Magn Reson Mater Phy (2018) 31:415–424 https://doi.org/10.1007/s10334-017-0659-3 RESEARCH ARTICLE T mapping of cerebrospinal fluid: 3 T versus 7 T 1 2 1 1 Jolanda M. Spijkerman · Esben T. Petersen · Jeroen Hendrikse · Peter Luijten · Jaco J. M. Zwanenburg Received: 6 July 2017 / Revised: 22 September 2017 / Accepted: 13 October 2017 / Published online: 6 November 2017 © The Author(s) 2017. This article is an open access publication Abstract Conclusion CSF T mapping is feasible at 7 T. The shorter Object Cerebrospinal fluid (CSF) T mapping can poten- peripheral T is likely a combined effect of partial vol - 2 2,CSF tially be used to investigate CSF composition. A previously ume and CSF composition. proposed CSF T –mapping method reported a T difference 2 2 between peripheral and ventricular CSF, and suggested that Keywords Cerebrospinal fluid · T2 relaxation · 7 T · 3 T · this reflected different CSF compositions. We studied the MRI performance of this method at 7 T and evaluated the influ - ence of partial volume and B and B inhomogeneity. 1 0 Materials and methods T -preparation-based CSF T -map- Introduction 2 2 ping was performed in seven healthy volunteers at 7 and 3 T, and was compared with a single echo spin-echo sequence Qin  proposed a fast 3-T MRI method to map the volume with various echo times. The influence of partial volume and T of cerebrospinal fluid (CSF) in the brain. A striking was assessed by our analyzing the longest echo times only. finding with this method was the observation of a shorter T B and B maps were acquired. B and B dependency of the in the peripheral CSF compared with the T of the CSF in 1 0 1 0 2 sequences was tested with a phantom. the lateral ventricles. Qin suggested that this T difference Results T was shorter at 7 T compared with 3 T. At is caused by differences in CSF composition between both 2,CSF 3 T, but not at 7 T, peripheral T was significantly shorter areas, implying that CSF T (T ) can be used as a nonin- 2,CSF 2 2,CSF than ventricular T . Partial volume contributed to this T vasive biomarker for CSF composition. This would be highly 2,CSF 2 difference, but could not fully explain it. B and B inhomo- relevant in the light of the recent attention to the clearance 1 0 geneity had only a very limited effect. T did not depend of brain waste products, in which CSF is involved [2–5]. 2,CSF on the voxel size, probably because of the used method to A method that can noninvasively assess CSF composition select of the regions of interest. would provide a noninvasive window on the brain clearance system, with great potential for applications in studying dis- eases related to dementia such as Alzheimer’s disease and Electronic supplementary material The online version of this article (doi:10.1007/s10334-017-0659-3) contains supplementary cerebral small vessel disease. If T is indeed useful as 2,CSF material, which is available to authorized users. a functional marker of the brain clearance system, it could be studied next to other advanced imaging markers of early * Jolanda M. Spijkerman brain damage such as microbleeds, microinfarcts, and hip- email@example.com pocampus subfield volumes and atrophy. As many of these Department of Radiology, University Medical Center advanced markers are acquired at 7 T [6–8], it is desirable Utrecht, HP E 01.132, P.O.Box 85500, 3508 GA Utrecht, to implement and evaluate CSF T mapping at 7 T as well. The Netherlands At 7 T, B inhomogeneity is considerable and may Danish Research Centre for Magnetic Resonance, Centre influence the T mapping results, despite the relative B 2 1 for Functional and Diagnostic Imaging and Research, insensitivity of the used CSF T mapping method. Even Copenhagen University Hospital Hvidovre, 2650 Hvidovre, 2 Denmark at 3 T, considerable B inhomogeneity in the brain can Vol.:(0123456789) 1 3 416 Magn Reson Mater Phy (2018) 31:415–424 be observed . Also, when T is measured in peripheral Materials and methods CSF, partial volume effects with tissue cannot be avoided. So, we hypothesized that these partial volume effects and Sequence B imperfections can explain the previously observed T 1 2 differences. De Vis et al. [ 10] obtained a rough estimation The CSF T mapping sequence used in this research is based of the influence of partial volume effects on the estimated on T preparation, and has been described elsewhere [1, 10]. T by scanning with two different resolutions. The For this study the method was further extended to improve 2,CSF higher resolution resulted in longer T times, suggesting the fit reliability of the long T times by implementation of a role for partial volume effects. Because the influence of longer refocusing pulse trains, yielding longer TEs. Briefly, B inhomogeneity and partial volume effects is not clear the sequence consists of four parts (Fig. 1). First, a set of yet, it remains uncertain to what extent T can be used four nonselective water suppression enhanced through T 2,CSF 1 to