Non-Invasive Measurement of Choroid Plexus Blood Flow with Arterial Spin Labeling


 Background The choroid plexus is a major contributor to the generation of cerebrospinal fluid (CSF) and the maintenance of its electrolyte and metabolite balance. Here, we sought to characterize the blood flow dynamics of the choroid plexus using arterial spin labeling (ASL) MRI to establish ASL as a non-invasive tool for choroid plexus function and disease studies.Methods Seven healthy volunteers were imaged on a 3T MR scanner. ASL images were acquired with 12 labeling durations and post labeling delays. Regions of the choroid plexus were manually segmented on high-resolution T1 weighted images. Choroid plexus perfusion was characterized with a dynamic ASL perfusion model. Cerebral gray matter perfusion was also quantified for comparison.Results Kinetics of the ASL signal were clearly different in the choroid plexus than in gray matter. The choroid plexus has a significantly longer T1 (2.33 ± 0.30 s vs. 1.85 ± 0.10 s, p < 0.02) and a trend of lower arterial transit time (1.24±0.20 s vs.1.31 ± 0.12 s) than the gray matter. Blood flow to the choroid plexus was measured to be 39.5 ± 10.1 ml/100 g/min and 0.80 ± 0.31 ml/min integrated over the posterior lateral ventricles in both hemispheres.Conclusions Our findings suggest that ASL can provide a clinically feasible option to quantify the dynamic characteristics of choroid plexus blood flow. It also provides useful reference values of the choroid plexus perfusion. The long T1 of the choroid plexus may suggest the transport of water from arterial blood to the CSF, potentially providing a method to quantify CSF generation.


Results
Kinetics of the ASL signal were clearly different in the choroid plexus than in gray matter. The choroid plexus has a significantly longer T1 (2.33 ± 0.30 s vs. 1.85 ± 0.10 s, p < 0.02) and a trend of lower arterial transit time (1.24±0.20 s vs.1.31 ± 0.12 s) than the gray matter. Blood flow to the choroid plexus was measured to be 39.5 ± 10.1 ml/100 g/min and 0.80 ± 0.31 ml/min integrated over the posterior lateral ventricles in both hemispheres.
Conclusions Our findings suggest that ASL can provide a clinically feasible option to quantify the dynamic characteristics of choroid plexus blood flow. It also provides useful reference values of the choroid plexus perfusion. The long T1 of the choroid plexus may suggest the transport of water from arterial blood to the CSF, potentially providing a method to quantify CSF generation.

