Quantification of T₁ and T₂ of subarachnoid CSF: implications for water exchange between CSF and brain tissues

Abstract

Purpose

To quantify the T₁ and T₂ values of CSF in the subarachnoid space (SAS) at 3.0T and interpret them in the context of water exchange between CSF and brain tissues.

Methods

CSF T₁ was measured using inversion recovery and CSF T₂ was assessed using T₂-preparation. T₁ and T₂ values in the SAS were compared to those in the frontal-horns of lateral ventricles, which have less brain-CSF exchange. Phantom experiments were performed to examine whether there were spatial variations in T₁ and T₂ that were unrelated to brain-CSF exchange. Simulations were conducted to investigate the relationship between the brain-CSF exchange rate and the apparent T₁ and T₂ values of SAS CSF.

Results

The CSF T₁ and T₂ values were 4308.7±146.9ms and 1885.5±67.9ms, respectively, in the SAS and were 4454.0±187.9ms and 2372.9±72.0ms in the frontal-horns. The SAS CSF had shorter T₁ (P=0.006) and T₂ (P<0.0001) than CSF in the frontal-horns. Phantom experiments showed negligible (<6ms for T₁; <1ms for T₂) spatial variations in T₁ and T₂, suggesting that the T₁ and T₂ differences between SAS and frontal-horns were largely attributed to physiological reasons. Simulations revealed that faster brain-CSF exchange rates lead to shorter apparent T₁ and T₂ of SAS CSF. However, the experimentally observed T₂ difference between SAS and frontal-horns was greater than that attributable to typical exchange effect, suggesting that the T₂ shortening in SAS may reflect a combined effect of exchange and deoxyhemoglobin susceptibility.

Conclusion

Quantification of SAS CSF relaxation times may be useful to assess the brain-CSF exchange.

1. INTRODUCTION

The brain’s waste clearance system is essential for the brain’s homeostasis, and the dysfunction of this system is thought to underlie various brain diseases such as Alzheimer’s disease.^1,2 The cerebrospinal fluid (CSF) is believed to play an important role in the brain’s waste clearance system. According to the classical view of CSF dynamics, CSF is produced by the choroid plexus in the cerebral ventricles. Subsequently, the CSF enters the subarachnoid space (SAS) through the foramen of Luschka and Magendie, and drains into the dural venous sinuses via arachnoid villi and granulations.³ Recent studies have suggested an alternative pathway for CSF flow, in which the CSF in the SAS enters the brain parenchyma through the perivascular spaces of penetrating arteries and extensively exchanges with the brain tissues. This exchange facilitates the removal of waste products (such as neurotoxic amyloid β and tau) from the brain parenchyma.^4,5 For example, the glymphatic system hypothesis proposes that CSF in the periarterial space exchanges with the interstitial fluid of the brain tissues via the aquaporin-4 channels located on the endfeet of astrocytes. This rapid exchange generates a convective flow that flushes the waste products from the arterial side towards the venous side. Subsequently, the waste products enter the perivenous space and are eventually removed from the brain by lymphatic vessels.^4,6 Therefore, there is a growing interest in developing imaging techniques to evaluate the water exchange between CSF and brain tissues.

MRI-based assessment of the brain-CSF exchange was first conducted using contrast-enhanced MRI with intrathecal injection of gadolinium contrast agent in both animal models⁷ and human patients.^8,9 However, intrathecal gadolinium injection bears the risk of gadolinium encephalopathy,¹⁰ and has not been approved for routine clinical use. Non-contrast, non-invasive MRI techniques to measure the brain-CSF exchange are therefore desirable for wide clinical applications. A few groups have used arterial spin labeling (ASL) MRI to measure the brain-CSF exchange at the choroid plexus, which is the major site of CSF production and also forms the blood–CSF barrier.^11-14 These studies mainly focused on the choroid plexus in the lateral ventricles. Li et al. assessed brain-CSF exchange around the ventricles using magnetic transfer indirect spin labeling or phase alternate labeling in rodent models.^15,16 Petitclerc et al., using multi-delay ASL with ultra-long echo time (TE), observed active water exchange between SAS CSF and nearby blood or grey matter in human subjects, although the scan time was relatively long (~1 hour).¹⁷