assess the composition of CSF. effects (WET) pulses are applied for saturation to prevent In this work we studied the performance of Qin’s slice history effects. These pulses are optimized for satura- CSF T mapping method at 7 T. The specific goals were tion of free water (applicable for T between 3 and 6 s); the to investigate the influence of B and B imperfections pulse angles are 156°, 71°, 109°, and 90° . Second, a 1 0 on the estimated T , to assess the influence of partial delay time (T ) follows, where T relaxation occurs, fol- 2,CSF delay 1 volume effects, and to evaluate to what extent the previ- lowed by crusher gradients. Third, T preparation is applied, ously observed difference in T between periphery and consisting of a nonselective 90° pulse, a set of 4, 8, 16, or 32 2,CSF ventricles can be explained by B , B , and partial volume nonselective refocusing pulses (R) according to the Malcolm 1 0 effects. B and B sensitivity was investigated with phan- Levitt (MLEV) phase cycling scheme, and a nonselective 1 0 tom measurements, and by comparison of the method for −90° pulse with a crusher gradient to crush any remaining 7 and 3 T in healthy humans. Partial volume effects were transverse magnetization . Each refocusing pulse R is a ◦ ◦ ◦ estimated by removal of the influence of partial volume composite pulse, consisting of 90 , 180 , 90 rectangular x y x with tissue, through selection of only the last (longest) ◦ ◦ ◦ pulses (or the inverse R ∶ 90 , 180 , 90 ). The duration -x -y -x echo times (TEs). Also, scanning was done at different of a single refocusing pulse was 2.6 ms. T relaxation occurs resolutions. 2 Presaturation Delay T₂ preparation Acquisition (WET) 180° 90° -90° 90° Readout RF Tdelay TE T2-prep crushing Fig. 1 Cerebrospinal fluid T mapping pulse sequence. The sequence tion occurs during TE . To perform T mapping the sequence 2 T2-prep 2 consists of four parts: water suppression enhanced through T effects was repeated, while the number of refocusing pulses (and therefore (WET) presaturation, a fixed delay with duration T , a T prepara- TE ) was increased for a fixed interpulse delay τ. The applied RF delay 2 T2-prep tion module with Malcolm Levitt (MLEV) phase cycling, and a spin pulses are shown on the RF axis, and the applied gradients are shown echo echo planar imaging (SE-EPI) image acquisition. T relaxa- on the frequency (F), phase (P), and slice (S) encoding axes 1 3 Magn Reson Mater Phy (2018) 31:415–424 417 during TE , which is determined by the number of refo- pronounced B inhomogeneity . The acquisition with T2-prep 1 cusing pulses and the spacing between the centers of the the highest B gradient yielded images free of FID artifacts, refocusing pulses (τ). To achieve long TE times, τ was which allowed us to study the B dependency of the CSF T2-prep 1 chosen as 150 ms. This resulted in TE durations of T mapping sequence. B and B maps were acquired, with T2-prep 2 0 1 600, 1200, 2400, and 4800 ms. Also, one scan was acquired use of identical resolution and FOV as for the T mapping without any refocusing pulses; this scan was not used in data acquisitions. The B map was obtained with a gradient echo analysis. The fourth part of the sequence is a single-shot 2D sequence with two different TEs (1.64 and 2.64 ms). The B spin echo (SE) echo planar imaging (EPI) readout. During mapping sequence was based on the actual flip angle method the EPI train, T decay also occurs; this can, however, be , with first repetition time (TR) of 40 ms, second TR of regarded as a constant factor, and was therefore disregarded 160 ms, TE of 0.96 ms, and a flip angle of 50°. in the analysis. Although the refocusing train in the T preparation is In vivo measurements relatively insensitive to B inhomogeneities because of the MLEV phase cycling scheme, the 90° rectangular pulses In vivo experiments were performed at both 7 and 3 T to before and after the train may fail in the case of B devia- test the feasibility of CSF T mapping at 7 T, to further 1 2 tions. Consequently, we hypothesized that a fraction of the assess the sensitivity to B inhomogeneities, and to explore magnetization may be unaffected by the T -preparation the influence of partial volume effects. Seven healthy vol- module, which could have a relatively large impact on the unteers (three men, mean age 34 ± 11 years, age range measured signal in the case of partial volume effects. Also, 21–54 years) participated in this study. Informed consent in the case of imperfect B , T -weighted stimulated echoes was given by all volunteers in accordance with the require- 1 1 may influence the T measurements, although this effect ments of the Institutional Review Board of the University is expected to be relatively small because of the long T Medical Center Utrecht (Utrecht, Netherlands). All volun- of CSF (4.4 s ). Therefore, T mapping with a single teers were scanned with both a 3-T Philips Achieva scanner echo SE-EPI sequence with various TEs