Reduced suprasellar cistern cerebrospinal fluid motion in patients with Parkinson’s disease revealed by magnetic resonance imaging with dynamic cycling of diffusion weightings

BACKGROUND: Parkinson’s disease is characterized by dopamine-responsive symptoms as well as aggregation and accumulation of a-synuclein protofibrils. New diagnostic methods assess a-synuclein aggregation characteristics from cerebrospinal fluid and recent pathophysiologic mechanisms suggest that cerebrospinal fluid circulation disruptions may precipitate a-synuclein retention. Here, we test the hypothesis that cerebrospinal fluid motion at the level of the suprasellar cistern is reduced in Parkinson’s disease relative to healthy participants and this reduction relates to choroid plexus perfusion. METHODS: Diffusion weighted imaging (spatial resolution=1.8×1.8×4 mm) magnetic resonance imaging with cycling of diffusion weightings (b-values=0, 50, 100, 200, 300, 700, and 1000 s/mm2) over the approximate kinetic range of suprasellar cistern neurofluid motion was applied at 3-Tesla in Parkinson’s disease (n=27; age=66±6.7 years) and healthy (n=32; age=68±8.9 years) participants. Wilcoxon rank-sum tests were applied to test the primary hypothesis that the decay rate of cerebrospinal fluid signal as a function of b-value, which reflects increasing fluid motion, is reduced in persons with versus without Parkinson’s disease and inversely relates to choroid plexus activity assessed from perfusion-weighted magnetic resonance imaging (Spearman rank-order correlation; significance-criteria: p<0.05). RESULTS: Consistent with the primary hypothesis, decay rates were higher in healthy (D=0.00328±0.00123mm2/s) relative to Parkinson’s disease (D=0.00256±0.0094mm2/s) participants (p=0.016). This finding was preserved after controlling for age and sex. An inverse correlation between choroid plexus perfusion and decay rate (p=0.011) was observed in Parkinson’s disease participants. CONCLUSIONS: Cerebrospinal fluid motion at the level of the suprasellar cistern is often reduced in adults with versus without Parkinson’s disease and this reduction correlates on average with choroid plexus perfusion.


INTRODUCTION
Parkinson's disease (PD) is a progressive neurodegenerative disorder with symptoms of cognitive and motor decline, and pathologic aggregation and accumulation of a-synuclein proto brils 1 .As a-synuclein is distributed throughout the cerebrospinal uid (CSF), and the topography of a-synuclein clearance and aggregation may contribute to disease propagation in PD, neuro uid motion assessments in PD may be useful to develop new pathophysiologic models of disease progression 2,3 .CSF, produced within the choroid plexus complexes, is the primary medium for clearance of waste products and proteins and traverses traditional bulk and recently identi ed perivascular pathways 4,5 .Arterial pulsatility is believed to be fundamental to anterograde neuro uid movement, whereas uid transport along the parasagittal dural spaces and cranial nerves has been proposed to have egress relevance 6,7 .Major intracranial arteries are bounded by free CSF at the level of the suprasellar cistern, and as such, quantitative imaging of neuro uid motion within the suprasellar cistern may provide one relevant marker of neuro uid circulation secondary to variation in macrovascular arterial pulsatility.
In vivo assessments of this motion have been di cult given limited imaging approaches for quantifying neuro uid motion non-invasively in vivo.Bulk CSF ow at the level of the cerebral aqueduct can be interrogated using quantitative phase contrast magnetic resonance imaging [7][8][9] , and perivascular motion can be estimated qualitatively from T 2 -weighted MRI or T 1 -weighted MRI before and after intrathecal contrast agent administration 10 .However, such imaging of perivascular motion generally does not provide quantitative measures of motion and additionally suffers from the caveats that intrathecally administered contrast is contraindicated for research studies in most centers and may not re ect in vivo neuro uid kinetics.
To begin to address this limitation, we implemented a novel diffusion weighted imaging (DWI) magnetic resonance imaging (MRI) protocol and analysis approach, which incorporates dynamic diffusion weighting cycling over the approximate kinetic range of CSF motion within the suprasellar cistern.This approach, which has previously been proposed in healthy adults using categorical scoring of signal attenuation for increasing diffusion weighting 11 , is applied here to (i) demonstrate sensitivity to regions with known differences in neuro uid motion and (ii) quantify relationships between neuro uid motion and choroid plexus perfusion in older adults with and without PD.The hypothesis to be investigated is that neuro uid motion within the suprasellar cistern of patients with PD is reduced relative to agematched controls.A secondary hypothesis is that markers of CSF production activity increase in the presence of reduced neuro uid movement within the suprasellar cistern, which would be consistent with additional CSF production activity required to clear CSF along less compliant perivascular pathways as has been suggested in separate vascular disorders 12 .Findings are discussed in the context of the growing literature on neuro uid circulation and neurodegeneration.