Background
Cerebrospinal fluid (CSF) plays essential roles in development, physiology, and pathology of the brain. Beyond the protection of the brain from trauma, CSF transports and regulates molecules that are essential for neuronal metabolism [1]. On one hand, CSF and blood-CSF barrier control the entry of iron, metabolites, electrolytes, and proteins into the brain [2]. On the other hand, CSF also serves as a drainage pathway of metabolic waste [3,4]. The CSF flowing through the perivascular spaces may remove amyloid-beta and other toxic molecules by actively exchanging with the interstitial fluid [5][6][7]. Recent studies further suggested that dysfunction of brain clearance mechanisms could be responsible for the accumulation of amyloid-beta and tau protein in Alzheimer's Disease (AD) [8][9][10][11].
CSF is mainly secreted by the choroid plexus, most of which is attached to the walls of the lateral ventricles. The anterior choroidal artery originating from the internal cerebral artery and the posterior choroidal artery originating from the posterior cerebral artery provide blood supply to the choroid plexus in the lateral ventricle. Distinct from the blood-brain barrier, capillaries of the choroid plexus are fenestrated and water exchanges freely between the choroid plexus stroma and the blood. The boundary between the choroid plexus and the CSF is composed of a monolayer of choroidal epithelial cells that have tight junctions between each other and form the blood-CSF barrier. Because of this unique vascular structure, the perfusion of the choroid plexus may show special characteristics compared to other regions of the brain. In addition, early studies showed that the choroid plexus blood flow was associated with CSF production [12,13]. Therefore, the characteristics of choroid plexus perfusion would be of great interest.
However, studies of the choroid plexus perfusion were rarely performed, which were partly obstructed by the shortage of safe methods. A small number of early studies were performed with radioactive tracers and animal models. Townsend et al. [14], Faraci et al. [15][16][17] and Williams et al. [18] injected radioactive microspheres and measured radiation doses of the choroid plexus on sacrificed animals. After years of medical imaging developments, only a few studies reported choroid plexus perfusion or permeability in human subjects using contrast-enhanced MRI with Gadolinium-based contrast agents (GBCA) [19,20]. However, these studies didn't provide absolute blood flow, which is challenging with GBCA because of the nonlinear dependence of signal on concentration and tissue distribution. More rapid outflow in the highly vascular choroid plexus may limit its accuracy. Concerns about the deposition of Gadolinium in tissue may also limit the use of the technique, especially in research studies with little individual benefit to subjects [21,22]. Therefore, it would be highly desirable to develop a clinically feasible and totally non-invasive method for measurement of choroid plexus perfusion.
Arterial spin labeling (ASL) perfusion MRI is an appealing non-invasive and quantitative option for the measurement of choroid plexus blood flow. In an ASL scan, the endogenous arterial blood water is labeled by radiofrequency pulses. When the labeled blood flows into the brain tissue, it results in reduced signal compared to a control scan with non-labeled blood. Through the measurement of the signal change between the control and label scans, ASL is able to quantify the blood flow using the endogenous arterial blood water signal. Therefore, ASL provides a safe technique without contrast agent administration and a clinically feasible option for characterizing choroid plexus perfusion.
Clearly visible choroid plexus signal has been observable in brain ASL imaging studies for many years, such as the Fig. 10 of Dai et al. [23], the Fig. 1 of Mutsaerts et al. [24], and the Fig. 9 of Amukotuwa et al. [25]. The ASL signal of the choroid plexus has been also observed in patients with Alzheimer's Disease [26]. In a recent restingstate functional study with ASL, our results further indicated that the ASL signal of the choroid plexus didn't participate in flow fluctuations present in brain networks [27]. However, sources of high ASL signal in the choroid plexus were not discussed in the above work.
Recently, we reported the feasibility to detect choroid plexus perfusion using ASL in abstract form [28,29], but the blood flow quantification methods were not fully developed. Johnson et al. [30] implemented this method to evaluate perfusion changes after angiogenesis, but they used a single delay ASL experiment which couldn't provide the dynamic characteristics of choroid plexus perfusion.
In this work, we sought to establish ASL as a tool to quantify the dynamic characteristics of choroid plexus perfusion. The perfusion characterization can serve as a preliminary reference for studies of physiological and pathological modulation of choroid plexus function.

Subject Recruitment
Seven healthy adult volunteers (5 females, mean age 35±18 years) were recruited for this study. All subjects were imaged following a protocol approved by the institutional committee on clinical investigations. Written informed consent was obtained from all subjects prior to MRI scanning.

MR Imaging Protocol
MR imaging was performed on a GE 3 T Discovery MR750 scanner (GE Healthcare, Chicago, IL) using a body coil for transmission and a 32-channel head array for signal reception. High-resolution T1 weighted images and 12 different ASL images were acquired in each subject.
First, whole brain T1-weighted images were acquired to segment the choroid plexus. A 3D inversion-prepared Spoiled gradient echo sequence was used to acquire high-resolution T1-weighted images. Scan parameters included: inversion time (TI) of 450 ms, echo time (TE) of 3.24 ms, and repetition time (TR) of 8.24 ms. Parallel imaging was used to reduce the scan time to approximately 4 min 20seconds. The spatial resolution was 0.94 mm x 0.94 mm x 1 mm.
Next, 12 ASL images were acquired to quantify the dynamic characteristics of the choroid plexus perfusion. The label/control of each ASL acquisition was performed using the unbalanced pseudo-continuous arterial spin labeling (pCASL) technique [23]. Arterial blood was labeled with repeated Hann RF pulses and static tissue signal was suppressed by background suppression pulses interleaved with ASL label/control pulses [31]. ASL images were acquired with a stack-of-spiral Fast Spin Echo acquisition. Each spiral arm used a readout of 8 ms with bandwidth 62.5 kHz and three interleaved arms were used. For a field-of-view 220 mm, this setting resulted in an in-plane resolution of 4 mm x 4 mm. 32 slices were acquired with slice thickness 4 mm. In each ASL scan, four pairs of control and label scans were acquired to improve the signal-to-noise ratio. TE was 28.4 ms and TR was from 4.24 s to 9.19 s. At the end of each ASL scan, one proton density reference image was acquired with a saturation time of 2 s. The total time of ASL scans was about 40 min.
The above ASL acquisition was performed at 12 different observation times (OT), including blood labeling durations (LD) and post-labeling delays (PLD) listed in Table 1. This design was able to track the inflow and washout process of the labeled arterial blood, similar to a previous work [32]. The minimum PLD was chosen to be 0.7 s, which was assumed to be shorter than the arterial transit time [31]. At such short delays, a part of the labeled blood was still in the large arteries when the ASL images were acquired. To quantify the perfusion signals more accurately, a set of vessel suppression pulses [31] was implemented right before the acquisition to remove the labeled blood signals in the large arteries.