Since the brain tissues have much shorter T₁ and T₂ than CSF,^18,19 the water exchange between SAS CSF and nearby brain tissues is expected to result in lower apparent relaxation times of CSF in the SAS compared to regions with less exchange. This difference in CSF apparent relaxation times may be used to assess the brain-CSF water exchange rate. In this work, we quantified T₁ and T₂ of SAS CSF using MRI pulse sequences that were designed to minimize partial volume contamination from brain tissues. The CSF T₁ and T₂ values in the SAS were compared to a reference region: the frontal-horns of lateral ventricles, which have less exchange due to the lack of choroid plexus.²⁰ We also conducted phantom experiments to understand the mechanisms of the T₁ and T₂ differences. Furthermore, we proposed an exchange model and performed simulations to relate the brain-CSF water exchange rate to the apparent relaxation times of SAS CSF.

2. METHODS

2.1. Study 1: In vivo Experiments

Seven healthy adults (3 females and 4 males, 24.3±3.1 years old) were scanned on a 3T Prisma scanner (Siemens Healthcare, Erlangen, Germany). The study protocol was approved by the Johns Hopkins University Institutional Review Board. Each subject gave written informed consent before participating in this study.

To measure the CSF T₁, we employed an inversion recovery sequence with spin-echo echo-planar-imaging readout. An ultra-long TE was used to minimize the partial volume contamination from brain tissues. A fixed recovery time of 8000ms was applied at the end of each TR. The sequence for CSF T₂ measurement was similar to that for T₁ measurement, except that the inversion recovery module was replaced with T₂-preparation (T₂-prep). To make the sequences insensitive to the flow of CSF and robust against B₀ and B₁ inhomogeneities, we employed a non-selective adiabatic inversion pulse for T₁ mapping and used non-selective composite refocusing pulses with MLEV phase cycling in T₂-prep for T₂ mapping.²¹ Details of the pulse sequences are described in Supporting Information and Figure S1.

The T₁ mapping sequence used the following parameters: 2D single-slice, field-of-view (FoV)=218×218mm², in-plane resolution=1.7×1.7mm², slice thickness=5mm, TE=700ms, 10 inversion times (TIs): 50, 500, 1100, 1800, 2600, 3600, 4900, 6800, 10000 and 15000ms, 2 averages and scan duration=4.8min. The imaging slice was approximately parallel to the anterior-commissure-to-posterior-commissure line and 10mm above the posterior commissure.

The T₂ mapping sequence used the same parameters as the T₁ mapping sequence, except for the following: the T₂ mapping scan used five effective TEs (eTEs) for the T₂-prep module: 0, 640, 1280, 1960, and 2560ms. Each eTE was acquired with 2 averages and the total scan time was 2min. To examine whether the spacing between refocusing pulses (τ_CPMG) affected CSF T₂ values, we measured T₂ using four different τ_CPMG values: 20, 40, 80, and 160ms, by varying the number of refocusing pulses while keeping the eTE value the same. The order of the τ_CPMG values was pseudo-randomized across participants.

To evaluate the reproducibility of CSF T₁ and T₂ measurements, each of the above sequences was repeated three times.

2.2. Study 2: Phantom Experiments

The interpretation of the in vivo data relies on the comparison of relaxation times between the SAS and the frontal-horns of lateral ventricles. To examine whether the sequences themselves could result in spatially dependent variations in T₁ and T₂ values (unrelated to brain-CSF water exchange), we performed a phantom experiment using a cylindric container filled with 0.6% agarose gel (Life Technologies, Carlsbad, California, United States). The phantom had a diameter of 136mm, mimicking the size of the adult brain. The T₁ and T₂ mapping sequences used the same parameters as in Study 1, except: (1) A minimal TE of 46ms was used in both T₁ and T₂ mapping to maximize the signal-to-noise ratio; (2) For T₂ mapping, five eTEs of 0, 80, 160, 240 and 320ms were used with τ_CPMG=20ms.

