Velocity (encoding) sensitization was set at 80 cm/s for the vessels, 10 cm/sec for the aqueduct, and 5 cm/sec for the cervical subarachnoid space. Sagittal scout view sequences were used as localizer to select the anatomical levels for flow quantification. The acquisition planes were selected perpendicular to the presumed direction of the flow. Sections through C2 to C3 subarachnoid spaces level and sylvius aqueduct were used for CSF, and sections through C2 to C3 level were used to measure vascular flows in left and right internal carotids, vertebral arteries, and internal jugular veins. The acquisition time for each flow series was approximately 2 mins, with slight fluctuation that depended on the participants heart rate.

Data Analysis

Materials and methods

Participants
The studied population, defined as normal elderly (NE), consisted of 12 elderly subjects (eight women and four men). The mean age was 7179 years. They were all admitted in the neurological department of our clinic, and underwent MRI of the brain to screen seizures, transient neurological deficit, or gait troubles. Exclusion criteria were a cognitive decline (all subjects underwent a neuropsychological screening with a Mini Mental State Examination (Folstein et al, 1975) and the scores were required > 26/30 for all), a relevant cerebral neurological disease (cerebrovascular accident, meningoencephalitis, tumor), or relevant cerebrovascular risk factors, except arterial hypertension controlled by medication. None of these subjects had ventricular enlargement on the morphological MRI sequences. A regular heart rate (expressed as beats per minute) was required in all participants, and was recorded using a peripheral gating device placed on the subjects forefinger, to trigger the sequences. The control population, hereby called normal young (NY) population, consisted of 19 healthy young volunteers (3 women and 16 men), who underwent cine PC-MRI. The mean age was 27.574.4 years. The exclusion criteria were any neurological, psychiatric or severe general disease, alcoholism, or abnormalities detected with clinical MRI exam.

Data Acquisition
All MRI exams were performed using a 1.5-T machine (Sigma; General Electric Medical System, Milwaukee, WI,

Data were analyzed using an in-house image processing software (Baledent et al, 2001) with an optimized CSF and blood flows segmentation algorithm, which automatically extracts the region of interest at each level and calculates its flow curves over the 32 segments of the CC (Figure 2). Then, the mean venous, arterial, and CSF flow curves were generated against either CC or image index. For the arterial flows, we calculated the mean cerebral arterial total cerebral blood flow (tCBF) in milliliters per minute, which was the sum of mean arterial flows in left and right internal carotids and vertebral arteries. For the venous flows, a correction factor was applied to the amplitude of the internal jugular flow to obtain equality between the mean total inflow arterial and the mean total outflow venous rates, as described by previous works (Alperin et al, 1996; Baledent et al, 2001). The total cerebral vascular flow curve was generated by calculating the difference between arterial and venous flows throughout the CC. The time integral of this arteriovenous (AV) curve gave the intracranial blood volume change during the CC. The absolute sum of this volume maximum change in the caudal and cranial directions during the CC was defined as the blood stroke volume, responsible for the dynamic coupling and succession of cerebral flows, starting with the CSF cervical venting. Similarly, CSF flow curves in both aqueductal and cervical levels were integrated, providing the CSF stroke volumes, which represent the CSF volumes displaced in both directions through the considered region of interest at each level (Nitz et al, 1992). These volumes represent the mobile compliance of ventricular and subarachnoid compartments and contribute to rapid regulation of intracranial pressure throughout the CC.
Journal of Cerebral Blood Flow & Metabolism (2007) 27, 15631572

Aging effects on CBF and CSF flows S Stoquart-ElSankari et al 1566

Figure 2 A snapshot of the graphic user interfaces of the home-made software. (A) A sagittal T1W image is used to localize the acquisition plane at the cervical C2 to C3 and aqueductal levels. At the cervical level (velocity encoding Venc at 80 cm/sec), we first get the PC-MRI image of the cervical vascular flows (B): black pixels correspond to arterial inflow, white pixels to venous outflow, and grey pixels denote static tissues. The useful image (i.e., right internal carotid) is manually selected in the considered region of interest, and zoomed out (C). Then a semiautomated extraction algorithm outlines the vessel outline (D), based on the temporal evolution of vascular flow velocities through the cardiac phases. Finally, the algorithm calculates the right carotid arterial flow curve (E). By applying this procedure to the right and left arteries and veins, we get the temporal evolution of total cerebral vascular flow over a CC. Changing the Venc to 10 cm/sec at the cervical level, and repeating the processing algorithm to the subarachnoidal spaces region of interest (F, G, and H), we obtain the cervical CSF flow curve (I).

For each of these curves, amplitude and temporal characteristical parameters were analyzed.

