T2 mapping of cerebrospinal fluid: 3 T versus 7 T

Object Cerebrospinal fluid (CSF) T 2 mapping can potentially be used to investigate CSF composition. A previously proposed CSF T 2–mapping method reported a T 2 difference between peripheral and ventricular CSF, and suggested that this reflected different CSF compositions. We studied the performance of this method at 7 T and evaluated the influence of partial volume and B 1 and B 0 inhomogeneity. Materials and methods T 2-preparation-based CSF T 2-mapping was performed in seven healthy volunteers at 7 and 3 T, and was compared with a single echo spin-echo sequence with various echo times. The influence of partial volume was assessed by our analyzing the longest echo times only. B 1 and B 0 maps were acquired. B 1 and B 0 dependency of the sequences was tested with a phantom. Results T 2,CSF was shorter at 7 T compared with 3 T. At 3 T, but not at 7 T, peripheral T 2,CSF was significantly shorter than ventricular T 2,CSF. Partial volume contributed to this T 2 difference, but could not fully explain it. B 1 and B 0 inhomogeneity had only a very limited effect. T 2,CSF did not depend on the voxel size, probably because of the used method to select of the regions of interest. Conclusion CSF T 2 mapping is feasible at 7 T. The shorter peripheral T 2,CSF is likely a combined effect of partial volume and CSF composition. Electronic supplementary material The online version of this article (doi:10.1007/s10334-017-0659-3) contains supplementary material, which is available to authorized users.


SE-EPI sequence
As a reference for the CSF T 2 mapping method a single echo SE-EPI sequence was used, equal to the readout used in the CSF T 2 mapping sequence, but with increasing (very) long TEs ( Figure S1). Crushers were applied before and after the refocusing pulse to crush the free induction decay signal from the refocusing pulse in case of an imperfect 180° pulse (inhomogeneous B 1 ). To minimize motion sensitivity, some modifications were made: the slice rewinder gradient was applied directly after the slice excitation pulse, while the phase encoding gradient was applied just prior to the EPI readout train.
The motion sensitivity of the crusher gradients around the 180° refocusing pulse is limited, due the relatively slow flow of CSF (around 2-4 mm/s [1]) and the short spacing between both crusher lobes.
The shortest used TE was aimed to be shorter than the first non-zero TE for the CSF T 2 mapping sequence, since the SE-EPI sequence has a higher diffusion sensitivity. The longest TE was aimed to be in the same range as the longest TE for the CSF T 2 mapping sequence. Therefore the TEs were heuristically defined according to the following formula: TE = 240 + 45·n 2 , for the n th acquisition. This resulted in the following TEs : 240, 285, 420, 645, 960, 1365, 1860, 3120, 3885, and 4740 ms. After the readout a fixed delay time (T delay ) was applied. The other parameters are specified at the description of the in vivo measurements.

Phantom experiment
A single slice was acquired with 3x3x6 mm 3

Data analysis
Data analysis was identical to the data analysis of the CSF T 2 mapping sequence.
The in vivo ROI masks were made by applying an intensity threshold (25% of the maximum intensity in the scan) to the first echo time for the SE-EPI sequence (TE SE-EPI = 0.24 s), leading to similar ROIs compared with the CSF T 2 mapping sequence.
For the partial volume assessment only TEs of at least 960 ms were taken into account in the analysis.
The minimum TE of 960 ms is shorter than the minimum TE of 1200 ms for the CSF T 2 mapping sequence, since shorter T 2 values were expected to be found for the SE-EPI sequence due to higher diffusion sensitivity.

Phantom measurements
Supplementary figure S2 shows the results of the phantom measurements for the B 1 ( Figure S2A) and B 0 gradient ( Figure S2B)  dependency, but considerably shorter T 2 values were found for increasing B 0 gradients.

In vivo measurements
In one volunteer the SE-EPI scan at 3T (both resolutions) could not be acquired due to time constraints. A total of 33 SE-EPI scans was acquired, for both field strengths, and the different resolutions. Per scan 3 fits were made, one per ROI, resulting in a total of 99 fits (36 at 3T, 63 at 7T).
Based on the strict requirement on minimum R 2 , 21 fits were excluded (8 at 3T, 13 at 7T), which is 21% of the total number of fits (22% at 3T, 21% at 7T), see Supplementary table S2 for a detailed overview.
The in vivo results for the SE-EPI scans with resolution 1×1×4 mm 3 are summarized in supplementary Figure S3. At both 3T and 7T the T 2 s measured in the periphery were significantly shorter compared to only the lateral ventricles. In all ROIs and for both field strengths, considerably shorter T 2 s were measured with the SE-EPI sequence compared to the CSF T 2 mapping sequence. The results for the other resolutions were not significantly different from the data shown here (all data, for both the SE-EPI sequence and the CSF T 2 mapping sequence, is shown in supporting tables S3, S4, and S5).

Partial volume assessment
The results for the additional analysis of peripheral CSF with the longest TEs only are shown in supplementary Figure S4. At 3T, the peripheral CSF T 2 increased with 83 ms at 3T, although this was not significant. At 7T, no difference was observed between the corrected and uncorrected peripheral CSF T 2 .
The results for this analysis for the phantom data resulted in shorter T 2 values (Supplementary figure   S5). This T 2 shortening was larger for larger B 0 gradients.

Discussion
For the SE-EPI sequence the peripheral T 2,CSF was significantly shorter than the ventricular T 2,SCF at both 3T and 7T.