Demographics
Adult participants with and without Movement Disorder Society clinical criteria for PD 13 with no other history of neurological disease were recruited and provided informed consent for this prospective Institutional Review Board (IRB)-approved study.Disease duration as well as the Uni ed Parkinson's disease rating scale (UPDRS 14 ) were both recorded in all participants.Exclusion criteria were: independent neurological or psychiatric condition including but not limited to Alzheimer's disease, multiple sclerosis, prior overt stroke, schizophrenia, or bipolar disorder; ow-limiting cervical or intracranial stenosis (i.e., stenosis > 70%) of the internal carotid arteries, vertebral or basilar arteries, or rst segment of the anterior, middle, or posterior cerebral arteries; independent condition expected to lead to death within one year; and any contraindication to 3-Tesla MRI.Presence of non-speci c white matter lesions was not an exclusion criterion, as these lesions become more prevalent with aging, and we sought our cohort to be generalizable and representative.Parkinsonism medications (e.g., dopamine agonist and dopamine replacement therapies) were withheld for 16-hours prior to scanning.Exclusion criteria were identical for healthy control participants and additionally expanded to include positive mental health negative for neuro-psychiatric conditions and no complaints of prior cognitive issues.
Non-imaging assessments PD and healthy control participants underwent cognitive screening with the Montreal Cognitive Assessment (MoCA) 15 and brief neuropsychological test examination with the Repeatable Battery for the Assessment of Neuropsychological Status (RBANS) 16 .PD participants were additionally screened by a board-certi ed neurologist (DOC) to ensure PD criteria were met 14 .

Experiment
Participants were scanned at 3-Tesla (Philips Healthcare, Best, The Netherlands) using body coil radiofrequency transmission and SENSE phased-array 32 channel reception.As it has been shown that CSF production activity is coupled to circadian rhythms 17 , we imaged consistently between 7:00 AM and 11:00 AM.

Analysis
Anatomical imaging and angiography scans were reviewed by a board-certi ed radiologist (CDM; experience = 9 years) to ensure that no exclusion criteria were met.Clinical history was reviewed by a board-certi ed neurologist (DOC; experience = 16 years) to ensure participants met healthy control or PD criteria.A board-certi ed neuropsychologist (CMC; experience = 5 years) reviewed all cognitive screening and neuropsychological test data to ensure stated inclusion criteria regarding cognition were met.
Choroid plexus segmentations were generated from a previously published 8 fully convolutional neural network with 3-D U-Net architecture utilizing co-registered T 1 -weighted and T 2 -weighted FLAIR images in 1 mm isotropic MNI152 space.All registrations were performed using the Advanced Normalization Tools (ANTs) nonlinear registration tools 18 .Choroid plexus perfusion was quanti ed within the atria of the lateral ventricles following previously reported procedures 8 .Brie y, the source pCASL control and label images were surround subtracted, slice delay corrected, M 0 -normalized, averaged, and t to a general kinetic model 19 using the following parameters: blood-brain partition coe cient (λ) = 0.9 mL blood/g brain, labeling e ciency from the dual-background suppressed sequence (α) = 0.8, and the 3T T 1,blood = 1,624 ms.The processed perfusion maps were co-registered to the T 1 -weighted images using the ANTs linear registration tools 18 , and the choroid plexus segmentation from the machine learning model was applied to the perfusion maps to calculate the mean perfusion values in the choroid plexus of the lateral ventricles.A minimum threshold of 10 mL/100 g/min was used for perfusion quanti cation to ensure that all values were above the noise oor.
Maps of neuro uid motion within the suprasellar cistern were calculated from the multi-shell DWI scan.
Images were corrected for bulk head motion 18 and an exponential decay model was t voxel-wise to the maps: where S is the signal as a function of b-value, D is the free diffusion coe cient for water, and S 0 is signal in the absence of diffusion-sensitizing gradients, and b is the b-value=(gGd) 2 D for the narrow pulse approximation with gradient amplitude G, the diffusion time Δ, and gradient pulse width, δ 20 .Note that in this tting procedure, which is analogous to the calculation of the apparent diffusion coe cient (ADC), larger decay rates (D) re ect faster decay across the b-value range and as such higher neuro uid motion.After model tting, resulting neuro uid motion maps were up-sampled to 1 mm isotropic resolution, and spatially normalized to the MNI space using symmetric normalization (SyN) 18 .Spatial transformations utilized the b = 50 s/mm 2 image as a template.The region comprising the suprasellar cistern was manually identi ed in template space by a board-certi ed radiologist (CDM; experience = 9 years) and decay values were averaged within this region for each participant.In addition, gray matter, white matter, and CSF masks were used to quantify the decay rates in each tissue region.