2.3. Study 3: Simulation

To investigate the relationship between the apparent T₁ of SAS CSF with the brain-CSF water exchange rate ($k$, in unit of min⁻¹), we simulated the magnetization of CSF spins in the SAS using the following exchange model based on modified Bloch equations:

$$\frac{{dM}_{z,c1}}{dt}=\frac{M_{0,c1}-M_{z,c1}}{T_{1,c1}}+k(\lambda M_{z,c2}-M_{z,c1})$$ [1]

$$\frac{{dM}_{z,c2}}{dt}=\frac{M_{0,c2}-M_{z,c2}}{T_{1,c2}}+k(\frac{M_{z,c1}}{\lambda }-M_{z,c2})$$ [2]

where the subscripts “c1” and “c2” denote the brain tissue and SAS compartments, respectively. T_1,c1 represents the brain tissue T₁ and was assumed to be 1209ms.¹⁸ The CSF T_1,c2 used the mean T₁ in the frontal-horns of lateral ventricles measured in Study 1. λ is the brain/CSF partition coefficient and was assumed to be 0.8 based on the proton density of brain tissue and CSF reported in the literature.²² The apparent T₁ of SAS CSF was then fitted using the simulated M_z,c2 values for $k$ ranging from 0 to 3 min⁻¹, with a step size of 0.01 min⁻¹. Additionally, we assessed the influence of the assumed tissue T₁ value on the calculated apparent T₁ of SAS CSF (please refer to the Supporting Information for details).

Similarly, we also examined the association of $k$ with the apparent T₂ of SAS CSF using the following exchange model:

$$\frac{{dM}_{xy,c1}}{dt}=-\frac{M_{xy,c1}}{T_{2,c1}}+k(\lambda M_{xy,c2}-M_{xy,c1})$$ [3]

$$\frac{{dM}_{xy,c2}}{dt}=-\frac{M_{xy,c2}}{T_{2,c2}}+k(\frac{M_{xy,c1}}{\lambda }-M_{xy,c2})$$ [4]

We assumed brain tissue T_2,c1 to be 88ms.¹⁸ The CSF T_2,c2 used the mean T₂ in the frontal-horns of lateral ventricles measured with τ_CPMG=20ms in Study 1. The apparent T₂ of SAS CSF was then fitted using the simulated M_xy,c2 values for $k$ ranging from 0 to 3 min⁻¹, with a step size of 0.01 min⁻¹.

Note that the models described by Equations 1-4 depict the water exchange between CSF and nearby brain tissues during T₁ or T₂ relaxation. In regions with minimal exchange, such as the frontal-horns of the lateral ventricles, k ≈ 0, and Equations 1-4 reduce to simple Bloch equations.

2.4. Data processing

All data processing used in-house MATLAB (Mathworks, Natick, MA) scripts.

In Study 1, voxel-wise CSF T₁ and T₂ maps were obtained by exponential fitting of the TI-dependent or TE-dependent data (see Supporting Information for details). We also quantified CSF T₁ and T₂ values in two manually defined regions-of-interest (ROIs): ROI1 encompassed the frontal-horns of lateral ventricles and ROI2 included all voxels in the SAS (Figure S2). As a metric for reproducibility, the coefficients-of-variation (CoVs) of ROI T₁ and T₂ values across the three repetitions were calculated.

In Study 2, T₁ and T₂ maps of the phantom were computed. To examine whether there were spatially dependent biases in T₁ and T₂ quantification, we compared the T₁ and T₂ values in a center ROI with those in a peripheral ROI in the phantom. The center ROI was a circle (radius=10 voxels), while the peripheral ROI was an annulus (width=3 voxels).

2.5. Statistical Analysis

All statistical analyses were conducted using MATLAB. A two-tailed P<0.05 was considered statistically significant.

In Study 1, the ROI T₁ or T₂ values were first averaged across the 3 repetitions for each subject. Paired t-test was then used to compare the T₁ values between the SAS ROI and the frontal-horn ROI. Two-way repeated-measures analysis-of-variance (ANOVA) followed by Tukey multiple comparison tests was used to examine whether the T₂ values differ between ROIs or among τ_CPMG values.