MR Image Processing
ASL images were reconstructed by the manufacturer's online algorithm. Pair-wise subtraction between label and control data was performed followed by summing of the ASL difference images. ASL images were reconstructed by a 1D Fourier transform along the slice direction and a gridding operator and 2D Fourier transform in each slice. The proton density reference images were reconstructed in the same manner. At each OT, the ASL and proton density images were scaled with the same factor to maximize the visual dynamic arrange, when they were exported to DICOM files.
Images were preprocessed using Statistical Parametric Mapping (SPM12) software (Wellcome Trust Centre for Neuroimaging, London) with default settings. The 12 ASL images were realigned for each subject. Since the proton density image was acquired immediately after the ASL image, we assumed negligible movement between the ASL image and the following proton density image. Therefore, the proton density image was aligned using the same parameters obtained in the alignment of corresponding ASL image.
The 12 realigned ASL images were co-registered to the high-resolution T1 weighted images for each subject to avoid potential partial volume effects, because the ASL images have lower resolution than the T1-weighted images. T1 weighted images were segmented into the gray matter, the white matter, the CSF, the skull and the fat regions using the segmentation tool from SPM12. The mean of the 12 realigned ASL images was registered with the gray matter probability map. The individual ASL and proton density images were then transformed to the T1-weighted image space using the above generated transformation matrix.
The choroid plexus regions were manually segmented based on the T1-weighted images using ITK-SNAP [33]. The most readily recognized choroid plexus regions were selected in the posterior part of the lateral ventricles, Fig. 1.
The manual segmentation was performed on axial, coronal and sagittal planes and was checked in a 3D view. The gray matter regions were selected by a threshold of 0.5 on the gray matter probability map.

Perfusion Characteristics Quantification
The blood flow, arterial transit time and T1 of the gray matter and the choroid plexus regions was quantified using a standard dynamic ASL model [31].
where is the ASL signal. Pixel-wise ASL signal was used to obtain the spatial distribution of perfusion. The mean ASL signal of the choroid plexus or the gray matter regions was used to calculate average perfusion. is the labeling efficiency, is the inversion efficiency ratio of the arterial blood [23]; is the background suppression efficiency [34]. is the longitudinal relaxation rate of the arterial blood [35]. is the post-labeling delay and is the labeling duration, Table 1. In this study, a constrained, pixel-wise nonlinear least squares fit (Scipy1.4, optimize.least_squares) was used to quantify three parameters: is the blood flow, constrained by a lower bound of 10 ml/100 g/min; is the arterial transit time, constrained by a lower bound of 0.7s [31]; and is the longitudinal relaxation rate of the labeled water after it is transported from the arterial blood to local tissue, constrained by a lower bound of 1s was used empirically. The goodness of fit was measured by a coefficient of determination R 2 .
is the equilibrium magnetization of arterial blood calculated from the proton density image.
where is the mean signal of the gray matter region and is calculated from the proton density image corresponding to the ASL image at each OT. is the brain blood partition coefficient [31], which is used to approximate the unknown choroid plexus blood partition coefficient. is the saturation recovery time in the proton density image scan and is the longitudinal relaxation rate of the gray matter [31,36]. The same was also used to quantify the perfusion of the gray matter and the choroid plexus, which was equivalent to using the gray matter as reference tissue [37].
Since the choroid plexus and the CSF intermingle at a microscopic level, it is not possible to exclude a partial volume of CSF in the choroid plexus voxels at any achievable MR imaging resolution. Here we choose to report two measures: One, a perfusion value, f, in ml/100 gm/min that includes CSF partial volume unavoidably but can be readily compared to ASL measured perfusion in the brain, and the second, the total blood flow of choroid plexus volume in the left and right posterior lateral ventricles, F, in ml/min. The total flow can be calculated, assuming zero blood flow in CSF, according to Where F is the total blood flow (ml/min), represents the partial volume effect in the nth voxel, which is the volume percentile of choroid plexus in the voxel. is the volume of each voxel and is the choroid plexus volume in the nth voxel. is the choroid plexus density, assuming to be the same as the brain parenchyma [38]. is the actual blood flow (ml/100g/min) in the choroid plexus. On the contrary, is voxel-wise blood flow measured in ASL experiments, including partial volume effects. is the average blood flow of the choroid plexus, which contains the partial volume effect. is the number of voxels in the choroid plexus.