was used as a truly with an eight-channel head coil (Philips Healthcare, Best, B -insensitive reference (shown in the electronic supplemen- Netherlands) and the 7-T scanner that was also used for the tal material). phantom study. The CSF T mapping scans were acquired Nonselective pulses were used where possible to mini- in a single coronal slice, planned through both the lateral mize motion sensitivity. Consequently, only the excitation ventricles and the fourth ventricle (Fig. 2a). The scanning pulse of the SE-EPI readout was selective. parameters are summarized in Table 1. The fixed T was delay 15 s, and TR ranged between 20 and 25 s depending on Phantom measurements TE . Other parameters were as follows: SENSE factor T2-prep 2.3 in the left–right direction and FOV of 240 × 240 mm . Phantom measurements were performed to test the B and Because of the long TR, the specific absorption rate (SAR) B sensitivity. The experiments were performed with a 7-T remained well within the specific absorption rate limits, also Philips Achieva scanner (Philips Medical Systems, Best, at 7 T. No additional methods were used to correct the scans Netherlands) with a 32-channel head coil (Nova Medical, for eddy currents. The low bandwidth of the scan may cause Wilmington, MA, USA), with a tap water phantom at room distortions in areas with poor shimming, such as the nasal temperature. The phantom size was 200 × 95 × 20 mm . cavities. However, shimming was good in the selected coro- The phantom was scanned with the same TEs as used nal slice. Also B and B maps were acquired. 0 1 for the in vivo experiments. A single slice was acquired with 3 × 3 × 6 mm resolution, field of view (FOV) of Data analysis 240 × 96 mm , T of 15 s, [which is more than three delay times the T of CSF (4.4 s )], and sensitivity encod- Phantom ing (SENSE)  (with SENSE factor 1, meaning that the coil sensitivities of the receive coils were used for optimal The resulting T estimates were analyzed as a function of coil combination without imaging acceleration). A series of B and B . B sensitivity was assessed with use of the scans 1 0 1 T maps with increasing through-plane B gradients were with the highest B gradient strength (0.5 mT/m), where 2 0 0 acquired to study the effect of diffusion for B of 100%. The no FID artifacts were present. The B range present in the 1 1 following through-plane B gradient strengths were applied scan was used. On the basis of B in each voxel, the voxels 0 1 by addition of this strength to the linear shim term in the user were sorted over eight bins of 5% B , leading to B bins 1 1 interface: 0, 0.05, 0.1, 0.2, 0.3, and 0.5 mT/m. The phantom ranging from (85 ± 2.5)% to (120 ± 2.5)%. The signal was appeared sensitive to free induction decay (FID) artifacts, averaged over each B bin, and T values were fitted over 1 2 because of the relatively large volume of water and more this averaged signal. B sensitivity was assessed with the 1 3 418 Magn Reson Mater Phy (2018) 31:415–424 Fig. 2 a Planning of the cerebrospinal fluid (CSF) T mapping scans, and 4.8 s), shown with equal intensity scaling. c The region of inter- through the lateral ventricles and the fourth ventricle. b CSF T map- est masks used: the periphery (PER; white), the lateral ventricles ping scans at 7 T for increasing echo times (TE ) (0.6, 1.2, 2.4, (LAT; yellow), and the fourth ventricle (FOU; red) T2-prep Table 1 Scan parameters used 3 Resolution (mm ) TE (ms) TE (ms) EPI factor Bandwidth (phase/ Scan readout T2-prep for the in vivo experiments frequency) (Hz/ duration for the cerebrospinal fluid T voxel) (min) mapping sequence (based on T preparation) 3 T 1 × 1 × 4 133 0–4800 105 8.1/961 2:59 3 × 3 × 6 42 0–4800 67 28.2/2308 2:59 7 T 1 × 1 × 2 127 0–4800 107 8.4/1082 3:04 1 × 1 × 4 126 0–4800 107 8.4/1082 3:04 3 × 3 × 6 23 0–4800 37 55.1/2642 3:04 EPI echo planar imaging, TE echo time, TE echo time of T -preparation T2-prep 2 The TE values used were 0, 600, 1200, 2400, and 4800 ms. T2-prep various applied B gradient strengths (0, 0.05, 0.1, 0.2, 0.3, volume and motion sensitivity. Erosion of the peripheral and 0.5 mT/m). In each scan, only voxels with B between ROIs was not feasible. 97.5% and 102.5% were included. An additional intensity The signal was averaged over each ROI, and T values threshold was applied on the scan with the longest TE were fitted over this averaged signal, with use of a single T2-prep to minimize the influence of artifacts in the lower B gradi- exponential decay model. Also mean B and B values were 0 0 1 ent scans. This threshold was set at 75% of the maximum determined for each ROI. To minimize the influence of, for intensity for the longest TE. The signal of all voxels included example, motion or partial volume effects on the data analy - was averaged over each scan, and T values were fitted over sis, only fit results with R of 0.99 or higher were considered. this averaged signal. Partial volume assessment In vivo In the peripheral