In addition, we evaluated the spatial dependence of CSF T₁ within the SAS using linear-mixed-effect (LME) analysis, in which the dependent variable was the T₁ value of a voxel in the SAS in the T₁ map, and the independent variable was the normalized anterior-posterior coordinate (Y) of this voxel (Y=0 for the anteriormost voxel in the SAS, Y=1 for the posteriormost voxel in the SAS). In this LME analysis, each voxel of the T₁ map for each repetition and each subject was considered as an individual data point. The LME model considered subject-specific random effects for both slope and intercept. We also examined the dependence of SAS T₁ on the normalized right-left coordinate (X=0 for the rightmost voxel in the SAS, X=1 for the leftmost voxel in the SAS) using a similar LME model. These LME analyses were repeated for SAS CSF T₂ measured with τ_CPMG=20ms.

3. RESULTS

3.1. Study 1

Figure 1A shows inversion recovery images at different TIs. The locations of the ROIs have also been illustrated. Figure 1B shows the signal fitting of the two ROIs. Figure 1C displays the quantitative T₁ map. Corresponding results for T₂ mapping are shown in Figure 1D-1F.

Figure 1.

Data of a representative subject. (A) Images acquired with all TIs for CSF T₁ mapping. Yellow arrows point to the ROIs. Due to the long TE used, only CSF signal is visible. (B) Magnitude of the CSF signals in the ROIs as functions of TIs. The dashed curves indicate the fitted T₁ relaxation curves. (C) CSF T₁ map. (D) Images acquired with all eTEs for T₂ mapping. (E) ROI CSF signals as functions of eTEs. The dashed curves represent the fitted T₂ decay curves. (F) CSF T₂ map.

Table 1 summarizes the ROI T₁ and T₂ values across all subjects. The CoVs of T₁ and T₂ measurements were less than 4% for both ROIs, suggesting that the parametric estimations were reliable.

Table 1.

ROI T₁ and T₂ values in the in vivo experiments.

T1	ROI1: frontal-horn	ROI2: SAS
T1 (ms)	4454.0 ± 187.9	4308.7 ± 146.9
CoV (%)	3.8 ± 2.5	3.8 ± 3.0

T2	τCPMG	ROI1: frontal-horn	ROI2: SAS
T2 (ms)	20ms	2372.9 ± 72.0	1885.5 ± 67.9
	40ms	2353.5 ± 72.3	1867.8 ± 60.1
	80ms	2293.1 ± 55.7	1853.9 ± 71.8
	160ms	2248.2 ± 123.0	1768.1 ± 99.0
CoV (%)	20ms	3.2 ± 2.1	1.8 ± 1.1
	40ms	3.1 ± 3.4	2.3 ± 0.8
	80ms	3.0 ± 1.6	2.5 ± 1.2
	160ms	3.5 ± 2.9	2.6 ± 1.2

Comparison of T₁ values showed that the SAS exhibited a shorter T₁ than the frontal-horns of lateral ventricles (P=0.006). For T₂, ANOVA revealed significant effects of both ROI (P<0.0001) and τ_CPMG (P=0.001) on T₂. The SAS had a shorter T₂ compared to the frontal-horns and a shorter τ_CPMG was associated with a longer measured T₂. There was no significant interaction effect between ROI and τ_CPMG (P=0.48).

We also observed an anterior-posterior “gradient” in SAS T₁ and T₂ values. Figure 2 shows the dependence of SAS CSF T₁ and T₂ values on the voxel location along the anterior-posterior direction. There was a decrease in both T₁ (T₁ = −147.3*Y+4327.8, P=0.02, Figure 2B) and T₂ (T₂ = −239.0*Y+1973.4, P<0.0001, Figure 2C) from the anterior to the posterior portions of the SAS. We found no dependence of either T₁ (P=0.94) or T₂ (P=0.61) on the voxel location along the right-left direction (Figure S3).

Figure 2.