Statistical considerations and hypothesis testing
Descriptive statistics, including means and standard deviations of continuous imaging parameters, as well as percentages and frequencies of categorical parameters, were calculated.A Wilcoxon rank sum test was applied to evaluate differences in continuous measures, whereas a chi-squared test was applied to evaluate differences in categorical measures.First, to ensure that the multi-shell DWI protocol was sensitive to known regions of increased neuro uid movement, voxel-wise rate (D) values in CSF, gray matter, and white matter were contrasted across all participants to test the hypothesis that motion (mm 2 /s) was highest in CSF compared to gray matter and white matter.Summary statistics which describe normative values in each region for PD and control participants were recorded.Second, to test the overarching hypothesis that suprasellar cistern neuro uid motion was reduced in PD relative to control participants, a Wilcoxon rank sum test was applied between mean rates in the suprasellar cistern for each cohort; separate linear regression analyses which modeled the dependence of the decay rate on (i) cohort, (ii) age, and (iii) sex were performed.Finally, to understand whether reduced suprasellar cistern motion was related to elevated choroid plexus activity, Spearman rank-order correlation test was applied using the decay rate as the dependent variable and choroid plexus perfusion as the independent variable, separately for healthy and PD participants.As an exploratory analysis, separate regression analyses were performed to evaluate any dependence of (i) UPDRS (motor function) or (ii) MoCA (cognitive function) on the independent variable of decay rates within the suprasellar cistern region in PD participants.For all comparisons, signi cance was de ned as two-sided p < 0.05.

RESULTS
The participants enrolled are summarized in Table 1, which comprised 59 participants, including 27 with PD (age = 66 ± 6.7 years) and 32 healthy adults (age = 68 ± 8.9 years).Participants between groups were matched for age (p = 0.465) and no participants had evidence of vasculopathy or met neurological or radiological exclusion criteria.One healthy control with atypical cognitive screen (MoCA < 24) 21 was determined to demonstrate clinically signi cant performance de cits on neuropsychological test examination (1 + score < 2 standard deviations, or, 2 + scores < 1.5 standard deviations) 22 .Five healthy controls with atypical cognitive screens (MoCA = 22-23) were determined to demonstrate performance on neuropsychological test examination within normal limits, and thus were included in subsequent analyses.1B. Figure 1C summarizes the range of diffusion relative to different water motion, including intravoxel incoherent motion, gaussian diffusion, and non-gaussian diffusion, and the relevant approximate diffusion weightings for each regime.Figure 1D shows accelerated signal decay at the level of the CSF relative to tissue for the low bvalues used here.It should be noted that since parenchymal motion in this range may be attributable to capillary ow or neuro uid ow, we focused our analysis on neuro uid motion within the suprasellar cistern.
Figure 2 depicts the common suprasellar cistern region of interest, both in terms of the location relative to arteries of the circle of Willis and cranial nerves.The common suprasellar cistern region of interest evaluated in all participants is shown in Fig. 2B and an example of the deep learning algorithm applied to isolate the choroid plexus at the level of the lateral ventricles is shown in Fig. 2C.Note that given intersubject variability in choroid plexus anatomy, this segmentation was performed for each participant.To con rm that the method could detect differences in regions with known uid differences, Fig. 2D shows the gray matter, white matter, and CSF segmentation, along with violin plots of decay rates in Fig. 2E (Eq.1).Decay rates were signi cantly elevated (p < 0.001) in CSF relative to other tissues, as well as in gray matter (p < 0.001) compared to white matter.This nding was observed both in the PD and healthy participants.
Figure 3 shows group level decay rate maps, separately for the healthy adults and PD participants.When group level values were considered in the suprasellar cistern, the decay rates were found to be signi cantly higher in healthy (D = 0.00328 ± 0.00123 mm 2 /s) relative to PD (D = 0.00256 ± 0.0094 mm 2 /s) participants (p = 0.016), consistent with the primary study hypothesis.There was no relationship observed for the decay rate and age (p = 0.342) or sex (p = 0.604) among PD and healthy participants on linear regression.Figure 3 also summarizes the distribution of decay rates across cohorts and against choroid plexus perfusion.When the total cohort was considered, there was no relationship that was observed between the decay rate and choroid plexus perfusion (p = 0.369).However, an inverse correlation between choroid plexus perfusion and decay rate (Spearman's r = -0.51;p = 0.011) was observed when only PD participants were considered.All analyses met criteria with and without the single healthy control that met stated criterial for mild cognitive impairment.
Regression analyses yielded a trend (Spearman's r = 0.264; p = 0.183) between UPDRS score and suprasellar cistern decay rate, consistent with higher levels of motor dysfunction being associated with reduced suprasellar uid motion, however, this did not meet stated signi cance criteria.No trend for a relationship was observed with the MoCA score (Spearman's r=-0.008;p = 0.968).
Case examples are shown in Fig. 4, whereby differences in suprasellar cistern uid motion are identi able on quanti ed maps.