CSF an additional assessment of the influ- Three regions of interest (ROIs) were defined on the acquired ence of partial volume was made by our performing a partial in vivo scans: the lateral ventricles, the fourth ventricle, and volume correction. Only TE values of at least 1200 ms T2-prep peripheral CSF. The ROI masks were made by our applying (excluding the shortest TE of 600 ms) were taken into T2-prep an intensity threshold to the first TE (TE = 0.6 s). The account in the analysis. Thereby, maximal nulling of, for T2-prep intensity threshold was set at 25% of the maximum inten- example, tissue signal was achieved, since the T values of sity in the image. Figure 2b and c shows the acquired CSF tissue are below 100 ms [17, 18], about ten times shorter T mapping scan at 7 T with a resolution of 1 × 1 × 4 mm than the minimal TE used. The analysis with only 2 T2-prep at all TEs for one volunteer, and the ROIs used. Conserva- the last TE times was also performed on the phantom T2-prep tive ROIs were used in the ventricles by our eroding the scans to check for any systematic errors for all B gradient intensity-based ROIs with one voxel to minimize both partial strengths applied and B between 97.5% and 102.5%. 1 3 Magn Reson Mater Phy (2018) 31:415–424 419 All data analysis was performed in MATLAB (version 7 T), which corresponds to 13% of the total number of fits 2015B, The MathWorks, Natick, MA, USA). IBM SPSS Sta- (10% at 3 T, 16% at 7 T); see Table 2 for a detailed overview. tistics (version 21.0) was used for statistical analysis. Median The in vivo results for the scans with a resolution of T values and full ranges are reported. Wilcoxon signed- 1 × 1 × 4 mm are summarized in Fig. 4. The results for the 2,CSF rank tests (significance level p < 0.05) were used to compare other resolutions were not significantly different from the CSF T values in the lateral and fourth ventricles with those data shown here (all data are shown in Tables S3, S4, S5). in the periphery to explore the observed T differences. Although T differences between the resolutions were not 2 2 significant, in most cases the shortest T times were observed for the largest voxel sizes. Results At 7 T significantly shorter T times were found compared with at 3 T. At 3 T the T times measured in the periphery Phantom measurements were significantly shorter than those measured in the lateral and fourth ventricles. The T times measured at 7 T were Figure 3 shows the results of the phantom measurements not significantly different between the three ROIs. At 3 T for the B dependency (Fig. 3a) and B gradient depend- the median B in the periphery was 85% (range 79–90%), 1 0 1 ency (Fig. 3b). The CSF T mapping sequence measured while in the lateral and fourth ventricles it was 109% (ranges a T of 1.71 s (95% confidence interval 1.66–1.76 s) for B of (100 ± 2.5)% and B gradient strength of 0 mT/m. 1 0 Table 2 Number of scans acquired and T fits performed, and the The sequence showed only minor B sensitivity (assessed number of excluded T fits per region of interest in the scans with B gradient strength of 0.5 mT/m), with T 0 2 ranging from 1.41 s (95% confidence interval 1.38–1.43 s) Resolution (mm ) Scans Fits Excluded fits at B of (85 ± 2.5)% to 1.49 s (95% confidence interval Lateral Fourth Periphery 1.40–1.57 s) at B of (105 ± 2.5)%. Also minor B gradient ventricles ventricle 1 0 dependency was observed. 3 T 1 × 1 × 4 7 21 0 1 0 In vivo measurements 3 × 3 × 6 7 21 2 1 0 Total 14 42 2 2 0 Thirty-five CSF T mapping scans were acquired, for both 7 T field strengths and the different resolutions. Per scan, three 1 × 1 × 2 7 21 0 1 0 fits were made, one per ROI, resulting in a total of 105 fits 1 × 1 × 4 7 21 1 3 0 (42 at 3 T, 63 at 7 T). On the basis of the strict requirement 3 × 3 × 6 7 21 1 4 0 on minimum R , 14 fits were excluded (four at 3 T, ten at Total 21 63 2 8 0 A B 2.0 2.0 1.6 1.6 1.2 1.2 0.8 0.8 0.4 0.4 0 0 85 90 95 100 105 110 115 120 00.1 0.20.3 0.40.5 B₁ [%] B₀ gradient [mT/m] Fig. 3 Results of the phantom measurements for the B (a) and B the B gradient was most apparent for the highest B gradient. The B 1 0 0 0 1 gradient dependency (b) showing the fitted T for different B values dependency was determined with a B gradient strength of 0.5 mT/m 2 1 0 and through-plane B gradient strengths, respectively. The error bars to avoid free induction decay (FID) artifacts in regions with B devi- 0 1 show the 95% confidence interval of the fitted T . The cerebrospinal ating from 100%. For a B gradient of 0.2 mT/m, the confidence 2 0 fluid T mapping sequence shows only minor sensitivity to B and to interval was greater because of FID artifacts 2 1 the through-plane B gradient (and thus to diffusion). The effect of 1 3 T₂ [s] T₂ [s] 420 Magn Reson Mater Phy (2018) 31:415–424 Lateral ventricles Fourth ventricle Periphery A B C 2.0 0.30 1.5 0.20 1.0 0.10 0.5 0 0 3T 7T 3T 7T 3T 7T Fig. 4 In vivo results: T (a), B (b), and B gradient (c) for