Statistical Analysis
Amplitude and temporal parameters comparison between NY and NE subjects was performed using a univariate analysis, with a nonparametric MannWhitney test. The level for statistical significance (P-value) was set at 0.05. The age effect on tCBF was evaluated using two different methods. When considering the two studied groups as a whole (NE and NY mixed together), the annual decrease in tCBF was calculated by dividing the mean tCBF difference ((tCBF of NE)(tCBF of NY)) by the mean age difference ((mean age of NE)(mean age of NY)). Conversely, when considering separately the NE or NY population, we used linear regression analysis, extracted the linear equation, and calculated the regression coefficient R2.

Amplitude Analysis: For the arterial flows, we analyzed tCBF, arterial systolic flows, and diastolic peak flows. For the venous flows, we analyzed mean cerebral venous jugular blood flow, venous systolic peak flows, and venous diastolic peak flows. For each of the CSF flow curves, the analyzed parameters were CSF peak fill flow (Fi) and CSF peak flush flow (Fl). Temporal Flow Analysis: The analysis of key temporal parameters gives the succession of dynamic volume changes in vascular and CSF compartments. Comparing the latencies of maximum flow peaks provides an indirect evaluation of the delays in wave propagation, that is the dynamic coupling among the arterial, venous, and CSF cerebral compartments (Figure 1). We defined the time reference (or zero time) as the averaged time of the arterial Fi peaks in the right and left arterial carotid arteries. Temporal parameters were the time to peak fill flow (LFi) and to peak flush flow (LFl) for CSF flow curves, the latencies of systolic and diastolic peaks (LAs and LAd) for arterial flows, and the systolic and diastolic peak flows (LVs and LVd) for venous flows. For the AV flow curve, we analyzed the arterio venous delay (AVD), representing the latency between the arterial systolic inflow peak and the outflow venous peak. All temporal parameters were expressed in terms of percentage of the CC.
Journal of Cerebral Blood Flow & Metabolism (2007) 27, 15631572

Results
Flow Curves

The mean flow curves of NY and NE populations were generated for aqueductal and cervical CSF, as well as for arterial, venous, and AV flows. Figure 3 demonstrates a global decrease in flow values in the NE population when compared with NY population.
Analysis of Blood Flow

The mean values of key parameters of arterial, venous, and AV blood flows in young and elderly groups are shown in Table 1.

Aging effects on CBF and CSF flows S Stoquart-ElSankari et al 1567

The mean arterial tCBF was significantly (P < 0.01) decreased in NE subjects (5097103 mL/min) when compared with NY adults (6887115 mL/min). The maximum peaks of arterial systolic and diastolic

0 -100 -200 -300 -400 -500 -600 -700 -800 -900 -1000 cranial flow ml/min -1100 diastolic period 3 13 22 31 41 50 59 69 78 88 97 % of cardiac cycle

Arterial flow

Systolic period

flows were significantly decreased in NE (Table 1), without difference in their latencies. Similarly, the mean out-fill jugular venous flow was decreased in NE subjects (3677124 mL/min) in comparison with NY (4827147 mL/min). The internal jugular systolic flow peak was significantly lower for NE, but occurred at the same latency observed in NY (Table 1). Considering the global AV cerebral flow curve, there were no significant differences in AV stroke volume between the NE (0.7270.3 mL) and the NY (0.870.2 mL) adults. Finally, the AVD was not significantly different between NE (22%713% of CC) and NY (17%77% of CC) populations.
Analysis of Cerebrospinal Fluid Flow

1100 1000 900 800 700 600 500 400 300 200 100 0 3

ml/min caudal flow

Venous flow
diastolic period

Systolic period

% of cardiac cycle 13 22 31 41 50 59 69 78 88 97

Values of fill and flush CSF peak flows, of their latencies, and of CSF stroke volumes are shown in Table 2. There were no significant differences in peaks latencies when comparing NE and NY flow curves. The unique noticeable difference in CSF flows concerned the Fl peak for cervical CSF, which was significantly lower in NE (P < 0.01). There was a significant decrease in CSF stroke volume with age at the cervical level as well as at the aqueductal level (P < 0.05). Finally, we calculated the ratio (cervical CSF stroke volume/aqueductal CSF stroke volume), and did not find a significant difference (P = 0.9) between NY (15.779.6) and NE (15.677.5).
Blood Flows and Age Relationships

200 150

ml/min caudal flow

Cervical CSF flow

100 50 Flush period 0 3 -50 -100 cranial flow ml/min caudal flow 13 22 31 41 50 % of cardiac cycle 59 69 78 88 97