DISCUSSION
Diffusion weighted imaging (DWI) MRI with dynamic cycling of diffusion weightings over the approximate kinetic regime of suprasellar cistern CSF was applied in adults with PD and age-matched healthy controls.Signi cantly attenuated motion was observed in the suprasellar cistern of PD relative to healthy participants and the extent of motion attenuation was associated with increased choroid plexus perfusion in PD participants.The methodology proposed, which is similar to an apparent diffusion constant calculation from low b-value DWI MRI, can be implemented on clinical scanners and could represent an additional non-invasive tool for assessment of neuro uid kinetics in the setting of neurodegeneration.

Expansion of methodologies to assess uid movement non-invasively in vivo
Modeling diffusion signal decay as a function of b-value in principle provides quantitative kinetic information and may present a candidate non-tracer technology for quantifying neuro uid movement non-invasively in vivo.This methodology should be considered in light of other MRI-based assessments of quantitative uid movement.Most commonly, phase contrast magnetic resonance angiography, generally with velocity encoding gradients of 50-100 cm/s or higher can be used to quantify peak systolic arterial ow velocities in the major cervical vessels (velocity = 75-125 cm/s) 23 or intracranial vessels of the circle of Willis (70-160 cm/s) 24 .In major venous structures, peak systolic velocity is lower and within a healthy range of 10-35 cm/s for most major venous structures 25 .Similar phase contrast methods can be applied at the level of the cerebral aqueduct to assess CSF ow en route from the third to fourth ventricle, as peak CSF ow velocity is frequently reported as 10-20 cm/s [7][8][9] .However, these phase contrast approaches are generally not practical for assessing CSF movement in other structures given much slower and non-laminar ow.Under these physiological conditions, su ciently low velocity encoding gradients are beyond the hardware limits of clinical scanners.Blood arrival to tissue, and even CSF ow 26 and peripheral lymphatic ow 27 , have been assessed using principles of spin labeling with multiple delay times.However, the duration of the generated magnetic label is governed by the T 1 relaxation time of the water spins in the separate uids, which, at 3-Tesla, is approximately 1,624ms for blood water 28 , 1,442ms for CSF 29 , and 610ms for lymphatic uid 27 .Therefore, principles of spin labeling can only be applied to assess uid movement over short distances or when velocity is high.
This work utilizes principles of diffusion and the slower and more isotropic uid motion in the suprasellar cistern to estimate neuro uid kinetics.The theory of estimating diffusion properties from different levels of diffusion weightings has been well characterized 30 and the approach proposed here is similar to a calculation of the apparent diffusion coe cient, albeit over a range of diffusion weightings relevant to CSF motion.Recent work has applied DWI at different b-values to qualitatively estimate neuro uid motion from categorical scoring of the rate of decay in ten healthy adults (aged = 25-58 years) in posterior fossa, suprasellar cistern, and Sylvian vallecula compared to the lateral ventricle and frontal and parietal CSF spaces 11 .This work also provided support for greater CSF uid motion in the third ventricles in younger versus older participants, however, the method required radiological scoring and noncontinuous, quantitative assessment was not proposed.Our work extends the approach to enable quantitative, continuous assessments of the decay rate.