the three different regions of interest. Outliers are represented by a square. Signifi- 2 1 0 cant differences in measured T were found between the periphery and the lateral and fourth ventricles at 3 T (indicated by an asterisk) 106–112% and 103–114%). The median B gradient in the The partial volume correction resulted in longer T times, 0 2 periphery was 0.13 mT/m (range 0.07–0.38 mT/m), while in with a significant increase of 118 ms at both 3 and 7 T. At the lateral and fourth ventricles it was 0.02 and 0.03 mT/m, 7 T the corrected peripheral CSF T was quite similar to respectively (range 0.02–0.06 mT/m and 0.01–0.03 mT/m, the ventricular T (1.01 s vs 1.05 s), while at 3 T the mean respectively). At 7 T, lower B values were observed in peripheral T was still approximately 200 ms shorter than 1 2 the periphery and the fourth ventricle [median 86% (range the ventricular T times [1.79 s (range 1.49–1.82 s) vs 2.03 s 75–94%) and 93% (range 62–101%), respectively], and (range 1.73–2.16 s), p = 0.02]. The results for this analy- higher B values were observed in the lateral ventricles sis of the phantom data are shown in Fig. 6. Both analysis [median 111% (range 109–116)]. Similar B gradients were methods (including all TEs or only the longest TEs) resulted observed in the three ROIs [median 0.07 mT/m (range in similar T values, indicating no systematic errors in the 0.03–0.10 mT/m), 0.06 mT/m (range 0.04–0.08 mT/m), and additional analysis with only the longest TEs. 0.06 mT/m (range 0.05–0.09 mT/m), for the lateral ventri- cles, the fourth ventricle, and the periphery, respectively]. Discussion Partial volume assessment In this research we have shown the feasibility of CSF T Figure 5 shows the results for the additional analysis of mapping at 7 T with a dedicated CSF T mapping sequence peripheral CSF to assess the influence of partial volume. based on T preparation, which was initially developed at 3T 7T A B 2.0 1.3 1.9 1.2 1.8 1.1 1.7 1.0 1.6 0.9 1.5 0.8 1.4 0.7 All TE’s Long TE’s All TE’s Long TE’s Fig. 5 T values of peripheral cerebrospinal fluid resulting from the increase in T can be observed. The asterisk indicates a significant 2 2 use of only the longest echo times (TEs) compared with the original difference with the original analysis (including all TEs) analysis. Outliers are represented by a square. At both 3 and 7 T an 1 3 T₂ [s] T₂ [s] B₁ [%] T₂ [s] B₀ gradient [mT/m] Magn Reson Mater Phy (2018) 31:415–424 421 through-plane B gradient showed only limited B gra- 0 0 2.0 dient dependency, except for the highest B gradient (0.5 mT/m), as shown in Fig. 3b. In the in vivo measure- ments, the B gradient was similar between the periphery 1.6 and the ventricles at 7 T, and differed by a maximum of 0.38 mT/m (median B gradient was 0.13 mT/m) at 3 T 1.2 0 (Fig. 4c). This difference in B homogeneity between 3 and 7 T is probably due to different shimming techniques: 0.8 image-based third-order shimming was used at 7 T, and linear shimming was used at 3 T. The low sensitivity to 0.4 All TEs B gradient shows that the T mapping sequence is rela- 0 2 Long TEs tively insensitive to diffusion. It is not likely that B gradi- 00.1 0.20.3 0.40.5 ents due to imperfect shimming contributed considerably B₀ gradient [mT/m] to the observed difference in T between periphery and ventricles. Fig. 6 T values of the phantom, resulting from the use of only the longest echo times (TEs; orange) compared with the original analysis (blue). Both analyses result in similar T values Partial volume effects 3 T. We investigated the sensitivity of this sequence for the The different resolutions used at both field strengths did not influence of B , diffusion (B gradient), and partial volume yield considerably different T values (Tables S3, S4, S5), 1 0 effects. The sequence appeared to be relatively insensitive although there is a trend of longer measured ventricular T to B and B inhomogeneity. Partial volume effects tend to times for smaller voxel sizes at 3 T, similarly to what was 1 0 lower the observed T values at the periphery. T was found by to De Vis et al. . As the ventricular ROIs were 2 2,CSF considerably shorter at 7 T than at 3 T in all three ROIs. The eroded, the voxels at the edges, where more partial volume peripheral T was significantly shorter than the ventricu- is expected, were discarded. For the periphery, however, 2,CSF lar T at 3 T (but not at 7 T). erosion was not feasible because of the thin shape of the 2,CSF The peripheral T increased considerably on partial ROI. Moreover, the ROI definition was based on an inten- 2,CSF volume correction, as obtained from analysis of long TEs sity threshold, which depends on the CSF fraction in each (more than ten times the tissue T ). The partial volume cor- voxel. Since the total subarachnoidal CSF volume is quite rection for the SE-EPI sequence, which was used as a rela- small, and distributed over a relatively large area , partial tively B -insensitive reference (data