The annual decrease in tCBF among all subjects (NY and NE together) was calculated and the annual decrease coefficient was of 4.1 mL/min per year.
Figure 3 Flow curves are represented in young (solid lines) and elderly (dash lines) subjects. Flows are plotted during the successive phases of a CC represented in %, with the time reference (or zero time) corresponding to the averaged time of the arterial fill-flow peaks in the right and left arterial carotid arteries. By convention, fill (or cranial) flows are represented with negative values, and flush (or caudal) flows with positive values. Flows are expressed in mL/min. Each flow curve for both populations represents mean flow values, and s.d. (not shown for commodity) is p30%. This figure shows a global decrease in flow values in the elderly population when compared with the young population. Vascular arterial flows (A) have similar patterns in the two populations, with peaks occurring at similar latencies. The only difference is smoother peaks in the elderly populations flow curve. Jugular venous flows (B) show, in a similar way, comparable patterns between the two populations. Cervical (C) and aqueductal (D) CSF flow curves show two separate and opposite phases through the CC: a flush period and a fill period. During the systolic phase, CSF flushes first into the cervical subarachnoid spaces at the same latency for both populations (88% of CC) and later through the aqueductal level (at 100% of CC for both populations). After a shift point, CSF fill period begins at the two levels.
Journal of Cerebral Blood Flow & Metabolism (2007) 27, 15631572

Fill period

15 10

Aqueductal CSF flow

5 0 3 -5

Flush period % of cardiac cycle 13 22 31 41 50 59 69 78 88 97

Fill period

-10 -15 cranial flow

Aging effects on CBF and CSF flows S Stoquart-ElSankari et al 1568

Table 1 Amplitude and temporal parameters of arterial, venous, and arteriovenous blood flow in normal elderly and normal young adults
NE subjects Mean heart rate Arterial total flow FA (mL/min) LA (% CC) FAd (mL/min) LAd (% CC) 78710 8677168 070 284786 7876 NY subjects 7178 10887153 070 460799 8173 6567343 22713 3377108 89710 P < 0.05 0.001 NS < 0.001 NS NS NS 0.036 NS
tCBF (ml/min)

1000 900 800 700 600 500 400 300 200 100 0 0

tCBF related to age in the NY and NE populations

NY NE Linear (NE) 20 40 60 80 100 y = -9,3507x + 1178,2 R2 = 0,7173

Age (years)

Venous cervical total flow FV (mL/min) 5267250 LV (% CC) 1777 FVd (mL/min) 2357110 LVd (% CC) 84717

For arterial flows, maximum arterial systolic (FAs) and diastolic (FAd) peaks are shown in mL/mins, and their latencies (LAs and LAd) expressed in % of cardiac cycle (CC). For venous flows, maximum venous systolic (FVs) and diastolic (FVd) peaks are represented in mL/min, and their latencies (LVs and LVd) are expressed in % of CC.

Figure 4 Linear reduction of total arterial cerebral blood flow (tCBF) with age in the normal young (NY) and normal elderly (NE) groups. For each subject, tCBF value is plotted, with the x axis showing subjects age. In the NY population, there was no linear regression of tCBF with age (slope = 0.065; intercept = 689.43; R2E0; P < 0.05). Conversely, when considering the elderly subjects, tCBF reduction was continuous and linear with age, as shown by a linear regression analysis (slope = 9.35; intercept = 1178.2; P = 0.0001).

Table 2 Amplitude and temporal values of cerebrospinal flows in normal elderly and normal young subjects
NE subjects Cervical C2C3 CSF flow Fl (mL/min) LFl (% CC) Fi (mL/min) Lfi (% CC) Stroke volume (mL/CC) Aqueductal CSF flow Fl (mL/mins) LFl (% CC) Fi (mL/mins) Lfi (% CC) Stroke volume (mL/CC) NY subjects P

144744 775 94723 7079 0.45770.147 9.874.6 2174 873 8473 0.03470.016

196742 473 106728 69710 0.57570.124 11.174.9 2075 10.476 8873 0.04870.023

< 0.01 NS NS NS < 0.05 NS NS NS NS < 0.01

For cervical and aqueductal cerebrospinal (CSF) flows, fill and flush peaks (respectively Fi and Fl) are represented in milliliters per minute (mL/min), and their latencies (LFi and LFl) expressed in % of cardiac cycle (CC). The stroke volume, representing the CSF total volume displaces throughout the cervical or aqueductal level throughout the CC, is expressed in mL/CC. NS (no significance) represents P-values > 0.05.

and reliable diagnosis tools to promote early diagnosis and to improve treatment outcome. Greitz (1969) suggested that most of these cerebral diseases are due to an unbalance in the mechanical intracranial coupling between vascular and CSF flows. However, to our knowledge, only few studies considered the coupling between blood and CSF flows in NPH (Baledent et al, 2004; Luetmer et al, 2002) and AD (Uftring et al, 2000) patients. Furthermore, in these populations studied, samples were too small to establish pathological patterns of CSF flows. In addition, patients were compared with NY adults in most of these studies (Baledent et al, 2004), which introduces age-related slants, because AD, NPH, and other dementias affect older populations. Barkhof was the first to use PC-MRI to evaluate effects of aging on aqueductal CSF flow (Barkhof et al, 1994). He found flow patterns in young and elderly subjects to be similar; however there was an earlier peak flush latency in the elderly. Our pilot study is the first to evaluate both vascular and CSF intracranial flows, and to correlate their mechanical coupling with aging.