Statistical Analysis
The mean values and standard deviations of the blood flow, arterial transit time and T1 were calculated across subjects. Non-parametric and paired comparisons of the above parameters were compared using the Wilcoxon sign-rank method (Scipy1.4, stats.wilcoxon) between the gray matter and the choroid plexus regions.

Results
ASL signal in the choroid plexus and the brain was readily detected and qualitatively consistent with expected behavior. The relative perfusion images, , show the perfusion evolution across multiple OTs, Fig. 2. In the first six OTs, when longer blood labeling duration was used, the ASL signal of the gray matter increased. This was because more labeled blood flowed into the brain tissue and had stronger effects than the ASL signal decay. In the later six OTs, when the PLD further increased, the ASL signal reduced when the labeled blood signal recovered towards the control blood value.
The red arrow in Fig. 2 highlights the prominent signal of the choroid plexus. In the early perfusion phases, the signal of choroid plexus was readily noticeable, but did not stand out compared to the gray matter and the thalamus. In the late perfusion phases, the ASL signal of the choroid plexus became more obvious. This indicates that the ASL signal of the choroid plexus had a slower signal decay compared to surrounding tissue. Our results at t8 and t9 showed a similar contrast between the choroid plexus and other regions as previous ASL studies with similar post-labeling delays [23][24][25].
The voxel-wised distributions of the blood flow, arterial transit time, T1 values, and coefficients of determination from all 7 subjects are shown in Fig. 4. The blood flow and the T1 showed distinguishable distributions between the gray matter and the choroid plexus, although voxel-vise perfusion contained higher noise than the average regional perfusion below.
In the average perfusion characteristics, the choroid plexus regions showed significantly lower blood flow (p < 0.02) and significantly longer T1 (p < 0.02) than the gray matter across subjects, Table 2. The choroid plexus showed a trend of decreased arterial transit time, but not significantly. Both the gray matter and choroid plexus regions showed high coefficients of determination () in the perfusion model fitting, which indicated high accuracy in ASL model parameters. Across the subjects, the volume of choroid plexus in the lateral ventricles was 2.08±0.56 ml, which was consistent with a previous study [19], and the total blood flow was 0.80±0.31 ml/min.