Spatial dependence of SAS T₁ and T₂ values along the anterior-posterior direction. (A) Illustration of the normalized X (right-left, X=0 for the rightmost voxel in the SAS, X=1 for the leftmost voxel in the SAS) and Y (anterior-posterior, Y=0 for the anteriormost voxel in the SAS, Y=1 for posteriormost voxel in the SAS) coordinates. The CSF image is displayed in the radiological convention. (B) Scatter plot showing the relationship between SAS T₁ and the normalized Y coordinate. Each dot represents one voxel in the T₁ map. Data from different subjects are indicated by different colors. The black solid line represents the fitted line. (C) Scatter plot demonstrating the association between SAS T₂ and the normalized Y coordinate.

3.2. Study 2

Figure 3 shows the T₁ and T₂ maps of the gel phantom. Both maps are spatially uniform. Their source images are displayed in Figure S4. Figure 3C displays the center and peripheral ROIs. We found negligible differences in T₁ (center T₁ = 2985.7ms vs. peripheral T₁ = 2980.6ms) and T₂ (center T₂ = 180.9ms vs. peripheral T₂ = 180.2ms) between the two ROIs, suggesting that the T₁ and T₂ differences observed in the in vivo data were not due to technical reasons such as B₁ or B₀ inhomogeneities.

Figure 3.

Phantom experiment results. (A) T₁ map of the phantom. (B) T₂ map. (C) The center (blue dots) and peripheral (red dots) ROIs.

3.3. Study 3

Figure 4A shows the simulation results of the relationship between the apparent CSF T₁ and the brain-CSF exchange rate $k$. It can be seen that a higher $k$ is associated with a lower apparent T₁ in SAS. This is because faster exchange will allow more CSF spins to recover at the T₁ of brain tissues, thereby accelerating the longitudinal relaxation process. The average SAS apparent T₁ of 4308.7ms measured in Study 1 corresponded to $k=0.94{\min }^{-1}$. When dividing the SAS along the anterior-posterior axis, the anteriormost SAS had a longer T₁ of 4327.8ms and a lower $k=0.81{\min }^{-1}$; while the posteriormost SAS had a shorter T₁ of 4180.5ms and a higher $k=1.83{\min }^{-1}$.

Figure 4.

Numerical simulation results. (A) Apparent SAS T₁ and (B) apparent SAS T₂ as a function of brain-CSF exchange rate. Higher exchange rate leads to lower apparent T₁ and T₂ values.

Our simulation results also demonstrated the influence of tissue T₁ value on the SAS CSF T₁ value. As depicted in Figure S5, for a specific k level, a shorter tissue T₁ led to a lower SAS CSF T₁. Furthermore, the effect of tissue T₁ variations on SAS CSF T₁ was more pronounced with larger k values. On the other hand, if the objective was to estimate k based on the SAS CSF T₁, assuming a shorter tissue T₁ yielded a lower estimated k value, as illustrated in Figure S6.

Figure 4B illustrates the relationship between apparent CSF T₂ and $k$. As shown, a larger $k$ corresponds to a shorter apparent T₂.

4. DISCUSSION

In this work, we measured the CSF T₁ and T₂ values in the human brain and observed that CSF in the SAS had lower apparent T₁ and T₂ values than CSF in the frontal-horns of lateral ventricles. In addition, within the SAS, CSF T₁ and T₂ values decreased from the anterior SAS to the posterior SAS. Phantom experiments showed that the pulse sequences had negligible spatial biases in T₁ and T₂ quantification, suggesting that the regional differences in CSF T₁ and T₂ values were largely attributed to physiological reasons. Simulations revealed that a faster brain-CSF water exchange rate was associated with shorter apparent T₁ and T₂ values of SAS CSF.