Neuro uid motion the suprasellar cistern
The suprasellar cistern is located at level of the circle of Willis, superior to the sella turcica, anterior and inferior to the hypothalamus, and bordered by the uncus of the temporal lobes.Importantly, the suprasellar cistern is lled with freely circulating CSF and additionally contains the optic chiasm, the infundibular stalk, and the cerebrovascular circle of Willis.
Growing evidence suggests that pulsatility of the intracranial vasculature is fundamental to neuro uid motion 31,32 .Compliance of the cerebrovasculature and healthy pulsatility assists with movement of neuro uids within the cerebrum, both at the level of the parenchyma and in large CSF-lled cisterns.In support of this, patients with bilateral intracranial vasculopathy have been observed to have elevated perfusion of the choroid plexus, which may be required to upregulate CSF production activity in the presence of reduced pulsatility of arteries secondary to vasculopathy 33 .It has also been shown that choroid plexus perfusion reduces after successful angiogenesis-inducing indirect surgical revascularization in these patients 33 , and also, that the choroid plexus activity may be reduced in the presence of hyper-vascularity and sickle cell anemia, where increases in cerebral blood volume are required to compensate for reduced blood oxygen content from anemia 12 .
The observation that neuro uid motion has high inter-subject variability in the cisterns has previously been shown in the intracranial vessel wall imaging literature 34 , whereby delay-alternating-nutation-withtailored-excitation uid suppression modules 35 were utilized to null basal cistern CSF signal in a velocitydependent manner, which when effective increases conspicuity of the intracranial vessel walls.However, the degree of uid attenuation has been reported to be highly participant-dependent, a nding attributed to high variability of inter-subject neuro uid motion within the basal cisterns 34 .This work extends these ndings to demonstrate that neuro uid motion within the suprasellar cistern varies between PD and agematched healthy adults, and importantly, this variation is inversely related to the perfusion of the choroid plexus in PD participants.These ndings taken together provide support for the choroid plexus upregulating activity in the presence of reductions in intracranial neuro uid motion and arterial pulsatility.

Neuro uid aberrations in Parkinson's disease
There is a growing literature reporting aberrant neuro uid circulation in PD, as well as other neurodegenerative proteinopathies 36 .Speci cally with regards to PD, dynamic contrast-enhanced magnetic resonance imaging has been applied to assess meningeal lymphatic ow in cognitively normal controls and patients with idiopathic PD or atypical Parkinsonian disorders, and it was observed that patients with idiopathic PD exhibited signi cantly reduced ow through the meningeal lymphatic vessels along the superior sagittal sinus and sigmoid sinus, as well as a notable delay in deep cervical lymph node perfusion, compared to patients with atypical Parkinsonian disorders 37 .These ndings suggest that differences in neuro uid circulation may partly underlie differences between idiopathic and atypical Parkinsonian disorders.Furthermore, mice injected with α-synuclein preformed brils demonstrated delayed meningeal lymphatic drainage, loss of tight junctions among meningeal lymphatic endothelial cells and increased in ammation of the meninges 37 .
Prior work using principles of diffusion to estimate water motion along perivascular spaces has provided evidence in support of reduced water motion along perivascular spaces in older adults with PD relative to essential tremor 2 .This nding was tentatively attributed to PD being a proteinopathy and that neuro uid movement aberrations may contribute to the retention of a-synuclein.This nding is partly consistent with other work: aquaporin-4 (AQP4) channels are likely essential for effective transport of uid between perivascular and interstitial spaces, and separate work has shown that a-synuclein deposition negatively correlates with AQP4 expression in patients with PD 38 .Additionally, Fang et al. 39 evaluated AQP4 polymorphisms in PD together with Positron Emission Tomography measures of beta-amyloid burden, sleep behavior, and CSF biomarkers and observed that AQP4 rs162009 may be a biomarker of cognitive decline in PD, possibly attributable to its role in changing neuro uid circulation.Finally, we observed no relationship between cognition, assessed with MoCA, and the imaging measures, however, this is not surprising given that most participants enrolled in this work had no or only very mild cognitive impairment.For motor impairment, assessed with UPDRS, we observed a trend for a relationship between increased motor dysfunction and slower neuro uid motion within the suprasellar cistern, however, the study was not powered to evaluate this relationship rigorously, which would require approximately 60 patients and likely a multiple regression analysis with additional control for other relevant covariates such as age, sex, and disease duration.As the proposed method can be readily applied in 5-10 min, future work that assesses how neuro uid motion within the cisterns relates to symptomatology, together with other established markers of disease severity, may be useful.

Limitations
First, the spatial resolution of typical DWI scans at the clinical eld strengths of 3 Tesla is 8-27 mm 3 , which is at or larger than the size of many structures relevant to perivascular ow.Therefore, we focused our study on the suprasellar cistern, which has a volume more than 300-fold larger than the spatial resolution of the voxel.Since the suprasellar cistern contains the major arteries of the circle of Willis, and is comprised of freely circulating CSF, it represents a logical location for assessment of neuro uid motion.However, we did not attempt to interrogate smaller structures with emerging relevance to glymphatic ow (e.g., cribriform plate, cranial nerves, or parasagittal dural spaces) nor perform spatial analysis of variance analyses given the sizes of these structures relative to the spatial resolution of the DWI acquisition.Second, given limited information available on kinetic neuro uid movement within the suprasellar cistern, it remains unclear which velocity or diffusion regime is most relevant.Work by others in 50 healthy adults and 50 patients with ventricular dilation demonstrated that at the b-value of 500 s/mm 2 , signal voids in 3rd and 4th ventricles, the cerebral sulci and the Sylvian ssure were observed, thereby providing evidence for this range being reasonable for assessing CSF ow 40 .Therefore, we applied diffusion weightings 0-1000 s/mm 2 and characterized the signal decay quantitatively.Future work will bene t from further de ning the precise range of diffusion weightings that are most relevant.
In conclusion, diffusion weighted imaging with dynamic cycling of b-values was applied to provide evidence supporting reduced CSF motion within the suprasellar cistern in participants with versus without Parkinson's disease.The extent of attenuation of uid movement corresponded with choroid plexus hyperemia in PD participants.These ndings suggest that low b-value DWI may provide a new imaging tool to interrogate neuro uid motion in the growing number of applications where neuro uid circulation dysfunction is implicated.shown to right, which highlights (yellow arrows) the high choroid plexus perfusion signal which is comparable to gray matter perfusion signal.Given the high variability in choroid plexus anatomy, structures were segmented in native space for each participant using previously-reported machine learning routines (see Methods).(D) Example of tissue segmentation used to con rm abilities to evaluate differences in neuro uid and tissue motion.(E-F) Separately in each group, decay rates are signi cantly elevated in cerebrospinal uid (CSF) relative to other tissue types (p<0.001), a well as in gray matter relative to white matter (p<0.001).The plots show boxplots overlaid on violin plots, with all participant data plotted as separate data points for completeness.See Table 1 for quantitative values.* p<0.001.

Figure
Figure1Ashows acquired representative DWI in a participant as a function of increasing b-value, whereby faster owing uid is suppressed in all b-values and slower owing uid is suppressed in larger b-values only.Corresponding decay maps (mm 2 /s), calculated using the prescribed low-to-intermediate b-values of 0-1000 s/mm 2 , in the native space of the participant are shown in Fig.1B.Figure1Csummarizes the range of diffusion relative to different water motion, including intravoxel incoherent motion, gaussian diffusion, and non-gaussian diffusion, and the relevant approximate diffusion weightings for each regime.Figure1Dshows accelerated signal decay at the level of the CSF relative to tissue for the low bvalues used here.It should be noted that since parenchymal motion in this range may be attributable to

Figure 3 Group
Figure 3

Figure 4
Figure 4 Source imaging and demographic data will be made available to investigators with appropriate Collaborative of Institutional Training Initiative Responsible Conduct of Human Research and Good Clinical Practice training upon request.

Table 1
Participant demographics and measures of suprasellar cistern motion and choroid plexus perfusion.n / a: not applicable.* signi cant at stated criteria two-sided p < 0.05.