shown in the electronic volume is probably present in all peripheral ROIs, indepen- supplementary material), did not significantly increase T dently of the voxel sizes used in this work. values, although the SE-EPI sequence showed an even larger The role of partial volume effects regarding the measured T difference between the periphery and ventricles. The peripheral T was investigated by use of the longest TEs 2,CSF ventricular T values measured with the CSF T mapping only (Fig. 5) to maximally remove the influence of partial 2 2 sequence at 3 T match with T values found in literature [1, volume. It could seem unexpected that the use of the late 10, 19, 20]. Given the results from our measurements and TEs reveals a considerable partial volume effect, since the analysis, we believe that the observed T difference between first TE is already relatively long compared with the T2-prep the ventricular and peripheral CSF could be partly due to tissue T . The T of gray matter is approximately 90 ms at 2 2 physiological differences. However, the different results for 3 T [18, 22] and 55 ms at 7 T [18, 23], while the first TE was different sequences and field strengths and the confounding 600 ms. However, it is possible that partial volume occurs influence of partial volume effects will make it challeng- with a compartment with a relatively long T in the cerebral ing to accurately isolate and quantify any true physiologi- cortex, like arterial blood (T around 150 ms at 3 T [24, 25]) cal effect from confounders. This will hamper applications or the outer rim of the cortex (unknown but long T , greater in research focusing on in vivo evaluation of the (regional) than 100 ms, at 7 T ). At the shortest TE (600 ms), T2prep composition of CSF. the signal of arterial blood has decayed to 2%. However, in the case of small partial volume fractions of CSF in the B and B dependency periphery, this could still have a considerable influence on 1 0 the measured T . The outer layer of the cerebral cortex (layer In the phantom measurements only minor B dependency I) may have a long T because it contains almost no neuronal was found for the CSF T mapping sequence, as shown cell bodies, and many glial cells instead, similarly to gliotic in Fig. 3a. Also, the measurements with an increasing lesions, which also have a long T . 1 3 T₂ [s] 422 Magn Reson Mater Phy (2018) 31:415–424 Peripheral versus ventricular CSF T and field strength Implications dependence Before CSF T mapping can be used as a parameter to study De Vis et al.  found a T difference of 609 ± 133 ms diseases such as cerebral small vessel disease, several uncer- between the periphery and the ventricles at 3 T, and Qin tainties need to be resolved. It is not yet clear to what extent found a T difference of 420 ± 155 ms at 3 T. Also in the T difference between ventricular and peripheral CSF 2 2 this work a shorter T was measured in the periphery reflects physiological differences in CSF composition. The 2,CSF compared with the ventricles, as shown in Fig. 4a. This CSF T mapping sequence shows a much smaller T dif- 2 2 T difference is larger at 3 T than at 7 T: the T difference ference compared with SE-EPI, while the difference also 2 2 is 365 ms at 3 T and 161 ms at 7 T, which corresponds to varies with field strength. Overall, the T difference between differences of 18% and 15% relative to the T in the lat- peripheral and ventricular CSF could (partly) be explained eral ventricles for 3 and 7 T, respectively. Partial volume by (a combination of) physiological differences. The possi- correction, which led to a peripheral CSF T increase of bility that the shorter peripheral T is entirely caused by an 2 2 118 ms at both field strengths (Fig. 5), resulted in remain- artifact, like B gradients caused by imperfect shimming and/ ing T differences of 247 ms and only 43 ms for 3 and 7 T, or partial volume effects between tissue, blood, and CSF, respectively. These correspond to a T difference of 12% seems unlikely. and 4% relative to the ventricular CSF T for 3 and 7 T, Care should be taken when one is interpreting T meas- 2 2 respectively. A relatively larger T difference was found urements of CSF, and more work is necessary to find the true when a SE-EPI sequence was used, and remained largely explanations for the T differences between 3 and 7 T and unchanged after partial volume correction (data shown in between the peripheral and ventricular CSF at 3 T. the electronic supplementary material). These results indicate a true T difference between Limitations peripheral and ventricular CSF. A potential physiologi- cal explanation for this observed T difference could The major limitation of this work is that it is an observa- be sought in differences in, for example, in the levels tional study, which limits the extent to which underlying of O , protein, and/or glucose, since these substrates are mechanisms causing the observations can be identified. known to decrease T [20, 