When the NE population was considered separately, as expected, we found by linear regression a significant gradual and continuous decrease with age (Figure 4). For these 12 subjects, the decrease in tCBF was about 9.38 mL/min per year (tCBF = 9.38 age + 1178.68; R2 = 0.71; P = 0.001). On the contrary, in the NY group, linear regression revealed no relation between tCBF and age (tCBF = 0.065 age + 689.43; R2 = 0; P < 0.05).

Vascular Flows

Discussion
With the dramatically increasing incidence of dementia, it seems necessary to develop powerful
Journal of Cerebral Blood Flow & Metabolism (2007) 27, 15631572

Phase-contrast magnetic resonance imaging enables tCBF measurement with a simple acquisition. In our study, in the NE population, a significant decrease in tCBF was observed. The first studies considering aging effects on cerebral circulation found that cerebral blood flow and oxygen consumption were maintained, whereas dementia (both AD and VD) was related to disturbances in brainblood flow and oxidative metabolism (Hoyer, 1982). These studies measured CBF in milliliters per gram of brain tissue per minute, which enables correction of tCBF for the

Aging effects on CBF and CSF flows S Stoquart-ElSankari et al 1569

brain volume and specific studies of regional and compartmental perfusion. Because aging is associated to a global cerebral atrophy, the observed decrease in tCBF reflects a global decrease in cerebral perfusion, without any disturbance of regional perfusion or oxygen consumption. Most of publications concerning tCBF measurements have reported decrease in tCBF with age, independent of the techniques used (Buijs et al, 1998; Kashimada et al, 1994; Scheel et al, 2000). Buijs et al (1998) used two-dimensional PC-MR angiography of the internal carotids to measure tCBF, and found a mean tCBF of 5277131 mL/min in a group ranging from 70 to 79 years, comparable to our findings (5097103 mL/min). Kashimada et al (1994) reported similar values of tCBF when applying the same technique in similarly aged patients. Scheel et al (2000) found a mean tCBF of 6037 131 mL/min in their study population (aged 52719), using transcranial color-coded sonography. These greater values found by Doppler are probably due to two reasons. The first is related to the technique itself; sonography is operator dependent, and flow measurements with Doppler may be hindered by cranial bones (especially for the vertebral arteries in the skull base) whereas PC-MR accuracy have been validated for flow measurement in vivo and in vitro studies (Bakker et al, 1995; Evans et al, 1993). The second reason is that the population studied with sonography is younger, with greater dispersion of ages. We calculated the decrease of tCBF with age, and we found a continuous and gradual decrease of 4.1 mL/min per year, similar to the linear regression results found by Buijs (4.8 mL/min per year) and Kashimada (3.9 mL/min per year), (Buijs et al, 1998; Kashimada et al, 1994). The lower annual decrease found by Scheel (3 mL/min per year) is probably related to the younger-aged population and to the errors related to sonography technique (Scheel et al, 2000). Conversely, when assessing separately the young and the elderly groups, we found different results. While there was no linear decrease of tCBF with age among young adults, we found an accentuation (almost doubling) of the linear regression straight lines slope among the elderly (9.38 mL/min per year). Although these results require confirmation in larger populations, they already suggest that tCBF decrease with age may not be really linear but may be exponential, with an acceleration of its decrease in extreme ages. Greitz et al (1992) considered that the intracranial blood and CSF flush and fill flows through the CC are initiated by the systolic intracerebral arterial inflow. Further studies, using PC-MRI, proposed a dynamic model for mechanical coupling between blood and CSF intracranial flows (Baledent et al, 2001) (Figure 1). The systolic arterial Fi peak in the carotid arteries results in an instantaneous increase of the intracranial pressure. The first and faster way to decrease intracranial pressure is a large CSF