The SAS and the connected perivascular space have received growing attention due to their important role in the brain’s waste clearance system.⁴ However, few studies have specifically quantified the SAS CSF T₁ and T₂ values.^23-25 The major technical challenges include partial volume contamination from nearby brain tissues and the ultra-long T₁ and T₂ of CSF. In this work, we employed an ultra-long TE (700ms) to minimize the brain tissue signal, and used large ranges of TIs (50ms to 15000mss) and eTEs (0ms to 2560ms) to provide accurate CSF T₁ and T₂ estimations. To the best of our knowledge, our study is the first to quantify SAS CSF T₁ at 3T. Zaharchuk et al. quantified CSF T₁ at 1.5T, and also observed that the SAS CSF had lower T₁ than the ventricular CSF.²³ For T₂, Qin²⁴ and Spijkerman et al.²⁵ reported CSF T₂ values of ~1600ms for SAS and ~2000ms for lateral ventricles, which were slightly lower than the T₂ values measured in this study. Both Qin²⁴ and Spijkerman et al.²⁵ found significantly lower CSF T₂ values in the SAS than those in the lateral ventricles, consistent with our observations.

Our simulations suggested that both T₁ and T₂ of SAS CSF may be affected by brain-CSF water exchange. However, in our experiments, we found that the extent of the T₂ difference between the SAS and the frontal-horns of lateral ventricles (by ~20%) was larger than that of the T₁ difference (by ~3%). A possible explanation is that, in addition to the exchange effects, the susceptibility effect of the deoxyhemoglobin in the venous network at the brain surface may also reduce the apparent T₂ of SAS CSF (by increasing the microscopic magnetic field inhomogeneity experienced by the nearby SAS CSF spins).²⁶ Therefore, for the purpose of assessing brain-CSF exchange, CSF T₁ may be a more specific measure. Based on our T₁ data, the brain-CSF exchange rate, $k$, is approximately 0.94min⁻¹. This rate is much slower than the water exchange rate across the blood-brain barrier (over 100min⁻¹ in human adults^27-31), but is consistent with the findings by Petitclerc et al., who reported that the water exchange time from the brain tissue (or blood) to the CSF was 60s in the SAS.¹⁷

Our simulation results also demonstrated the impact of tissue T₁ on the SAS CSF T₁ value. Because the SAS is close to the cortical gray matter, in the simulations, we assumed the tissue compartment to have a gray matter T₁. We used the T₁ value of 1209ms, which was reported by Lu et al. using an inversion recovery sequence similar to ours.¹⁸ It is worth noting that the gray matter T₁ values at 3.0T range from 900ms to 1800ms in the literature.³² These variations in literature T₁ values are partly due to methodological differences. For example, T₁ values measured using variable flip angle approaches tend to be longer than those estimated using inversion recovery methods.³³ Note that the intended use of our proposed technique is to use SAS CSF T₁ as a surrogate biomarker for brain diseases, rather than to quantify the absolute k value.

One interesting finding of the present study is that SAS CSF T₁ and T₂ values decreased from the anterior to the posterior end. This would suggest that the occipital lobe had a higher brain-CSF water exchange rate than the frontal lobe. To the best of our knowledge, no prior studies have compared the brain-CSF water exchange rate between different brain regions. We hypothesize that this regional difference in SAS T₁ and T₂ values may be related to the supine position of the subjects. In this position, the brain may shift towards the posterior end due to gravity, allowing more frequent water exchange between CSF and the occipital brain tissues. To test this hypothesis, future work should measure the SAS CSF T₁ and T₂ values in different body positions (e.g., supine vs. prone; left-facing vs. right-facing). Another possible explanation for this anterior-posterior gradient in SAS CSF relaxation times is that the occipital gray matter may have shorter T₁ and T₂ compared to frontal gray matter, as our simulations showed the influence of tissue T₁ on the SAS CSF T₁. However, few studies have compared the relaxation times of gray matter among different cortical regions, and the results are mixed.^18,34

Besides the brain-CSF exchange, the apparent relaxation times of the SAS CSF may also be affected by other physiological factors. For example, two previous studies showed that during hyperoxic gas inhalation, SAS CSF T₁ was reduced but the lateral ventricular CSF T₁ was unchanged.^23,35 Therefore, it has been suggested that the lower T₁ in the SAS might be partly caused by a higher oxygen partial pressure.^23,35 To examine whether the T₁ and T₂ differences between SAS and the frontal-horns of lateral ventricles are truly caused by brain-CSF exchange, it is important to apply our techniques to physiological or pathological conditions that have been shown to alter the brain-CSF exchange rate. For example, Xie et al. demonstrated in mice that natural sleep or anesthesia increased the convective exchange of CSF with interstitial fluid of brain tissues.³⁶ In human, Ringstad et al. showed that patients with idiopathic normal pressure hydrocephalus had reduced glymphatic clearance.^8,9 Further investigations in these physiological or pathological conditions will be the goals of our future studies.

In this work, our primary focus was to investigate the water exchange between SAS CSF and the nearby brain tissues. However, it is important to acknowledge the substantial brain-CSF exchange that occurs at the choroid plexus, which serves as the brain-CSF barrier. Considering the shorter relaxation times of the choroid plexus compared to CSF,¹¹ it is possible to assess CSF production activity at the choroid plexus by quantifying the relaxation times of neighboring CSF in comparison with CSF in distant regions. It is worth noting that, given the net production of CSF at the choroid plexus, the exchange models described by Equations 1-4 need to be adjusted accordingly. Specifically, the exchange rate from the choroid plexus to CSF should be larger than the rate from CSF back to the choroid plexus. Another important structure related to CSF flow dynamics is the arachnoid villi and granulations, which serve as sites for CSF absorption into the cerebral venous system. However, it should be mentioned that because our sequences used an ultra-long TE (700ms), the spins that have been absorbed into the venous system through arachnoid villi and granulations will not contribute to the measured MRI signals due to the short T₂ of venous blood (typically 50-80ms).^37,38

There are a few limitations of the present study. First, our sample size is relatively small (N=7) and we only studied young healthy adults. Future work should include a larger cohort and patients with brain diseases, such as hydrocephalus. Second, in this proof-of-principle study, we acquired only a single relatively thick slice (5mm). To expand the spatial coverage and decrease the voxel size, future pulse sequence development could integrate simultaneous multi-slice acquisition, gradient-and-spin-echo acquisition, or 3D high-resolution turbo-spin-echo readout. Third, we only examined the variations in SAS CSF T₁ and T₂ values along the anterior-posterior and right-left axes, but we did not evaluate these spatial variations using more sophisticated brain parcellations, such as brain lobes, arterial perfusion territories and watershed areas. Exploring these aspects will be the focus of our future studies.

5. CONCLUSIONS

This study demonstrated that the apparent CSF T₁ and T₂ values in the SAS were lower than those in the frontal-horns of lateral ventricles. Further comparison of the in vivo data with phantom and simulation studies suggested that SAS T₁ shortening may be predominantly attributed to brain-CSF exchange. These findings suggest that quantification of SAS CSF relaxation times may be useful to assess the fluid homeostasis in the brain.

Supplementary Material

Supinfo

Figure S1: Diagrams of pulse sequences for CSF T₁ and T₂ mapping. (A) The T₁ mapping sequence. (B) The T₂ mapping sequence.

Figure S2: Illustration of the ROIs on the CSF mask. ROI1 include all voxels in the frontal-horns of the lateral ventricles (blue dots) and ROI2 include all voxels in the SAS (red dots).

Figure S3: Spatial dependence of SAS T₁ and T₂ values along the right-left direction. (A) Illustration of the normalized X (right-left, X=0 for the rightmost voxel in the SAS, X=1 for the leftmost voxel in the SAS) and Y (anterior-posterior, Y=0 for the anteriormost voxel in the SAS, Y=1 for posteriormost voxel in the SAS) coordinates. The CSF image is displayed in the radiological convention. (B) Scatter plot between SAS T₁ and the normalized X coordinate. Each dot represents one voxel in the T₁ map. Data from different subjects are represented by different colors. The black solid line indicates the fitted line. (C) Scatter plot between SAS T₂ and the normalized X coordinate.

Figure S4: Images of the phantom. (A) Images with all TIs for T₁ mapping. (B) Images with all eTEs for T₂ mapping.

Figure S5: Dependency of SAS CSF T₁ on tissue T₁ at various k levels.

Figure S6: Relationship between the assumed tissue T₁ value and the estimated exchange rate k that corresponds to the measured SAS T₁ value.