B 1 and B 0 dependency
In the phantom measurements no B 1 dependency was found for the SE-EPI sequence, similar to the CSF T 2 mapping sequence, as shown in Supplementary figure S2. The measurements with an increasing through-plane B 0 gradient resulted in shorter T 2 values for larger B 0 gradients. Only a relatively small gradient was needed to find shorter T 2 values for the SE-EPI sequence, while a much larger gradient was needed to find shorter T 2 values for the CSF T 2 mapping sequence (shown in Figure 3B). This can be explained by the longer echo spacings compared to the CSF T 2 mapping sequence, which increases the sensitivity for diffusion [3]. Higher diffusion sensitivity for the SE-EPI sequence is also apparent from the analysis including only the longest TEs: shorter T 2 values were obtained for the SE-EPI sequence when only the longest echo spacings (longest TEs) were included. In contrast, the CSF T 2 mapping sequence showed similar results when only the longest TEs were included, for all B 0 gradient strengths. The difference in diffusion sensitivity of both sequences, is further illustrated by the relative difference in b-value for both sequences which can be readily computed from the echo spacing and the number of refocusing pulses [4]: the b-value of the SE-EPI sequence is 20 to 986 times larger compared with the CSF T 2 mapping sequence, for TEs 600-4800 ms, respectively. The difference in diffusion sensitivity is also visible in the different behavior of the long TE analysis on the phantom measurements. For the SE-EPI a shorter T 2 is measured if only long TEs (with stronger diffusion weighting) is used, but this is not the case for the CSF T 2 mapping sequence (compare Figure 6 with supplementary Figure S5). The difference in diffusion sensitivity is also visible in the different behavior of the long TE analysis on the phantom measurements. For the SE-EPI a shorter T 2 is measured if only long TEs (with stronger diffusion weighting) is used, but this is not the case for the CSF T 2 mapping sequence (compare Figure 6 with supplementary Figure S5).

Partial volume effects
Partial volume correction showed a relatively small T 2 increase (approximately 80 ms) at 3T, but this was not significant. At 7T the T 2 values did not change. This suggests that for the SE-EPI sequence the observed T 2 difference between ventricular and peripheral CSF is (predominantly) not caused by partial volume effects. However, the T 2 values measured with this sequence are much shorter compared to the CSF T 2 mapping sequence, probably due to its high sensitivity for diffusion and flow. If the partial volume effect in the CSF T 2 mapping sequence is indeed caused by arterial blood or relatively free water in the outer rim of the cortex, the diffusion sensitivity of the SE-EPI may decrease the sensitivity for partial volume effects from this compartment. Alternatively, it could be that the microscopic gradients around the (venous) vasculature at the brain surface induce a shorter T 2 due to the diffusion sensitivity of the SE-EPI sequence, and that this effect cannot be corrected for by choosing longer TEs.

Peripheral versus ventricular CSF T 2 and field strength dependence
The T 2 difference between peripheral and ventricular CSF was 517 ms at 3T, and 160 ms at 7T, a relative difference of 40% and 26% compared to the T 2 in the lateral ventricles. After partial volume correction, the T 2 difference decreased to 433 ms for 3T, a difference of 33% relative to the ventricular CSF T 2 , and was unchanged at 7T. Thus, a much larger T 2 difference between peripheral and ventricular CSF was found compared with the CSF T 2 mapping sequence. The T 2 difference at 3T can be partly explained by the larger B 0 gradient in the periphery, but at 7T the B 0 gradients are similar B 0 gradients were found for all ROIs. Furthermore, overall shorter T 2 values were found compared with the CSF T 2 mapping sequence. The SE-EPI sequence is more sensitive to flow and diffusion than the CSF T 2 mapping sequence, which uses multiple refocusing pulses with a fixed echo spacing that is shorter than the shortest TE used in the SE-EPI sequence. The higher flow and diffusion sensitivity of SE-EPI might (partly) cause the larger T 2 difference between peripheral and ventricular CSF. The CSF flow in the periphery is lower compared to the ventricles and cannot explain a shorter peripheral T 2 , but diffusion effects around the blood vessels on the cortex may be relatively strong. Since the ventricular walls are not covered with blood vessels, diffusion effects may be smaller in the ventricles in areas away from the choroid plexus at the base of the ventricles. Based on Kiselev, et al. [5], the level of signal dephasing due to diffusion around blood vessels is larger at 7T compared to 3T. This would imply a larger difference between peripheral and ventricular T 2 at 7T, contrary to our observations. It is conceivable, however, that the relative contribution of macroscopic field inhomogeneity to diffusion related T 2 shortening is larger at 7T than at 3T for the SE-EPI T 2 mapping, which could partially mask regional differences in microscopic field inhomogeneity. This is, however, not supported by the measured B 0 gradients from the B 0 maps that were acquired in vivo. As we used image based third order shimming at 7T, the shimming at the periphery seemed to be better than the shimming at 3T, which was linear shimming.

Limitations
The SE-EPI sequence resulted in different T 2 s compared with the CSF T 2 mapping sequence, partly due to a different sensitivity to e.g. flow and diffusion effects. The effects of these confounding factors on the observed T 2 were not studied thoroughly in this work.
Moreover, for the phantom measurements with increasing B 0 background gradient, the acquired signal decay profile showed some deviation from a single exponential decay profile due to relatively stronger diffusion effects for longer TEs. For the in vivo data however, the acquired signal decay only showed minor deviation from a single exponential decay profile. Some in vivo scans were excluded based on the minimum R 2 of 0.99, this was mainly due to motion and/or the presence of flow voids in the data.
Finally, for the SE-EPI sequence at 7T, the measured peripheral CSF T 2 was 450 ms. For the partial volume correction a minimum TE of 960 ms was used, almost double the CSF T 2 . This may decrease the sensitivity for small T 2 changes.

Supplementary tables
Supplementary