venting in the subarachnoid spaces, which drops in cerebral subarachnoid spaces pressure. Then flush flows occur in cerebral venous and aqueductal CSF compartments. Our results suggest a decrease in the systolic intracerebral dynamic arterial impulsion with aging, as reflected by decline in arterial systolic flow values observed in the elderly. This decrease is demonstrated by the peaks loss of sharpness in the arterial curves in the elderly. These dynamic agerelated changes may be a consequence of structural modifications suggested by laboratory animal studies, showing that cerebral vessels become stiffer and less distensible with aging (Kalaria, 1996). Most of these studies showed thickening of connective tissues in vessels walls, reduction in vascular content of smooth muscle and elastin, and loss of energetic reserve in vessels mitochondria, which probably result in cerebral vessels loss of distensibility (Kalaria, 1996). Modifications of veins with aging have been less studied, because structures and functioning of veins remain partly unknown. Some authors believe that veins are collapsible, and thus passively change their cross-sectional configuration depending on applied pressure, and that their pulsations are mainly consequent of concomitant arteries pulsations (Schaller, 2004). Bateman (2000) stretched the role played by the venous system in the mechanical exchanges among intracranial blood and CSF compartments, and suggested that diseases such as NPH may be initiated by venous dysfunction. Similarly, our results argue that venous pulsations are impulsed by arterials, as venous curves in both young and elderly populations closely resemble arterial curves (Figure 3). Aging morphological effects in veins have rarely been studied in animals (Otsuka et al, 2000) and humans. They seem to consist in endothelium thinning and capillary diameter reduction, but the difference in behavior of venous system in elderly brains compared with young ones is not yet understood (Schaller, 2004). Although the direct study of the whole venous outflow system remains difficult, our results showing a decrease in the cerebral venous flow with aging are scientifically reliable since intracranial venous flow modifications are modeled on arterials. Furthermore, Bateman (2000) considered the AV delay as the total compliance of the intracerebral compartments. He pointed out that this AVD is reduced in NPH, as a result of the increase in intracranial pressure. On the other hand, agingrelated atrophy has no effect on AVD, thanks to the arteriolar and capillary flow regulation. In normally aged brain, blood flow is regulated by the arteriolar constriction, before the capillary bed, and pressure transmission to venous and CSF compartments is preserved. Bateman suggests that atrophy in the elderly brains is related to interstitial fluid flow from the cerebral capillaries to the venules, which enlarges sulci. Our results showed that AVDs were
Journal of Cerebral Blood Flow & Metabolism (2007) 27, 15631572

Aging effects on CBF and CSF flows S Stoquart-ElSankari et al 1570

comparable in the young and the elderly groups (respectively 17%77% and 22%713% of CC; P > 0.05), identical to Batemans results (AVD in healthy subjects: 89740 ms/AVD and in elderly patients with cerebral atrophy: 103760 ms/P = 0.3). Thus our results suggest preservation of cerebral compliance in aging. Finally, we found an effective although not significant decline in the AV stroke volume with aging. This parameter reflects the driving current, defined by Greitz (1993) and studied through an electrical model by Alperin (Alperin et al, 1996). This driving AV current induces currents in the other limbs of the circuit, to regulate intracranial volume and pressure exchanges. However, despite preservation in cerebral compliance with aging, fluid and brain motions are less dynamic, because of the loss in arterial impulsion pulsatility.

1998), defined as the difference between the maximum rostral and caudal flux.

Cervical Level

Cerebrospinal Fluid Flow

Limits: Technical and measurement conditions should be taken into account when studying CSF flows, both on aqueductal and cervical levels. First, aqueductal CSF evaluation may be hampered by small size of the aqueduct and ventricular foramina. However, previous phantoms studies (Barkhof et al, 1994; Luetmer et al, 2002) supported good reliability and reproducibility of PC-MRI in evaluating CSF flows at the aqueductal level. Furthermore, in a previous study comparing our semiautomated CSF segmentation algorithm and manual tracing, CSF pulsatile patterns were homogeneous when processed in the narrow part of the aqueduct where velocity dispersion is minimized (Baledent et al, 2001). The second challenge was measuring the cervical CSF flows, because of their main contribution to the early systolic regulation of intracranial pressure, as suggested by many previous studies (Baledent et al, 2001; Greitz et al, 1992). Despite the technical difficulties in cervical dynamic evaluations (Alperin et al, 1996), our home-made segmentation software assesses CSF flows and stroke volumes in a reliable way (Baledent et al, 2001). Finally, stroke volume values were variable in previous studies because of varying measurement methods. Our approach was similar to the measurement method used by Luetmer et al, (2002). These investigators measured the average fill and flush CSF flow at the level of interest (aqueduct or cervical level) throughout the CC, and estimated the heart rate in the population. Then, they calculated the corresponding CSF stroke volume first by dividing the CSF flow by the heart rate, and then by two. This method is less dependent on heart rate variation, and thus has advantages compared with flux difference technique used by EgelerPeerdeman and Barkhof (Egeler-Peerdeman et al,
Journal of Cerebral Blood Flow & Metabolism (2007) 27, 15631572

We found a significant decrease in cervical CSF flush peak, and a loss of sharpness in the cervical CSF curve peaks in the elderly group. Similar to venous flows, CSF flows discharge the arterial impulsion volume, and model their variation on arterials. Age-related arterial loss of pulsatility results in smoother CSF peaks, consistent with cervical CSF compartments loss of pulsatility. On the contrary, the latencies of cervical CSF Fis and Fls were preserved with aging, suggesting a preserved compliance of the cervical compartment in the elderly. According to Greitzs theory and to Alperins electrical model, the AV stroke volume induces modulation of cervical CSF stroke volume, through the cranial closed box, which is characterized by resistance and compliance. In steady state, these two components balance the entering driving current (represented by the AV stroke volume) and the proportional flushing CSF stroke volume. Thus, inappropriate modulations between AV and CSF stroke volumes as seen in pathologies will result from abnormal resistance and/or compliance of the intracranial compartments. In our study, AV stroke volume is not significantly reduced with normal aging, whereas cervical CSF stroke volume is reduced. This inappropriate modulation means that aging induces modification of either compliance or resistance of cranial contents. Because the AVD, reflecting total cerebral compliance, is preserved in our studied populations, we suggest that aging results in increased cerebral resistance. Similar results have been found by Uftring et al (2000), using a mathematical method based on transfer functions. These investigators studied simulated changes in brain compliance and resistance to CSF flows, and found attenuation of CSF oscillations in the elderly. They related this result to an increased resistance to CSF flows at the skull base in the elderly. Although they have studied a small population of AD patients, Uftring et al (2000) found that vascular pulsations induce larger CSF oscillations than in the elderly. Furthermore, some studies suggest that AD is related to CSF production decrease, which induces lack of Ab clearance and increased deposition of Ab in the meninges. The consequence is a greater resistance to CSF outflow (Silverberg et al, 2003). In a randomized, controlled pilot study, the same authors suggested that among AD patients, there is a subgroup of AD-NPH hybrid patients, meeting clinical criteria for AD, with elevated CSF pressure (suggesting early hydrocephalus). The implantation of a ventriculoperitoneal shunt showed stabilization in psychometric test scores in

Aging effects on CBF and CSF flows S Stoquart-ElSankari et al 1571

these patients, concomitant to a decrease in the CSF concentration of biochemical markers of AD (Silverberg et al, 2002). Although it is premature to speculate about therapeutic implications in AD, these studies suggest that AD may be associated to CSF dynamic disorders.

Aqueductal Level

dynamic intracerebral pattern of flow movements in the elderly. This pattern may afterwards be compared with patterns observed in pathological states, such as AD or HPN. Thus, it would be helpful for determining if atrophy and ventricular enlargement observed on morphological MRI in elderly patients are due to normal age-related atrophy, degenerative cerebral diseases, or ventricular dysfunction as NPH.

Interestingly, Fis and Fls were not significantly decreased in the elderly. This might be related to the fact that values of aqueductal flows are low, and they are close to the limits of accuracy of the PC-MRI (Luetmer et al, 2002). The aqueductal stroke volume was reduced significantly in the elderly, similar to the cervical level. In addition, stroke values measured in our healthy young subjects (48 mL/CC) were comparable to those of Luetmer (53 mL/CC), with a comparable method (2002). Finally, the cervical to aqueductal stroke volume ratio was preserved in the NE when compared with the NY, suggesting that aging results in proportional loss of compliance in cervical and aqueductal compartments. It would be interesting to study this ratio in degenerative dementia, particularly NPH. We suggest that NPH may be related to a disturbed coupling between AV and aqueductal flow dynamics, resulting in a disproportional decrease of aqueductal stroke volume, thus in an abnormally increased ratio, characterized by ventricular dilatation on morphological imaging. This imbalance in the distribution of pulsatility between the ventricular and the subarachnoid spaces associated to hydrocephalus have been recently suggested by Wagshul et al (2006). Furthermore, they suggest considering the ratio cervical volume to aqueductal stroke volume in shunt placement decisions in NPH, because there are patients in whom shunt placement can provide clinical improvement, despite normal aqueductal flow (with probably unbalanced cervical to aqueductal stroke volume ratio).

References
Alperin N, Vikingstad EM, Gomez-Anson B, Levin DN (1996) Hemodynamically independent analysis of cerebrospinal fluid and brain motion observed with dynamic phase contrast MRI. Magn Reson Med 35: 74154 Bakker CJ, Kouwenhoven M, Hartkamp MJ, Hoogeveen RM, Mali WP (1995) Accuracy and precision of timeaveraged flow asmeasured by nontriggered 2D phasecontrast MR angiography, a phantom evaluation. Magn Reson Imaging 13:95965 Baledent O, Gondry-Jouet C, Meyer ME, De Marco G, Le Gars D, Henry-Feugeas MC, Idy-Peretti I (2004) Relationship between cerebrospinal fluid and blood dynamics in healthy volunteers and patients with communicating hydrocephalus. Invest Radiol 39:4555 Baledent O, Henry-Feugeas MC, Idy-Peretti I (2001) Cerebrospinal fluid dynamics and relation with blood flow: a magnetic resonance study with semiautomated cerebrospinal fluid segmentation. Invest Radiol 36:36877 Barkhof F, Kouwenhoven M, Scheltens P, Sprenger M, Algra P, Valk J (1994) Phase-contrast cine MR imaging of normalaqueductal CSF flow. Effect of aging and relation to CSF void on modulus MR. Acta Radiol 35:12330 Bartolini S, Inzitari D, Castagnoli A, Amaducci L (1982) Correlation of isotopic cisternographic patterns in multiple sclerosis with CSF IgG values. Ann Neurol 12:4869 Bateman GA (2000) Vascular compliance in normalpressure hydrocephalus. AJNR Am J Neuroradiol 21:1574 85 Bornebroek MBM (2004) Epidemiology of non-AD dementia. Clin Neurosci Res 3:34961 Bradley Jr WG, Scalzo D, Queralt J, Nitz WN, Atkinson DJ, Wong P (1996) Normal-pressure hydrocephalus: evaluation with cerebrospinal fluid flow measurements at MR imaging. Radiology 198:5239 Buijs PC, Krabbe-Hartkamp MJ, Bakker CJ, de Lange EE, Ramos LM, Breteler MM, Mali WP (1998) Effect of age on cerebral blood flow: measurement with ungated two-dimensional phase-contrast MR angiography in 250 adults. Radiology 209:66774 CDC, Public Health and Aging (2003) Trends in agingUnited States and worldwide. Journal of American Medical Association 289:13713 Egeler-Peerdeman SM, Barkhof F, Walchenbach R, Valk J (1998) Cine phase-contrast MR imaging in normal pressure hydrocephalus patients: relation to surgical outcome. Acta Neurochir Suppl 71:3402 Evans AJ, Iwai F, Grist TA, Sostman HD, Hedlund LW, Spritzer CE, Negro-Vilar R, Beam CA, Pelc NJ (1993)
Journal of Cerebral Blood Flow & Metabolism (2007) 27, 15631572

Conclusion
In conclusion, our investigation used reliable and validated flow segmentation technique of PC-MRI to perform a complete study of cerebral vascular and CSF flows in elderly subjects. Our results confirmed the reduction in cerebral blood flows in the elderly reported by previous studies. Interestingly, our investigation showed that the temporal interaction between vascular and CSF flows, that is, total cerebral compliance, are preserved. The magnitude of cervical CSF flows was reduced in the elderly as CSF oscillations are modeled on arterials. Surprisingly, this result was not found at the aqueductal level, probably due to technical reasons. These results are interesting because they establish a

Aging effects on CBF and CSF flows S Stoquart-ElSankari et al 1572

Magnetic resonance imaging of blood flow with a phase subtraction technique. In vitro and in vivo validation. Invest Radiol 28:10915 Folstein MF, Folstein SE, McHugh PR (1975) Mini-mental state. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12:18998 Greitz D (1993) Cerebrospinal fluid circulation and associated intracranial dynamics. A radiologic investigation using MR imaging and radionuclide cisternography. Acta Radiol Suppl 386:123 Greitz D (2004) Radiological assessment of hydrocephalus: new theories and implications for therapy. Neurosurg Rev 27:14565; discussion 15962. Greitz D, Wirestam R, Franck A, Nordell B, Thomsen C, Stahlberg F (1992) Pulsatile brain movement and associated hydrodynamics studied by magnetic resonance phase imaging. The Monro-Kellie doctrine revisited. Neuroradiology 34:37080 Greitz T (1969) Effect of brain distension on cerebral circulation. Lancet 1:8635 Henry-Feugeas MC, Idy-Peretti I, Baledent O, PonceletDidon A, Zannoli G, Bittoun J, Schouman-Claeys E (2000) Origin of subarachnoid cerebrospinal fluid pulsations: a phase-contrast MR analysis. Magn Reson Imaging 18:38795 Hoyer S (1982) The abnormally aged brain. Its blood flow and oxidative metabolism. A reviewpart II. Arch Gerontol Geriatr 1:195207 Ikeda T, Furukawa Y, Mashimoto S, Takahashi K, Yamada M (1990) Vitamin B12 levels in serum and cerebrospinal fluid of people with Alzheimers disease. Acta Psychiatr Scand 82:3279 Jellinger K, Schnaberth G, Traugott U, Turnheim M (1976) Clinical picture, cerebrospinal fluid findings, and pathological anatomy of subacute demyelinating encephalitis (mixed form of multiple, diffuse and concentric sclerosis). Nervenarzt 47:11823 Kalaria RN (1996) Cerebral vessels in ageing and Alzheimers disease. Pharmacol Ther 72:193214 Kashimada A, Machida K, Honda N, Mamiya T, Takahashi T, Kamano T, Inoue Y, Osada H (1994) Measurement of cerebral blood flow in normal subjects by phase contrast MR imaging. Nippon Igaku Hoshasen Gakkai Zasshi 54:111625 Lobo A, Launer LJ, Fratiglioni L, Andersen K, Di Carlo A, Breteler MM, Copeland JR, Dartigues JF, Jagger C, Martinez-Lage J, Soininen H, Hofman A (2000) Prevalence of dementia and major subtypes in Europe: a collaborative study of population-based cohorts. Neurologic Diseases in the Elderly Research Group. Neurology 54:S49 Luetmer PH, Huston J, Friedman JA, Dixon GR, Petersen RC, Jack CR, McClelland RL, Ebersold MJ (2002) Measurement of cerebrospinal fluid flow at the cerebral aqueduct by use of phase-contrast magnetic resonance imaging: technique validation and utility in diagnosing idiopathic normal pressure hydrocephalus. Neurosurgery 50:53443; discussion 53840.