27–29]. However, relatively Despite our efforts to separate the effects of partial volume large concentration differences are necessary to bridge and true physiological differences, it remains uncertain to the difference between peripheral and ventricular T . what extent the observed shorter peripheral T is due to 2,CSF 2,CSF So although differences in CSF composition may partly different CSF compositions. cause the observed T difference, it seems unlikely that Furthermore, the statistical power of this study was lim- these are the only contributor. ited by the low number of volunteers combined with the The shorter in vivo CSF T at 7 T than at 3 T (Fig. 4a) stringent R criterion, which resulted in a relatively large is in line with published in vivo measurements by Daoust dropout of ROIs. et al. . However, Daoust et al. suggested that the T Moreover, only macroscopic B gradients could be deter- 2 0 of CSF is not field strength dependent, but that residual mined in the in vivo scans, and the magnitude of micro- field gradients cause errors in in vivo measurements at scopic, subvoxel B gradients remains unknown. higher field strengths. If the T measurements are strongly Finally, no in vitro CSF sample was used to validate the dependent on residual gradients, one might expect that the in vivo measurements. In vitro CSF is prone to changes in, T difference between periphery and ventricles observed for example, O content, compared with in vivo CSF, which 2 2 at 3 T is also largely due to residual field gradients, such may induce T differences between in vitro and in vivo CSF. as B gradients. However, the CSF T mapping sequence 0 2 used in our study showed negligible B gradient depend- ency for the measured T up to 0.3 mT/m, while the Conclusion observed B gradients in the brain were between 0.07 and 0.38 mT/m, and on average well below 0.20 mT/m. CSF T mapping with a dedicated sequence is feasible at The limited diffusion sensitivity of the CSF T mapping both 3 and 7 T, and yields shorter CSF T times at 7 T com- 2 2 sequence is also visible from the results of the long TE pared with 3 T. At 3 T, shorter T times were found for analysis on the phantom measurements. The measured T peripheral CSF compared with ventricular CSF; at 7 T this remained unchanged when only long TEs (with stronger effect was much smaller. Partial volume effects can partly diffusion weighting) were used (see Fig. 6). explain this T difference, but a physiological contribution to the difference in T between ventricular and peripheral CSF is possible. The different results for different sequences 1 3 Magn Reson Mater Phy (2018) 31:415–424 423 microbleed detection on 7 T MR imaging: reliability and effects of and field strengths, and the confounding influence of par - image processing. Am J Neuroradiol 34:E61–E64 tial volume, will make it challenging to accurately isolate 7. van Veluw SJ, Biessels GJ, Luijten PR, Zwanenburg JJM (2015) and quantify any true physiological effect for applications Assessing cortical cerebral microinfarcts on high resolution MR in research focusing on in vivo evaluation of the (regional) images. J Vis Exp 105:e53125 8. Wisse LEM, Biessels GJ, Heringa SM, Kuijf HJ, Koek DL, Lui- composition of CSF. jten PR, Geerlings MI (2014) Hippocampal subfield volumes at 7 T in early Alzheimer’s disease and normal aging. Neurobiol Aging Funding The research leading to these results received funding from 35:2039–2045 the European Research Council (ERC) under the European Union’s 9. Saekho S, Boada FE, Noll DC, Stenger VA (2005) Small tip angle Seventh Framework Programme (2007-2013)/ERC grant agreement no. three-dimensional tailored radiofrequency slab-select pulse for 337333 (SmallVesselMRI), and the European Union’s Horizon 2020 reduced B1 inhomogeneity at 3 T. Magn Reson Med 53:479–484 program/ERC grant agreement no. 637024 (HEARTOFSTROKE) and 10. De Vis JB, Zwanenburg JJ, van der Kleij LA, Spijkerman JM, Bies- under grant agreement no. 666881 (SVDs@target). sels GJ, Hendrikse J, Petersen ET (2015) Cerebrospinal fluid volu- metric MRI mapping as a simple measurement for evaluating brain Authors’ contribution JMS: Protocol/project development, Data atrophy. Eur Radiol 26:1254–1262 collection, Data analysis. ETP: Protocol/project development, Data 11. Golay X, Petersen ET, Hui F (2005) Pulsed star labeling of arterial analysis. JH: Protocol/project development, Data analysis. PL: Pro- regions (PULSAR): a robust regional perfusion technique for high tocol/project development. JJMZ: Protocol/project development, Data field imaging. Magn Reson Med 53:15–21 collection, Data analysis. 12. Levitt MH, Freeman R, Frenkiel T (1982) Broadband heteronuclear decoupling. J Magn Reson 47:328–330 Compliance with ethical standards 13. Rooney WD, Johnson G, Li X, Cohen ER, Kim SG, Ugurbil K, Springer CS Jr (2007) Magnetic field and tissue dependencies of human brain longitudinal H O relaxation in vivo. Magn Reson Med Conflict of interest The authors declare that they have no competing 57:308–318 interests. 