Discussion
In this work, we quantified the dynamic characteristics of the choroid plexus perfusion using ASL. This work demonstrated the value of ASL as a noninvasive tool to evaluate the choroid plexus perfusion. The blood flow and T1 in the choroid plexus are significantly different from the gray matter.
This work provided the first non-invasive measures of the choroid plexus perfusion characteristics. Average blood flow of 39.5 ml/100 g/min and total blood flow of 0.8 ml/min was found at choroid plexus in the lateral ventricle. While choroid plexus perfusion was characterized using large-size tracers and ex-vivo animal models in the early studies [15][16][17], they were not able to provide spatial distributions of blood flow or dynamic perfusion information, such as arterial transit time. Our study clearly demonstrated the kinetics of perfusion labeled signal and its spatial distribution. The blood flow measured in this work was in agreement with Johnson et al.'s ASL study [30], although studies based on radioactive tracers and animal models reported higher blood flow. Total blood flow of the choroid plexus showed about 38% variance, which could be from the inter-subject variation, including the choroid plexus volume variation and the average blood flow variation. Our results showed about 25% variation for either, which was consistent with the previous studies [19,30].
Total blood flow represents the function of the choroid plexus more accurately than blood flow maps. Because of the low resolution of ASL images and the sheet-like structure of the choroid plexus, ASL images have considerable partial volume effects, which result in a reduced ASL signal at each voxel. Since volume percentile of the choroid plexus is unknown and can be different at each voxel, voxel-wise blood flow is not accurate. However, the total ASL signal across the choroid plexus region is not influenced by partial volume effects, so the total blood flow provides the choroid plexus blood flow more accurately.
The longer T1 at the choroid plexus may be evocative of CSF generation. A slow ASL signal decay was noted at the choroid plexus compared to the gray matter. Table 2 shows that the T1 of the choroid plexus was longer than the gray matter and the arterial blood. Since CSF is the only brain tissue with a T1 longer than 2 s, this result may suggest that the labeled water transfers from arterial blood to CSF. Therefore, ASL may provide a feasible method to detect CSF generation, but further investigation is required. For example, recent work showed that increases of choroid plexus weight coexisted with decreases of CSF generation [39]. This may suggest a complex relationship between choroid plexus perfusion, mass, and CSF generation.
Several limitations of this study should be noted. First, the cohort size of this work was relatively small. The variances of the blood flow and the choroid plexus volume result in a large variation in blood flow estimation. More reliable reference characteristics of the choroid plexus perfusion can be obtained with increased sample size and age distribution of controls. Second, this study focused on the choroid plexus in the lateral ventricles. Third, the choroid plexus region selection on the ASL images has limited accuracy. The choroid plexus region was transferred from T1 weighted images to the ASL domain based on the volume registration between ASL and the gray matter. However, this may not be an effective registration for the choroid plexus region and the registration quality can be limited by the ASL resolution. A surface constrained registration [40] and high-resolution ASL images [41] may provide improved accuracy. Fourth, the shorter arterial transit reflected the unique vascular supply of the choroid plexus. However, current work cannot exclude the impact of the different feeding arteries of the gray matter and the choroid plexus. Arterial transit time was reported to be shorter in the deep gray matter [42]. Therefore, further evaluations are required to understand the shorter transit time at the choroid plexus. Finally, a number of parameters (e.g. relaxation times) were approximated or assumed in the quantification. However, we anticipate that the errors from these approximations will not affect the qualitative conclusions of our work (e.g., T1 and arterial transit time), Further studies aimed at detailed quantification models for choroid plexus will be merited if the measure proves useful.

Conclusion
This work demonstrated arterial spin labeling MRI as a feasible and safe tool for choroid plexus perfusion studies. The dynamic ASL method is readily feasible to use and can help studies of choroid plexus lesions and function. This work also provided useful reference values of the choroid plexus perfusion characteristics. Compared to the gray matter, the choroid plexus has significantly lower blood flow and longer apparent T1, which may indicate transport of the label into CSF by diffusion or CSF generation.

List Of Abbreviations
List of abbreviations

Declarations
Ethics approval and consent to participate All subjects were imaged following a protocol approved by the institutional committee on clinical investigations. Written informed consent was obtained from all subjects prior to MRI scanning.

Consent for publication
Available on request.
Availability of data and materials The datasets generated and/or analyzed during the current study are not publicly available due IRB regulation, but are available from the corresponding author on reasonable request.
Competing interests David C. Alsop is an inventor of pseudo-continuous ASL and receives postmarket royalties through his institution from GE Healthcare, Philips Healthcare, Siemens Healthineers, Hitachi Medical, and Animage LLC. He also receives research support from GE Healthcare.

Funding
This work was supported in part by the National Institute of Mental Health through R01MH080729 and the Alzheimer's Association through AARF-18-566347 Authors' contributions LZ designed the study, collected and analyzed the data and draft the manuscript. MT was a contributor to the data analysis. WD was a major contributor to the quantification model. DP was a contributor to the study design. DA was a major contributor to the study design, data analysis and manuscript. All authors read and approved the final manuscript.

Figure 2
Relative perfusion images at multiple OTs. The choroid plexus regions become more prominent comparing to the gray matter in the scans with the longer blood labeling and longer post labeling delay. The choroid plexus region is highlighted by the red arrow.  Histograms of voxel-wise ASL measures in choroid plexus and gray matter of 7 volunteers.