Marmarou A, Bergsneider M, Klinge P, Relkin N, Black PM (2005) The value of supplemental prognostic tests for the preoperative assessment of idiopathic normalpressure hydrocephalus. Neurosurgery 57:S1728; discussion iiv. Nitz WR, Bradley Jr WG, Watanabe AS, Lee RR, Burgoyne B, OSullivan RM, Herbst MD (1992) Flow dynamics of cerebrospinal fluid: assessment with phase-contrast velocity MR imaging performed with retrospective cardiac gating. Radiology 183:395405 Otsuka H, Nakase H, Nagata K, Ueda K, Kempski O, Sakaki T (2000) Effect of age on cerebral venous circulation disturbances in the rat. J Neurosurg 93:298304 Relkin N, Marmarou A, Klinge P, Bergsneider M, Black PM (2005) Diagnosing idiopathic normal-pressure hydrocephalus. Neurosurgery 57:S416; discussion iiv. Resnick SM, Pham DL, Kraut MA, Zonderman AB, Davatzikos C (2003) Longitudinal magnetic resonance imaging studies of older adults: a shrinking brain. J Neurosci 23:3295301 Reynolds E (1979) Cerebrospinal fluid folate: clinical studies. New York: Raven Press Schaller B (2004) Physiology of cerebral venous blood flow: from experimental data in animals to normal function in humans. Brain Res Brain Res Rev 46:24360 Scheel P, Ruge C, Petruch UR, Schoning M (2000) Color duplex measurement of cerebral blood flow volume in healthy adults. Stroke 31:14750 Serot JM, Bene MC, Foliguet B, Faure GC (2000) Morphological alterations of the choroid plexus in late-onset Alzheimers disease. Acta Neuropathol (Berl) 99:1058 Serot JM, Christmann D, Dubost T, Couturier M (1997) Cerebrospinal fluid transthyretin: aging and late onset Alzheimers disease. J Neurol Neurosurg Psychiatry 63:5068 Silverberg GD, Levinthal E, Sullivan EV, Bloch DA, Chang SD, Leverenz J, Flitman S, Winn R, Marciano F, Saul T, Huhn S, Mayo M, McGuire D (2002) Assessment of low-flow CSF drainage as a treatment for AD: results of a randomized pilot study. Neurology 59:113945 Silverberg GD, Mayo M, Saul T, Rubenstein E, McGuire D (2003) Alzheimers disease, normal-pressure hydrocephalus, and senescent changes in CSF circulatory physiology: a hypothesis. Lancet Neurol 2:50611 Uftring SJ, Chu D, Alperin N, Levin DN (2000) The mechanical state of intracranial tissues in elderly subjects studied by imaging CSF and brain pulsations. Magn Reson Imaging 18:9916 Vanneste JA (1994) Three decades of normal pressure hydrocephalus: are we wiser now? J Neurol Neurosurg Psychiatry 57:10215 Vanneste JA (2000) Diagnosis and management of normalpressure hydrocephalus. J Neurol 247:514 Wagshul ME, Chen JJ, Egnor MR, McCormack EJ, Roche PE (2006) Amplitude and phase of cerebrospinal fluid pulsations: experimental studies and review of the literature. J Neurosurg 104:8109

Journal of Cerebral Blood Flow & Metabolism (2007) 27, 15631572