14. Pruessmann KP, Weiger M, Scheidegger MB, Boesiger P (1999) SENSE: sensitivity encoding for fast MRI. Magn Reson Med Ethical approval All procedures performed in studies involving 42:952–962 human participants were in accordance with the ethical standards of 15. Dale BM, Brown MA, Semelka RC (2015) MRI basic principles the institutional and/or national research committee and with the 1964 and applications, 5th edn. Wiley, Hoboken Helsinki declaration and its later amendments or comparable ethical 16. Yarnykh VL (2007) Actual flip-angle imaging in the pulsed steady standards. state: a method for rapid three-dimensional mapping of the transmit- ted radiofrequency field. Magn Reson Med 57:192–200 Informed consent Informed consent was obtained from all indi- 17. MacKay A, Laule C, Vavasour I, Bjarnason T, Kolind S, Mädler B vidual participants included in the study. (2006) Insights into brain microstructure from the T distribution. Magn Reson Imaging 24:515–525 18. Cox E, Gowland P (2008) Measuring T and T ′ in the brain at 2 2 Open Access This article is distributed under the terms of the 1.5 T, 3 T and 7 T using a hybrid gradient echo-spin echo sequence Creative Commons Attribution 4.0 International License (http://crea- and EPI. In: Proceedings of the 16th annual meeting of ISMRM, tivecommons.org/licenses/by/4.0/), which permits unrestricted use, Toronto, Canada, p 1411 distribution, and reproduction in any medium, provided you give appro- 19. Whittall KP, Mackay AL, Graeb DA, Nugent RA, Li DKB, Paty DW priate credit to the original author(s) and the source, provide a link to (1997) In vivo measurement of T distributions and water contents the Creative Commons license, and indicate if changes were made. in normal human brain. Magn Reson Med 37:34–43 20. Daoust A, Dodd S, Nair G, Bouraoud N, Jacobson S, Walbridge S, References Reich DS, Koretsky A (2017) Transverse relaxation of cerebrospi- nal fluid depends on glucose concentration. Magn Reson Imaging 44:72–81 1. Qin Q (2011) A simple approach for three-dimensional mapping 21. Yamada S, Ishikawa M, Yamamot K (2015) Optimal diagnostic of baseline cerebrospinal fluid volume fraction. Magn Reson Med indices for idiopathic normal pressure hydrocephalus based on the 65:385–391 3D quantitative volumetric analysis for the cerebral ventricle and 2. Iliff JJ, Wang M, Liao Y, Plogg BA, Peng W, Gundersen GA, Ben- subarachnoid space. Am J Neuroradiol 36:2262–2269 veniste H, Vates GE, Deane R, Goldman SA, Nagelhus EA, Neder- 22. Stanisz GJ, Odrobina EE, Pun J, Escaravage M, Graham SJ, Bron- gaard M (2012) A paravascular pathway facilitates CSF flow through skill MJ, Henkelman RM (2005) T, T relaxation and magnetization the brain parenchyma and the clearance of interstitial solutes, includ- 1 2 transfer in tissue at 3 T. Magn Reson Med 54:507–512 ing amyloid β. Sci Transl Med 4:147ra111 23. Visser F, Zwanenburg JJM, Hoogduin JM, Luijten PR (2010) High- 3. Iliff JJ, Nedergaard M (2013) Is there a cerebral lymphatic system? resolution magnetization-prepared 3D-FLAIR imaging at 7.0 tesla. Stroke 44:2013–2016 Magn Reson Med 64:194–202 4. Xie L, Kang H, Xu Q, Chen MJ, Liao Y, Thiyagarajan M, O’Donnell 24. Chen JJ, Pike GB (2009) Human whole blood T relaxometry at 3 J, Christensen DJ, Nicholson C, Iliff JJ, Takano T, Deane R, Ned- tesla. Magn Reson Med 61:249–254 ergaard M (2013) Sleep drives metabolite clearance from the adult 25. Krishnamurthy LC, Liu P, Xu F, Uh J, Dimitrov I, Lu H (2014) brain. Science 342:373–377 Dependence of blood T on oxygenation at 7 T: in vitro calibration 5. Spector R, Robert Snodgrass S, Johanson CE (2015) A balanced and in vivo application. Magn Reson Med 71:2035–2042 view of the cerebrospinal fluid composition and functions: focus on 26. van Veluw SJ, Fracasso A, Visser F, Spliet WGM, Luijten PR, Bies- adult humans. Exp Neurol 273:57–68 sels GJ, Zwanenburg JJM (2015) FLAIR images at 7 tesla MRI 6. De Bresser J, Brundel M, Conijn MM, Van Dillen JJ, Geerlings highlight the ependyma and the outer layers of the cerebral cortex. MI, Viergever MA, Luijten PR, Biessels GJ (2013) Visual cerebral Neuroimage 104:100–109 1 3 424 Magn Reson Mater Phy (2018) 31:415–424 27. Hopkins AL, Yeung HN, Bratton CB (1986) Multiple field strength 29. Yadav NN, Xu J, Bar-Shir A, Qin Q, Chan KWY, Grgac K, Li W, in vivo T and T for cerebrospinal fluid protons. Magn Reson Med McMahon MT, van Zijl PCM (2014) Natural d -glucose as a biode- 1 2 3:303–311 gradable MRI relaxation agent. Magn Reson Med 72:823–828 28. Yilmaz A, Ulak FŞ, Batun MS (2004) Proton T and T relaxivities 1 2 of serum proteins. Magn Reson Imaging 22:683–688 1 3
Magnetic Resonance Materials in Physics, Biology and Medicine – Springer Journals
Published: Nov 6, 2017
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
All the latest content is available, no embargo periods.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud